Design to accommodate “intelligent adaptive experiments” with future-proof hardware for deep learning-enabled imaging and neuroscience.
In other words, how to design future-proof measurement systems that are both easy to setup and are scalable for more advanced measurement paradigms of the future. And how you would like to think of structuring your data acquisition to be used efficiently with deep learning in neuroscience.
Alternative download link:
https://www.dropbox.com/s/j5r8vifvh6e7bfp/animal_instrumentation.pdf?dl=0
From traditional desktop to novel optical designs in small form factors. Towards portable low-cost fundus imaging designs with computational imaging techniques for image quality improvement.
Case for chronobiological experiments for rodents and
non-human primates.
Practical guide for selecting components for your lighting system design to be used with other behavioral measures.
Alternative download link:
https://www.dropbox.com/s/34j71dchhstvy0d/animal_lighting.pdf?dl=0
Developing an inexpensive optical measurement device to estimate ocular media density.
Selection of HDR camera, deep learning software stack for Purkinje image detection and segmentation, NVIDIA Jetson as the embedded computer triggering either LED or LASER lights to be projected on the cornea
Alternative download link:
https://www.dropbox.com/s/fh7r8szuc2pctfr/purkinje_imaging_inPractice.pdf?dl=0
Multimodal RGB-D+RF-based sensing for human movement analysisPetteriTeikariPhD
Combining RGB-D based computer vision with commodity Wifi for pose estimation and human movement analysis for action recognition.
Think of applications especially in healthcare settings, where existing Wifi Access Point already exist and adding USB Wifi dongles to Raspberry Pi (or dedicated chips) is a very easy way to create "operational awareness" of all your patients.
Alternative download link:
https://www.dropbox.com/s/awkqqfhibesjcb9/multimodal_remote_MovementSensing.pdf?dl=0
Short intro for some design considerations around hyperspectral retinal imaging. Both for research-grade desktop setups built around supercontinuum laser and AOTF tunable filter, and for mobile low-cost retinal imagers.
Available also from:
https://www.dropbox.com/s/5brchl9ntqno0i9/hyperspectral_retinal_imaging.pdf?dl=0
High-level overview of the specifications for a portable multispectral/hyperspectral fundus imaging based on the custom-developed MEMS filter developed by VTT, Finland.
Additionally one could do a multispectral purkinje imaging for quick estimation of spectral ocular media transmittance.
The device could be used for structural imaging ("classical fundus") as well as functional vessel response imaging with oximetry.
Alternative download link:
https://www.dropbox.com/s/g0ladhqw075w59j/multispectral_fundus_camera_rev2.pdf?dl=0
Practical Considerations in the design of Embedded Ophthalmic DevicesPetteriTeikariPhD
Practical level introduction for ophthalmic device design.
How in the future, multimodal measurement devices will be replacing unimodal devices with simple decision tree indices.
Some examples of embedded cameras, measurement illumination, stimulus presentation illumination are shown.
Alternative download link:
https://www.dropbox.com/s/lt76ohoeusopkoo/practicalConsiderations_embeddedOphthalmicDevices.pdf?dl=0
Novel deep learning-powered diagnostics hardware for assessing retinal health.
The impact of deep learning and artificial intelligence for the design practice itself is covered better in https://algorithms.design/ and the focus of this presentation is in the visual function diagnostics.
How is the future looking for your high-street optician's (e.g. Specsavers, Boots) vision exam going beyond simple refraction correction, and how possibly in the future AR glasses could allow design of "smarter" every-day eyewear also for health monitoring.
Talk given for “Future of Eyecare: How we see and how we want to be seen” organized by Flora McLean.
Royal College of Art - London UK
From traditional desktop to novel optical designs in small form factors. Towards portable low-cost fundus imaging designs with computational imaging techniques for image quality improvement.
Case for chronobiological experiments for rodents and
non-human primates.
Practical guide for selecting components for your lighting system design to be used with other behavioral measures.
Alternative download link:
https://www.dropbox.com/s/34j71dchhstvy0d/animal_lighting.pdf?dl=0
Developing an inexpensive optical measurement device to estimate ocular media density.
Selection of HDR camera, deep learning software stack for Purkinje image detection and segmentation, NVIDIA Jetson as the embedded computer triggering either LED or LASER lights to be projected on the cornea
Alternative download link:
https://www.dropbox.com/s/fh7r8szuc2pctfr/purkinje_imaging_inPractice.pdf?dl=0
Multimodal RGB-D+RF-based sensing for human movement analysisPetteriTeikariPhD
Combining RGB-D based computer vision with commodity Wifi for pose estimation and human movement analysis for action recognition.
Think of applications especially in healthcare settings, where existing Wifi Access Point already exist and adding USB Wifi dongles to Raspberry Pi (or dedicated chips) is a very easy way to create "operational awareness" of all your patients.
Alternative download link:
https://www.dropbox.com/s/awkqqfhibesjcb9/multimodal_remote_MovementSensing.pdf?dl=0
Short intro for some design considerations around hyperspectral retinal imaging. Both for research-grade desktop setups built around supercontinuum laser and AOTF tunable filter, and for mobile low-cost retinal imagers.
Available also from:
https://www.dropbox.com/s/5brchl9ntqno0i9/hyperspectral_retinal_imaging.pdf?dl=0
High-level overview of the specifications for a portable multispectral/hyperspectral fundus imaging based on the custom-developed MEMS filter developed by VTT, Finland.
Additionally one could do a multispectral purkinje imaging for quick estimation of spectral ocular media transmittance.
The device could be used for structural imaging ("classical fundus") as well as functional vessel response imaging with oximetry.
Alternative download link:
https://www.dropbox.com/s/g0ladhqw075w59j/multispectral_fundus_camera_rev2.pdf?dl=0
Practical Considerations in the design of Embedded Ophthalmic DevicesPetteriTeikariPhD
Practical level introduction for ophthalmic device design.
How in the future, multimodal measurement devices will be replacing unimodal devices with simple decision tree indices.
Some examples of embedded cameras, measurement illumination, stimulus presentation illumination are shown.
Alternative download link:
https://www.dropbox.com/s/lt76ohoeusopkoo/practicalConsiderations_embeddedOphthalmicDevices.pdf?dl=0
Novel deep learning-powered diagnostics hardware for assessing retinal health.
The impact of deep learning and artificial intelligence for the design practice itself is covered better in https://algorithms.design/ and the focus of this presentation is in the visual function diagnostics.
How is the future looking for your high-street optician's (e.g. Specsavers, Boots) vision exam going beyond simple refraction correction, and how possibly in the future AR glasses could allow design of "smarter" every-day eyewear also for health monitoring.
Talk given for “Future of Eyecare: How we see and how we want to be seen” organized by Flora McLean.
Royal College of Art - London UK
From unimodal image classification to integrative multimodal deep learning pipelines in disease classification, disease management and predictive personalised healthcare.
Purkinje imaging for crystalline lens density measurementPetteriTeikariPhD
Brief introduction for the non-invasive, inexpensive and fast Purkinje image -based method for measuring the spectral transmittance of the human crystalline lens density in vivo.
Alternative download link:
https://www.dropbox.com/s/588y7epy13n34xo/purkinje_imaging.pdf?dl=0
Using off-the-shelf ultrasound imagers, and transition to portable system-on-chip ultrasound imagers such as Butterfly IQ.
Embedded devices such as Butterfly IQ can be further improved by integrating deep learning / artificial intelligence at device level, and naturally at the post-processing and analysis levels
Alternative download link:
https://www.dropbox.com/s/rlwv7m29mh6y2w6/pupillometry_throughTheEyelids.pdf?dl=0
Time-resolved biomedical sensing through scattering mediumPetteriTeikariPhD
Time-resolved biomedical sensing through scattering medium | Case study with pupillometry through closed eyelids for neurological monitoring
Download link: https://www.dropbox.com/s/x0f5q6cz5ax33s4/timeResolvedSensing.pdf?dl=0
Beyond Broken Stick Modeling: R Tutorial for interpretable multivariate analysisPetteriTeikariPhD
“R Tutorial” for Interpretable multivariate analysis with t-SNE and Random Forests mainly for ophthalmic data modeling.
Bust through the fetish for indices and easy scalar human-readable interpretations of data.
Alternative download link:
https://www.dropbox.com/s/wyg5k0k35qxdcyx/beyond_brokenStick.pdf?dl=0
Possible future avenues for ophthalmic imaging combining advanced techniques and deep learning. "Bubbling under the surface, and inspiration from ‘bioimaging’ in general"
Hardware landscape from computer vision to wearable sensors, and a light intro for UX requirements to ensure adherence and engagement.
At the intersection of new sensors, big data, deep learning, gamification, behavioral medicine and human factors.
Applications benefiting from "quantitative sensorimotor training", "precision exercise", "precision physiotherapy" or whatever you are calling this, include weight and strength training, powerlifting, bodybuilding, martial arts, yoga, dance, musical instrument training, post-surgery rehabilitation for ACL tears, etc.
Alternative download link:
https://www.dropbox.com/s/wcfrzdjkn58xjdq/physio_pipeline_hw.pdf?dl=0
Instead of talking about artificial intelligence at the organizational level in hospitals and in research laboratories, the focus for non-machine learning practitioner should be on understanding the data pipes and what is involved around the model training.
alternative download link:
https://www.dropbox.com/s/9tv673sxkxcnojj/dataStrategyForOphthalmology.pdf?dl=0
Shallow introduction for Deep Learning Retinal Image AnalysisPetteriTeikariPhD
Overview of retinal imaging techniques such as fundus photography, optical coherence tomography (OCT) along with future upgrades such as multispectral imaging, OCT angiography, adaptive optics imaging and polarization-sensitive OCT. This is followed by an overview of deep learning image analysis methods suitable to be used with retinal imaging techniques.
Alternative download link: https://www.dropbox.com/s/n01w02cjaf68vbo/retina_deepLearning_pipeline.pdf?dl=0
Deep Learning for Biomedical Unstructured Time SeriesPetteriTeikariPhD
1D Convolutional neural networks (CNNs) for time series analysis, and inspiration from beyond biomedical field. Short intro for various different steps involved in Time Series Analysis including outlier detection, imputation, denoising, segmentation, classification and forecasting.
Available also from:
https://www.dropbox.com/s/cql2jhrt5mdyxne/timeSeries_deepLearning.pdf?dl=0
Pre-processing with outlier detection,
denoising and analysis using deep learning
Alternative download link:
https://www.dropbox.com/s/nvp70vyslvggte1/PLR_overview.pdf?dl=0
Overview of the emerging field of smartphone-powered ophthalmic diagnostics that has the potential to bring down the cost and improve access to healthcare especially in developing countries.
Alternative download link: https://www.dropbox.com/s/c3ef13nlw2mywa7/iphoneFundusCamera.pdf?dl=0
High-level concepts For applications such as:
1) Myopia, 2) Jetlag, 3) Seasonal Affective Disorder (SAD)
If you want to add some tech to eyewear / glasses / sunglasses design projects, this slideshow serves as a high-level introduction for technical details
Alternative download link:
https://www.dropbox.com/s/qe8dpji6gwh1s8v/lightTreatmentGlasses_concepts.pdf?dl=0
Using physics-based OCT Monte Carlo simulation and wave optics models for synthesising new OCT volumes for ophthalmic deep learning.
Alternative download link:
https://www.dropbox.com/s/ax15qy47yi76eex/OCT_MonteCarlo.pdf?dl=0
Image restoration techniques covered such as denoising, deblurring and super-resolution for 3D images and models.
From classical computer vision techniques to contemporary deep learning based processing for both ordered and unordered point clouds, depth maps and meshes.
Tracking times in temporal patterns embodied in intra-cortical data for cont...IJECEIAES
Brain-machines capture brain signals in order to restore communication and movement to disabled people who suffer from brain palsy or motor disorders. In brain regions, the ensemble firing of populations of neurons represents spatio-temporal patterns that are transformed into outgoing spatio-temporal patterns which encode complex cognitive task. This transformation is dynamic, non-stationary (time-varying) and highly nonlinear. Hence, modeling such complex biological patterns requires specific model structures to uncover the underlying physiological mechanisms and their influences on system behavior. In this study, a recent multi-electrode technology allows the record of the simultaneous neuron activities in behaving animals. Intra-cortical data are processed according to these steps: spike detection and sorting, than desired action extraction from the rate of the obtained signal. We focus on the following important questions about (i) the possibility of linking the brain signal time events with some time-delayed mapping tools; (ii) the use of some suitable inputs than others for the decoder; (iii) a consideration of separated data or a special representation founded on multi-dimensional statistics. This paper concentrates mostly on the analysis of parallel spike train when certain critical hypotheses are ignored by the data for the working method. We have made efforts to define explicitly whether the underlying hypotheses are actually achieved. In this paper, we propose an algorithm to define the embedded memory order of NARX recurrent neural networks to the hand trajectory tracking process. We also demonstrate that this algorithm can improve performance on inference tasks.
From unimodal image classification to integrative multimodal deep learning pipelines in disease classification, disease management and predictive personalised healthcare.
Purkinje imaging for crystalline lens density measurementPetteriTeikariPhD
Brief introduction for the non-invasive, inexpensive and fast Purkinje image -based method for measuring the spectral transmittance of the human crystalline lens density in vivo.
Alternative download link:
https://www.dropbox.com/s/588y7epy13n34xo/purkinje_imaging.pdf?dl=0
Using off-the-shelf ultrasound imagers, and transition to portable system-on-chip ultrasound imagers such as Butterfly IQ.
Embedded devices such as Butterfly IQ can be further improved by integrating deep learning / artificial intelligence at device level, and naturally at the post-processing and analysis levels
Alternative download link:
https://www.dropbox.com/s/rlwv7m29mh6y2w6/pupillometry_throughTheEyelids.pdf?dl=0
Time-resolved biomedical sensing through scattering mediumPetteriTeikariPhD
Time-resolved biomedical sensing through scattering medium | Case study with pupillometry through closed eyelids for neurological monitoring
Download link: https://www.dropbox.com/s/x0f5q6cz5ax33s4/timeResolvedSensing.pdf?dl=0
Beyond Broken Stick Modeling: R Tutorial for interpretable multivariate analysisPetteriTeikariPhD
“R Tutorial” for Interpretable multivariate analysis with t-SNE and Random Forests mainly for ophthalmic data modeling.
Bust through the fetish for indices and easy scalar human-readable interpretations of data.
Alternative download link:
https://www.dropbox.com/s/wyg5k0k35qxdcyx/beyond_brokenStick.pdf?dl=0
Possible future avenues for ophthalmic imaging combining advanced techniques and deep learning. "Bubbling under the surface, and inspiration from ‘bioimaging’ in general"
Hardware landscape from computer vision to wearable sensors, and a light intro for UX requirements to ensure adherence and engagement.
At the intersection of new sensors, big data, deep learning, gamification, behavioral medicine and human factors.
Applications benefiting from "quantitative sensorimotor training", "precision exercise", "precision physiotherapy" or whatever you are calling this, include weight and strength training, powerlifting, bodybuilding, martial arts, yoga, dance, musical instrument training, post-surgery rehabilitation for ACL tears, etc.
Alternative download link:
https://www.dropbox.com/s/wcfrzdjkn58xjdq/physio_pipeline_hw.pdf?dl=0
Instead of talking about artificial intelligence at the organizational level in hospitals and in research laboratories, the focus for non-machine learning practitioner should be on understanding the data pipes and what is involved around the model training.
alternative download link:
https://www.dropbox.com/s/9tv673sxkxcnojj/dataStrategyForOphthalmology.pdf?dl=0
Shallow introduction for Deep Learning Retinal Image AnalysisPetteriTeikariPhD
Overview of retinal imaging techniques such as fundus photography, optical coherence tomography (OCT) along with future upgrades such as multispectral imaging, OCT angiography, adaptive optics imaging and polarization-sensitive OCT. This is followed by an overview of deep learning image analysis methods suitable to be used with retinal imaging techniques.
Alternative download link: https://www.dropbox.com/s/n01w02cjaf68vbo/retina_deepLearning_pipeline.pdf?dl=0
Deep Learning for Biomedical Unstructured Time SeriesPetteriTeikariPhD
1D Convolutional neural networks (CNNs) for time series analysis, and inspiration from beyond biomedical field. Short intro for various different steps involved in Time Series Analysis including outlier detection, imputation, denoising, segmentation, classification and forecasting.
Available also from:
https://www.dropbox.com/s/cql2jhrt5mdyxne/timeSeries_deepLearning.pdf?dl=0
Pre-processing with outlier detection,
denoising and analysis using deep learning
Alternative download link:
https://www.dropbox.com/s/nvp70vyslvggte1/PLR_overview.pdf?dl=0
Overview of the emerging field of smartphone-powered ophthalmic diagnostics that has the potential to bring down the cost and improve access to healthcare especially in developing countries.
Alternative download link: https://www.dropbox.com/s/c3ef13nlw2mywa7/iphoneFundusCamera.pdf?dl=0
High-level concepts For applications such as:
1) Myopia, 2) Jetlag, 3) Seasonal Affective Disorder (SAD)
If you want to add some tech to eyewear / glasses / sunglasses design projects, this slideshow serves as a high-level introduction for technical details
Alternative download link:
https://www.dropbox.com/s/qe8dpji6gwh1s8v/lightTreatmentGlasses_concepts.pdf?dl=0
Using physics-based OCT Monte Carlo simulation and wave optics models for synthesising new OCT volumes for ophthalmic deep learning.
Alternative download link:
https://www.dropbox.com/s/ax15qy47yi76eex/OCT_MonteCarlo.pdf?dl=0
Image restoration techniques covered such as denoising, deblurring and super-resolution for 3D images and models.
From classical computer vision techniques to contemporary deep learning based processing for both ordered and unordered point clouds, depth maps and meshes.
Tracking times in temporal patterns embodied in intra-cortical data for cont...IJECEIAES
Brain-machines capture brain signals in order to restore communication and movement to disabled people who suffer from brain palsy or motor disorders. In brain regions, the ensemble firing of populations of neurons represents spatio-temporal patterns that are transformed into outgoing spatio-temporal patterns which encode complex cognitive task. This transformation is dynamic, non-stationary (time-varying) and highly nonlinear. Hence, modeling such complex biological patterns requires specific model structures to uncover the underlying physiological mechanisms and their influences on system behavior. In this study, a recent multi-electrode technology allows the record of the simultaneous neuron activities in behaving animals. Intra-cortical data are processed according to these steps: spike detection and sorting, than desired action extraction from the rate of the obtained signal. We focus on the following important questions about (i) the possibility of linking the brain signal time events with some time-delayed mapping tools; (ii) the use of some suitable inputs than others for the decoder; (iii) a consideration of separated data or a special representation founded on multi-dimensional statistics. This paper concentrates mostly on the analysis of parallel spike train when certain critical hypotheses are ignored by the data for the working method. We have made efforts to define explicitly whether the underlying hypotheses are actually achieved. In this paper, we propose an algorithm to define the embedded memory order of NARX recurrent neural networks to the hand trajectory tracking process. We also demonstrate that this algorithm can improve performance on inference tasks.
Finger Vein Recognition Based on PCA Feature using Artificial Neural Networkrahulmonikasharma
Personal recognition technology is developing rapidly as a security system. Traditional methods such as authentication key; password: card is not secure enough, because they could be stolen or easily forget. Biometrics has been applied to a wide range of systems. According to various researchers, vein biometrics was a good technique from other biometric authentication system used, such as fingerprints, hand geometry, voice, etc. of the DNA. Root Authentication systems can be designed in different ways. All methods include the matching stage. A neural network is an effective way of matching Personal identification authentication system. The finger vein pattern is unique biometric identity of the human beings. The finger vein recognition is a popular biometric technique which is used for authentication purposes in various applications. In the propose work an algorithm is proposed to find the accuracy, FRR and FAR of finger vein recognition. The performances of PCA, threshold segmentation, pre-processing and testing & training techniques has been validate and compared with each other in order to determine the most accurate results in terms of finger vein recognition.
MOTION PREDICTION USING DEPTH INFORMATION OF HUMAN ARM BASED ON ALEXNETgerogepatton
The development of convolutional neural networks(CNN) has provided a new tool to make classification and prediction of human's body motion. This project tends to predict the drop point of a ball thrown out by experimenters by classifying the motion of their body in the process of throwing. Kinect sensor v2 is used to record depth maps and the drop points are recorded by a square infrared induction module. Firstly, convolutional neural networks are made use of to put the data obtained from depth maps in and get the prediction of drop point according to experimenters' motion. Secondly, huge amount of data is used to trainthe networks of different structure, and a network structure that could provide high enough accuracy for drop point prediction is established. The network model and parameters are modified to improve the accuracy of the prediction algorithm. Finally, the experimental data is divided into a training group and a test group. The prediction results of test group reflect that the prediction algorithm effectively improves the accuracy of human motion perception.
Motion Prediction Using Depth Information of Human Arm Based on Alexnetgerogepatton
The development of convolutional neural networks(CNN) has provided a new tool to make classification
and prediction of human's body motion. This project tends to predict the drop point of a ball thrown out by
experimenters by classifying the motion of their body in the process of throwing. Kinect sensor v2 is used to
record depth maps and the drop points are recorded by a square infrared induction module. Firstly,
convolutional neural networks are made use of to put the data obtained from depth maps in and get the
prediction of drop point according to experimenters' motion. Secondly, huge amount of data is used to train
the networks of different structure, and a network structure that could provide high enough accuracy for
drop point prediction is established. The network model and parameters are modified to improve the
accuracy of the prediction algorithm. Finally, the experimental data is divided into a training group and a
test group. The prediction results of test group reflect that the prediction algorithm effectively improves the
accuracy of human motion perception.
Application of Artificial Neural Networking for Determining the Plane of Vibr...IOSRJMCE
In this paper a new approach for Artificial Neural Networking using Feed Forward Back Propagation Method and Levenberg-Marquardt backpropagation training function has been developed using Java Programming, where by directly feeding the RMS and Phase values of vibration, the unbalance plane can be detected with minimum error. In a Machine Fault Simulator RMS value and phase values of vibrations are collected from the four accelerometers placed in X and Y direction of Left and Right Bearings .Further these data are fed into the neural network for training purpose. In the testing phase of the neural network, the plane of vibration has been determined using different training algorithms available in MATLAB. Their prediction values have been compared with the actual value, errors for different training algorithms are calculated and a conclusion has been drawn for the best training function available for this current research work.
Similar to Instrumentation for in vivo intravital microscopy (20)
From lung/heart/ambient source separation to clinical unimodal
classification
Alternative download link:
https://www.dropbox.com/scl/fi/8s7uq4h0fi8lgqbzqwg83/wearableMic_signal.pdf?rlkey=l2tqg5yffd4e0w224g3cs6pfl&dl=0
Next Gen Ophthalmic Imaging for Neurodegenerative Diseases and OculomicsPetteriTeikariPhD
Shallow literature analysis on recent trends in (multimodal) ophthalmic imaging with focus on neurodegenerative disease imaging / oculomics. Open-ended literature review on what you could be building next.
#1/2: Hardware
#2/2: Computational imaging (coming)
Alternative download link:
https://www.dropbox.com/scl/fi/ebp5xkhm3ngfu80hw0lvo/retina_imaging_2024.pdf?rlkey=eeikf3ewxdb481v06wxm34mqu&dl=0
Next Gen Computational Ophthalmic Imaging for Neurodegenerative Diseases and ...PetteriTeikariPhD
Shallow literature analysis on recent trends in computational ophthalmic imaging with focus on neurodegenerative disease imaging / oculomics.
Open-ended literature review on what you could be building next.
#1/2: Hardware
#2/2: Computational imaging
Alternative download link:
https://www.dropbox.com/scl/fi/d34pgi3xopfjbrcqj2lvi/retina_imaging_2024_computational.pdf?rlkey=xnt1dbe8rafyowocl9cbgjh3p&dl=0
Skin temperature as a proxy for core body temperature (CBT) and circadian phasePetteriTeikariPhD
Using distal temperature (wrist temperature with smartwatch / finger temperature with smart ring as Oura) to estimate core body temperature (CBT).
We can then use the wrist temperature shifts as circadian phase shift estimates in circadian phase management. For example when prescribing melatonin or/and light exposure to mitigate the effects of jet lag
Alternative download link:
https://www.dropbox.com/scl/fi/es7174291yws262rhr568/cbt_estimation.pdf?rlkey=846yeed1wrqsjgkx7kp8ccc2y&dl=0
Summary of "Precision strength training: The future of strength training with...PetteriTeikariPhD
Short visual summary of the preprint:
Petteri Teikari and Aleksandra Pietrusz (2021)
“Precision Strength Training: Data-driven Artificial
Intelligence Approach to Strength and Conditioning.”
SportRxiv. May 20. https://doi.org/10.31236/osf.io/w734a
Precision strength training: The future of strength training with data-driven...PetteriTeikariPhD
Visual presentation of the preprint:
Petteri Teikari and Aleksandra Pietrusz (2021)
“Precision Strength Training: Data-driven Artificial
Intelligence Approach to Strength and Conditioning.”
SportRxiv. May 20. https://doi.org/10.31236/osf.io/w734a
Alternative download link:
https://www.dropbox.com/scl/fi/47nqp579t1b4m1zs0irhw/precision_strength_training.pdf?rlkey=05mzzw2ep8id71mq86936hvfi&dl=0
Intracerebral Hemorrhage (ICH): Understanding the CT imaging featuresPetteriTeikariPhD
Overview of CT basics and deep learning literature mostly focused on the analysis of ICH.
Intracerebral hemorrhage (ICH), also known as cerebral bleed, is a type of intracranial bleed that occurs within the brain tissue or ventricles. Intracerebral bleeds are the second most common cause of stroke, accounting for 10% of hospital admissions for stroke.
For spontaneous ICH seen on CT scan, the death rate (mortality) is 34–50% by 30 days after the insult,and half of the deaths occur in the first 2 days. Even though the majority of deaths occurs in the first days after ICH, survivors have a long term excess mortality of 27% compared to the general population.
Deep learning and computational steps roughly can be categorized to 1) Preprocessing, 2) Image Restoration (denoising, deblurring, inpainting, reconstruction), 3) Diffeomorphic registration for spatial normalization, 4) Hand-crafted radiomics and texture analysis, 5) Hemorrhage segmentation, among other relevant head CT issues
Alternative download link: https://www.dropbox.com/s/8l2h93cl2pmle4g/CT_hemorrhage.pdf?dl=0
Clinical applications with a focus on rheumatoid arthritis (RA) management. Quick overview of hand pose tracking for managing rheumatoid arthritis.
For best clinical outcome, you might want to think how to integrate additional modalities like surface electromyography (sEMG) and hand function assessments (like hand grip strength, and finger extension strength) to the clinical prognostics model.
Alternative download link:
https://www.dropbox.com/s/rexzt3d5tsm1vgc/hand_tracking_arthritis_management.pdf?dl=0
Creativity as Science: What designers can learn from science and technologyPetteriTeikariPhD
What personality traits do creative people share? Is creativity skill like any other? Is creativity suppressed in our world, is creativity misunderstood by "dinosaur companies" stuck with their legacy systems? Are "creatives" actually that creative in the end? Can fashion design exist in some romantic old school silo where no tech understanding is needed?
Alternative download link:
https://www.dropbox.com/s/ghiyeo3nyrtutzt/RCA_creativity.pdf?dl=0
Outlining the common challenges encountered when structuring clinical and research datasets for deep learning training.
Typically the datasets are so unstructured that they are impossible to analyze by any deep learning practitioners. And the cleaning and data wrangling ends up taking most of the time which could have been planned properly even before the clinical data acquisition.
One could argue that especially for medical data, the annotated data is the new gold, and not just the Big Data scattered all over the place. This is practice translates to efforts to design as intelligent as possible data labelling pipelines for efficient use of expert clinician annotation work.
Alternative download link:
https://www.dropbox.com/s/bbgc21yc86h0t14/Efficient_Ocular_Data_Labelling.pdf?dl=0
Interactive business intelligence visualizations with R Shiny and beyond with scalable big data architectures. Going beyond MS Excel and other non-scalable proprietary solutions.
Practical steps for non-machine learners on how to prepare your medical image dataset for deep learning modelling.
Here we use a fundus image dataset as an example that might have controls (healthy eyes) and glaucomatous fundus images with three different severities. In glaucoma, the optic disc is of a special interest so we want to annotate that from the images using a bounding box to help the deep learning training.
DevOps and Testing slides at DASA ConnectKari Kakkonen
My and Rik Marselis slides at 30.5.2024 DASA Connect conference. We discuss about what is testing, then what is agile testing and finally what is Testing in DevOps. Finally we had lovely workshop with the participants trying to find out different ways to think about quality and testing in different parts of the DevOps infinity loop.
Accelerate your Kubernetes clusters with Varnish CachingThijs Feryn
A presentation about the usage and availability of Varnish on Kubernetes. This talk explores the capabilities of Varnish caching and shows how to use the Varnish Helm chart to deploy it to Kubernetes.
This presentation was delivered at K8SUG Singapore. See https://feryn.eu/presentations/accelerate-your-kubernetes-clusters-with-varnish-caching-k8sug-singapore-28-2024 for more details.
Essentials of Automations: Optimizing FME Workflows with ParametersSafe Software
Are you looking to streamline your workflows and boost your projects’ efficiency? Do you find yourself searching for ways to add flexibility and control over your FME workflows? If so, you’re in the right place.
Join us for an insightful dive into the world of FME parameters, a critical element in optimizing workflow efficiency. This webinar marks the beginning of our three-part “Essentials of Automation” series. This first webinar is designed to equip you with the knowledge and skills to utilize parameters effectively: enhancing the flexibility, maintainability, and user control of your FME projects.
Here’s what you’ll gain:
- Essentials of FME Parameters: Understand the pivotal role of parameters, including Reader/Writer, Transformer, User, and FME Flow categories. Discover how they are the key to unlocking automation and optimization within your workflows.
- Practical Applications in FME Form: Delve into key user parameter types including choice, connections, and file URLs. Allow users to control how a workflow runs, making your workflows more reusable. Learn to import values and deliver the best user experience for your workflows while enhancing accuracy.
- Optimization Strategies in FME Flow: Explore the creation and strategic deployment of parameters in FME Flow, including the use of deployment and geometry parameters, to maximize workflow efficiency.
- Pro Tips for Success: Gain insights on parameterizing connections and leveraging new features like Conditional Visibility for clarity and simplicity.
We’ll wrap up with a glimpse into future webinars, followed by a Q&A session to address your specific questions surrounding this topic.
Don’t miss this opportunity to elevate your FME expertise and drive your projects to new heights of efficiency.
GraphRAG is All You need? LLM & Knowledge GraphGuy Korland
Guy Korland, CEO and Co-founder of FalkorDB, will review two articles on the integration of language models with knowledge graphs.
1. Unifying Large Language Models and Knowledge Graphs: A Roadmap.
https://arxiv.org/abs/2306.08302
2. Microsoft Research's GraphRAG paper and a review paper on various uses of knowledge graphs:
https://www.microsoft.com/en-us/research/blog/graphrag-unlocking-llm-discovery-on-narrative-private-data/
Elevating Tactical DDD Patterns Through Object CalisthenicsDorra BARTAGUIZ
After immersing yourself in the blue book and its red counterpart, attending DDD-focused conferences, and applying tactical patterns, you're left with a crucial question: How do I ensure my design is effective? Tactical patterns within Domain-Driven Design (DDD) serve as guiding principles for creating clear and manageable domain models. However, achieving success with these patterns requires additional guidance. Interestingly, we've observed that a set of constraints initially designed for training purposes remarkably aligns with effective pattern implementation, offering a more ‘mechanical’ approach. Let's explore together how Object Calisthenics can elevate the design of your tactical DDD patterns, offering concrete help for those venturing into DDD for the first time!
Builder.ai Founder Sachin Dev Duggal's Strategic Approach to Create an Innova...Ramesh Iyer
In today's fast-changing business world, Companies that adapt and embrace new ideas often need help to keep up with the competition. However, fostering a culture of innovation takes much work. It takes vision, leadership and willingness to take risks in the right proportion. Sachin Dev Duggal, co-founder of Builder.ai, has perfected the art of this balance, creating a company culture where creativity and growth are nurtured at each stage.
Connector Corner: Automate dynamic content and events by pushing a buttonDianaGray10
Here is something new! In our next Connector Corner webinar, we will demonstrate how you can use a single workflow to:
Create a campaign using Mailchimp with merge tags/fields
Send an interactive Slack channel message (using buttons)
Have the message received by managers and peers along with a test email for review
But there’s more:
In a second workflow supporting the same use case, you’ll see:
Your campaign sent to target colleagues for approval
If the “Approve” button is clicked, a Jira/Zendesk ticket is created for the marketing design team
But—if the “Reject” button is pushed, colleagues will be alerted via Slack message
Join us to learn more about this new, human-in-the-loop capability, brought to you by Integration Service connectors.
And...
Speakers:
Akshay Agnihotri, Product Manager
Charlie Greenberg, Host
UiPath Test Automation using UiPath Test Suite series, part 3DianaGray10
Welcome to UiPath Test Automation using UiPath Test Suite series part 3. In this session, we will cover desktop automation along with UI automation.
Topics covered:
UI automation Introduction,
UI automation Sample
Desktop automation flow
Pradeep Chinnala, Senior Consultant Automation Developer @WonderBotz and UiPath MVP
Deepak Rai, Automation Practice Lead, Boundaryless Group and UiPath MVP
1. Instrumentation
for in vivo intravital
microscopy
Design to accommodate
“intelligent adaptive
experiments” with future-proof
hardware for deep learning-
enabled imaging and neuroscience
Petteri Teikari, PhD
Singapore Eye Research Institute (SERI)
Visual Neurosciences group
http://petteri-teikari.com/
Version “Mon 5 November 2018“
Figuresfrom
https://doi.org/10.1186/1752-0509-2-74
https://doi.org/10.3389/fphys.2015.00147
https://www.nikonsmallworld.com/people/wim-va
n-egmond
3. DAQDataAcQuisition System
System that converts your analog signal (e.g. EEG, ECG,
temperature, blood pressure, etc.) to a digital signal
stored ona computer
NationalInstruments
DAQs
http://www.ni.com/data-acqui
sition/
Worksbestwiththe LabVIEW
(developedbyNational
Instruments)virtual
instrumentation software
LabJackDAQs
https://labjack.com/products/c
omparison-table
Worksforexamplewith
PsychoPy(Python)ifyouare
running behavioral
experiments.
BitScope and Raspberry Pi
http://www.bitscope.com/blog/DI/?p=DI25A
BitScopecan capturemultipleanalog and digital
signalsatveryhigh samplerates(up to 40MSpsin
somecases)withoutloading theRaspberryPi CPU
or requiring areal-timeoperating systemfor low
jittersampling.
9. Input–sensor /
Output– actuator,trigger,stimulus,etc.
inLight/ Circadian
studiesyou wantto
makesurethat
luminance
distribution,
intensityand
spectralcontent
arethedesired
http://dx.doi.org/10.12688/wellcomeopenres.9892.2
COMPASS: Continuous Open Mouse Phenotyping of Activity and Sleep Status
“Finally the authors would like to thank the open-source communities
connected to Arduino, Processing, Python/PyData stack and Blender for the
toolsusedto illustratethemethodsinthispaper.”
12. Input–sensor /
Output– actuator,trigger,stimulus,etc.
inLight/ Circadian
studiesyou wantto
makesurethat
luminance
distribution,
intensityand
spectralcontent
arethedesired
DemoKit for MAS
AS726xSpectral
sensing
https://ams.com/as726xdemokit
Multi-spectral colour sensoroptimisedfor
blue-light well-being
AMS has created a full-colour sensor AS7264N
that matches eye response with RGB, adds special
blue sensors for 440nm and 490nm, and another
fornear-infra-red.
“The sensor also accurately measures blue-light
wavelengths, which researchers have linked to
important health effects such as disruption or
management of the circadian rhythm, accelerated
eyeaging, andeyestrain.”
14. Input–sensor /
Output– actuator,trigger,stimulus,etc.
Pose&Skeleton
estimation
Usefulformany
applicationsthen
Markerlesstrackingofuser-defined featureswithdeeplearning
AlexanderMathis, PranavMamidanna, TaigaAbe, KevinM.Cury, VenkateshN.Murthy, MackenzieW.Mathis, MatthiasBethge(
Submittedon9 Apr 2018) | https://arxiv.org/abs/1804.03142
We demonstrate the versatility of this framework by tracking various body parts in a
broad collection of experimental settings: mice odor trail-tracking, egg-laying behavior in
drosophila, and mouse hand articulation in a skilled forelimb task. For example, during
the skilled reaching behavior, individual joints can be automatically tracked (and a
confidence score isreported).
TheRolesof Supervised
MachineLearningin
SystemsNeuroscience
https://arxiv.org/abs/1805.08239
Automated leg tracking reveals distinct
conserved gait and tremor signatures
in Drosophila models of Parkinson's
DiseaseandSpinocerebellarataxia3
https://doi.org/10.1101/425405
Different mutationsproduced tremorsindistinct legpairs,
indicatingthat differentmotorcircuitsareaffected. Almost
190,000videoframes weretrackedin thisstudy,
allowing, forthefirsttime,high-throughputanalysisofgait
andtremorfeaturesin Drosophilamutants.As an efficient
assayofmutantgaitand tremorfeaturesinanimportant
modelsystem,FLLITwillenabletheanalysisofthe
neurogenetic mechanismsthat underliemovement
disorders.
18. ExampleoftheDAQsystemwithabitofintelligence
Newtechniquesformotion-artifact-freeinvivocardiac
microscopy
Claudio Vinegoni,SungonLee,AaronD.AguirreandRalphWeissleder
Centerfor SystemsBiology,MassachusettsGeneral HospitalandHarvardMedicalSchool,Boston,MA,USA
Front. Physiol., 12 May2015 https://doi.org/10.3389/fphys.2015.00147
Scheme of principle for motion compensation in laser scanning microscopy (LSM). (A)
DAQ, data acquisition card; ECG, electro-cardiogram; V, mechanical ventilator. (B)
Time-gated windows, coincident with the time window corresponding to the end-
diastole, are isolated in the recorded ECG. (C) In LSM images are acquired pixel by pixel
intherealspace
Scheme of principle and timing diagram for retrospectively double gated
(cardiac and respiratory) sequential segmented laser scanning microscopy. Due to
the combined effect of cardiac and respiratory motion, segments from raw images need
to be chosen in correspondence to atime-gated window, which is the intersection of two
distinct temporal windows present in the ECG and the ventilator pressure diagram.
Adapted from Lee et al. (2012a).
19. ”Simplegating” mightnotbeenough for sharp images
All-opticalmicroscopeautofocusbasedonanelectricallytunable
lensandatotallyinternallyreflectedIRlaser
M.Bathe-Peters,P.Annibale,andM.J.Lohse
OpticsExpressVol. 26, Issue 3, pp. 2359-2368 (2018)
https://doi.org/10.1364/OE.26.002359
Active motion stabilization removes relative movement between the imaging device and the imaged tissue by active motion of the objective lens
and tracking of the imaged tissue, leading to motion-free images. (A) A high-speed camera with 955 fps was utilized to track the movement
of the tissue, and a piezoactuator-driven positioner was designed for precise and fast movement of the objective lens. Adapted from
Leeetal. (2008). (B) A contact-type sensor consisting of three cantilevers beams with strain gauges was designed to measure the three
dimensional movement of the tissue instead of the previous high-speed camera. This sensor also works as a passive stabilizer, reducing the
movementwithsoftpressure.
“Active
motion
stabilization”
20. ElectricalTunable Lens(ETL) for auto-focusing
All-opticalmicroscopeautofocusbasedonanelectrically
tunablelensandatotallyinternallyreflectedIRlaser
M.Bathe-Peters,P.Annibale,andM.J.Lohse
OpticsExpressVol. 26, Issue 3, pp. 2359-2368 (2018)
https://doi.org/10.1364/OE.26.002359
We propose here a truly all-optical microscope
autofocus taking advantage of an electrically
tunable lens (ETL, Optotune EL-16-40-TC) and a
totally internally reflected infrared probe beam. We
implement a feedback-loop based on the lateral
position of a totally internally reflected infrared laser
on a quadrant photodetector, as an indicator of the
relativedefocus.
We show here how to treat the combined
contributions due to mechanical defocus and
deformation of the tunable lens. As a result, the
sample can be keptin focus withoutany mechanical
movement, at rates up to hundreds of Hertz. The
device requires only reflective optics and can be
implemented at a fraction of the cost required for
acomparablepiezo-basedactuator.
The manufacturer (Optotune) discusses only coma as a possible geometry induced
aberration in their lenses. In our hands, the dominant aberration observed was
astigmatism
23. SAIIProvidesacompletekitto imaginggating
SAII SmallAnimalInstruments,Inc.
Model1035MR-compatibleMonitor forVeterinary use
MONITORING
●
Fiber OpticECG
●
Respiration
●
Fiber OpticTemperature
●
Fiber OpticPulse
Oximetry
●
Non-InvasiveBlood
Pressure
●
InvasiveBloodPressure
●
Capnography
GATING
●
ECG
●
Respiratory
●
ECG&Respiratory
●
AuxiliaryInputs
Options include a fluid heating system which can
regulate the temperature of the animal and invasive
blood pressure measuring the cardiac waveform,
heart rate, systolic, diastolic and mean arterial
pressure..
TTL IN
ForFluoviewIOBox
*TTL
Transistor–transistor logic
HIGH/ LOW whentoimage
24. Optimally, youwould haveboth digital and analog output
TTLIN
ForFluoviewIOBox
*TTL
Transistor–transistor logic
HIGH/ LOW when toimage
ANALOGIN(s)
ForFluoviewIO
box
Usethe digitalinput duringexperiments
Butsave the rawanalogsignals aswellto
thediskalong with thedetecteddigitalHIGH /
LOWsoyou willhavesometrainingmaterialif
youwantto trainsome machine learning
for peak classificationorfor denoising,
25. Alternatives for SAII 1035: BIOPAC
BiopacDTU200
MRIGatingSystemfor TwoSignals
RespirationandECGorBP
Thesearedualchannelgatingsystemsfor small
animal.Itsendscardiactrigger pulsestotheMRI
whenarespirationsignalisinthequietphase.
Pre-processing filtersandgaincontrolsfurther
refinethequalityofthesignalandensurereliable
triggering.Includesadaptercablesfor
monitoringwithaBIOPACResearch
System.
Signal Monitoring There are outputs for the cardiac and
respiration conditioned signals (available at BNC ports:
Buffered ECG/BP and Buffered RSP) and the respective
triggers. The conditioned signals are in the ±10 volt
level range and trigger outputs are 0-5 volts. Seven
BNC to 3.5 mm monitoring cables (CBL102) and CBL122
adapters* are included.
Compatibility The unit will interface with either aBIOPAC
MP160or MP150 system. It will also work with third-party
amplifiers and data acquisition systems that operate in
the±10voltrange.
Dialsontheunitallow
conditioning of theinput
signals. Cardiacand
respiratory signals can
beamplified upto 10X.
Both inputchannelscan
belowpassfiltered
(cardiac 10-100Hz;
respiratory 1-10Hz)and
high passfiltered
(cardiac 0.1-1 Hz;
respiratory 0.05-0.5 Hz).
Conditioned signalscan
bemonitored in real
timethrough analog
inputsto theMP
system.
28. BIOPAC comes with an array ofavailable mouse sensors
The BIOPACMP160
https://www.biopac.com/application/magnetic-resonance-imaging-with-biopa
c-equipment/advanced-feature/mri-small-animal-monitoring/
System supports small animal MRI monitoring system for ECG,
Heart Rate, EMG, blood pressure, respiration, temperature, pulse
oximetry, CO2 and O2 gas analysis, electrical stimulation, and MRI
triggering. BIOPAC has a range of options that can be used in the
MRI for small animal monitoring. The modular MP160 system is
configurable to meet your exact requirements. It is also possible to
interface withexistingMRI-compatible lab equipment.
31. Pulse Oximetry
Images displaying the clip sensors used by
the pulse oximeter systems. (a) In the base
of the mouse or (b) in the centre ofthe footin
rat. The MouseOx®murinepulseoximeter
system from Starr Life Sciences® Corp.
(Oakmont, PA, USA) provides
measurements of O2 saturation, pulse rate,
respiration and pulse and breathe
distension. (c) ProfileofarterialO2 saturation
measurement in rat during MRI acquisitions
at 100% and 21% O2 during inhalation
anaesthesia with isoflurane.
https://doi.org/10.1186/2191-219X-2-44
Pulse Oximetry allows noninvasive monitoring of
arterial blood oxygen saturation. Fiber optic oximetry
sensors are used to transmit pulses of red and infrared
light through the animal’s peripheral vascular region.
Oxygen saturation is determined by measuring the
differentialabsorption of thered and infraredlight.
http://www.i4sa.com/web_app/main/defaultProduct.aspx?ID=34&PT=3
32. RespirationMonitoring
Minimallyinvasivehighlyprecisemonitoringof respiratory rhythm in
themouseusing an epithelialtemperatureprobe 10.1016/j.jneumeth.2016.02.007
Respiratorygating,SAII
Respiration Pad Transducer | TSD110 |
Research| BIOPAC
The TSD110 consists of a differential pressure transducer (TSD160B),
sensor (RX110), and tubing (AFT30). The TSD110 interfaces to an
MP150/MP100 via a DA100C amplifier. The Pressure Pad/Respiration
Transducer (TSD110) requires no electrical connections and works on a
numberof bodylocations(affixwith TAPE1).
https://www.biopac.com/product/pressure-pad-respiration-trans/
Extra-smallimplantsFor
usewithmiceandother
similarlysizedanimals.
DSI(divisionofHarvard
Bioscience) MouseTelemetry
https://www.datasci.com/prod
ucts/implantable-telemetry/mo
use-(miniature)
33. Respirationgating with Ventilator
Vinegonietal.(2015)
https://doi.org/10.1038/nprot.2015.119
"...Olympus microscope, and it is interfaced with a secondary
PC that records physiological and timing signals and
provides cardiac pacing capability through a custom-written
Labviewsoftwareinterface
A differential amplifier (WarnerInstrumentsDP-301) is
configured to provide a single-lead ECG ( (ADInstruments, cat.
no. MLA1213). Animal ventilation is performed with a
volume-control ventilator (ASVInspira55–7058), which
providesthesynchronizationoutput.
The secondary PC uses a data acquisition card (NI PCI-6229)
to record the animal’s ECG, as well as the analog input
synchronization signals from the microscope power supply
unit (FV10-PSU, Frame Active signal) and the ventilator
(Sync Out signal). Cardiac pacing is performed by supplying
an analog output voltage waveform to a stimulus isolator (
AMSystems,2200 stimulator) operating in voltage-to-current
conversionmode."
35. Non-Invasive Blood Pressure (NIBP) ADI#2
The analog inputs receive external
signals up to ±10 V. Each input has
an independently programmable
gain amplifier, filtering, and AC/DC
coupling. Set up each input with the
software, for your requirements. Input
signals can be as low as the
microvolt (µV) range without
the need for external
amplification.
36. PowerLab4/35DAQ ADI
~31,950steps ~0.0094mmHg
In theory thesmallestblood
pressurechangedetectable
AMPLIFIER+DAQ
resolution,thesensor
itselfmightbeworse,
butthis limitcannot be
exceedintheend
LSB
Leastsignificantbit
https://en.wikipedia.org/wiki/
Bit_numbering#Least_signifi
cant_bit
https://www.adinstruments.com/products/powerlab
38. Non-Invasive Blood Pressure (NIBP) Biopac
A lot of the NIBP setups on the market seem to use their own
proprietary software being "dumb devices" in terms of system design
with no outputs that could be hooked directly to a DAQ (Like Fluoview
IO Box),likethe Visitech andMuromachi
Biopak seemsto havemore intelligentoptionwith the amplifier
(NIBP200A) andtail cuff(NIBP250):
https://www.biopac.com/wp-content/uploads/NIBP200A-NIBP250.pdf. https://scholar.google.co.uk/scholar?hl=en&as_sdt=0%2C5&q=NIBP200A+biopac&btnG=
39. Non-Invasive Blood Pressure (NIBP)
Unsuitable for “intelligent” contemporaryDAQ systems
The CODA tail-cuff bloodpressure systemutilizes
Volume Pressure Recording(VPR) sensortechnologyto
measurethemouseorrat tail bloodpressure. Non-
invasivebloodpressure devicesthat utilizeVPRarea
valuabletoolinresearchandwillcontinuetobebeneficial
inmanystudyprotocols.
KentDeviceDataManagementGuide
https://www.kentscientific.com/Customer-Content/www/CMS/files/Data_Manag
ement_Guide_February_2016.pdf
Your Kent Scientific Device supports a robust and
customizable set of data collection, storage and upload
features: History –stores the most recent roughly 1000
records from your runs. This data can be sent to a PC
through the USB port. Upload –uploads real-time data to
your computerthroughtheUSBport.
HarvardApparatus
BloodPressureAnalysisSystemfor
MouseandRat(SC1000)
https://www.harvardapparatus.c
om/blood-pressure-analysis-sy
stem-for-mouse-and-rat-sc100
0.html
Muromachi
MODEL MK-2000ST
NP-NIBP Monitor for Mice & Rats
https://muromachi.com/e
n/archives/english/1798/
Non-InvasiveBloodPressureSystemforRodents
HarvardApparatusPanlabNIPBsystem
https://www.harvardapparatus.com/non-invasive-blood-pressure-system-for-rodents-1.html
PressureandpulseBNCanalogsignaloutput
andRS-232serialport
“Borderlineusableasthiscomeswith
analogoutput”
BP-2000Blood Pressure
AnalysisSystemTM
http://www.visitechsystems.com/
41. Youmight wouldlike to image withoutthecornealcontact
Correction-freeremotelyscannedtwo-photonin
vivomouseretinalimaging
Adi SchejterBar-Noam,NairouzFarah&ShyShoham
Light: Science & Applicationsvolume 5, pagee16007 (2016)
https://doi.org/10.1038/lsa.2016.7 → Citedby16
To scan axially without requiring the objective to come into
contact withthe cornea of theanimal, aconvex electrical tunable
lens (ETL, EL-C-10-30-VIS-LD, Optotune AG), and a concave
offset lens (−100 or −50 mm, plano-concave,Thorlabs) were
positioned in front of a 10× water immersion objective (Nikon,0.3NA,
WD = 3.5 mm). The objective lens was positioned horizontally and
coupled to the eye while the animal faced sideways (a ;→
alternatively, the objective was vertical and the eye of the animal was
facing upwards).
This analysis (c ) showed that the vast majority of available water-→
dipping objectives will be focused by the crystalline lens in
front of the retina even when the objective comes in contact with
the cornea; the only exception in our set were the low-magnification
10×objectivesfromZeiss(0.45NA,WD =1.8 mm) and from Nikon
(0.3NA, WD = 3.5 mm), and the latter provided a much wider working
range and a superior ease of use. Indeed, we were unable to
image theretina except whenusingthese objectives
Using the paraxial model, which was validated by the ray-tracing Zemax model, it
is possible to translate changes in the axial scan parameters to ‘real-world’
coordinates in the eye, which is not trivial as indicated by the 4.4 ratio between the
axial focal shifts without and inside the eye. One benefit of our approach is that it
allows for simple integration of accessory optical systems, such as
photostimulation, photo-coagulation, and optical coherence tomography (OCT),
becausetheycanbeseamlesslycombinedintothesameopticalpath.
42. ElectricalTunable Lens requiresadriver
ApplicationNote:
Opticalfocusinginmicroscopywith
Optotune’sfocustunablelensEL-10-30
https://www.optotune.com/images/products/Optotune%20application%20not
e%20for%20microscopy.pdf
TheEL-10-30canbeeasilycomputer-controlledby
usingaprecisionconstantcurrentdriverfor laser
diodes(e.g.EdmundOpticsNT56-804,Thorlabs
LD1255R,$155) anda0-250mAprogrammableanalog
output.Forsimplefocusingapplications,acalibrated
lookup-tablerelating controlcurrenttofocuspositionsis
sufficient
The Lens Driver 4 offers a simple yet precise way to control Optotune’s electrically tunable lenses, in particular the EL-
6-18 and EL-10-30 series. Communication with the driver follows an open simple serial protocol, which can be
implemented in any programming language on Windows or Linux (C#, Labview and Python source code
available). As a compact USB-powered current source, it also serves for driving LEDs or laser diodes. Comes with I2
C
sensor read-oute.g. for temperature compensation
Designed for industrial use, this LensController by Gardasoft is the ideal solution for
machine vision customers. GigE Vision, RS232 and analog inferfaces as well as
numerous SDKs allow for easy integration. The trigger input and fast response time of the
controller make it also interesting for Z-stacking in microscopy and life science
applications.
SDKs: C++, C#, VB, Labview, Cognex VisionPro, Teledyne Dalsa
Sherlock,Stemmer ImagingCVB
https://www.optotune.com/products/focus-tunable-lenses/lens-drivers
LaserDiodeDriver
DemoBoard
https://www.edmundoptics.com/p/laser-diode-
driver-demo-board-RCD-05P/39965/
43. DriverSchematics
Designed for industrial use, this LensController byGardasoft
is the ideal solution for machine vision customers. GigE
Vision, RS232 and analog inferfaces as well as numerous
SDKs allow for easy integration. The trigger input and fast
response time of the controller make it also interesting for Z-
stackinginmicroscopyand life science applications.
SDKs: C++, C#, VB, Labview, Cognex VisionPro,
TeledyneDalsaSherlock,Stemmer ImagingCVB
https://www.optotune.com/products/focus-tunable-lenses/lens-drivers
The Lens Driver 4 offers a simple yet precise way to control Optotune’s electrically tunable
lenses, in particular the EL-6-18 and EL-10-30 series. Communication with the driver follows an
open simple serial protocol, which can be implemented in any programming language on
Windows or Linux (C#, Labview and Python source code available). As a compact USB-
powered current source, it also serves for driving LEDs or laser diodes. Comes with I2
C sensor
read-out e.g. for temperature compensation
CONSTANT
CURRENT
DRIVE
https://www.optotune.com/images/products/Optotune%20Lens%20Driver%204%20manual.pdf
https://www.optotune.com/Gardasoft_TR_CL180_Datasheet_v001.pdf
One channel, including constant current lens drive and lens EEPROM data communications.
Automatically reads data from EEPROM inside lens which calibrates the controller response.
The performance of the controller istherefore automaticallytailored toeach individual lens.
45. HighFrame Rates forgood images
Invivomultiphotonmicroscopyof cardiomyocyte
calciumdynamicsinthe beatingmouseheart
Smalletal.(2018)https://doi.org/10.1101/251561
(b) Electrocardiogram (ECG) and ventilator pressure are recorded simultaneously during image acquisition
allowing image reconstruction. Red vertical lines indicatethe start of each frame; red arrow indicates the
peak of R-wave used as the start of the cardiac cycle for the frame displayed below; blue arrow indicates
the end of respiratory exhalation that was used as the marker of respiratory cycle. (c) Single raw image
frames with colored boxes indicating the image segments, with corresponding timing of the acquisition
indicated on the ECG and ventilator pressure traces. (d) A plane reconstructed using 512 x 33 pixel
segments, 5% of the cardiaccycle, restricted to 70-100% of therespiratory cycle, and averaged across 4 µm
in z.
We demonstrated intravital multiphoton microscopy in the beating
heart in an intact mouse and optically measured action potentials with
GCaMP6f, a genetically-encoded calcium indicator. Images were
acquired at 30 fps with spontaneous heart beat and continuously
runningventilatedbreathing.
Higher frame rate imaging shows reduced in-frame motion due to heart contraction. Raw
image frames showing same cardiac vessel with (a) standard galvonometric scanning
and (b) resonant scanning. Green dotted lines indicate the timing of the peak of the R wave from the
electrocardiogram which align with image artifacts.
Resonant scanning (Cambridge Technology) data acquisition was performed using a National
Instruments digitizer (NI-5734), FPGA (PXIe-7975), and multifunction I/O module (PXIe-6366) for
device control, mounted in a PXI chassis (PXIe-1073) controlled by ScanImage 2016b. A Ti:Sapphire
laser (Chameleon, Coherent) with the wavelength centered at 950 nm, was used to simultaneously
exciteGCaMP6fand Texas-Red fluorescence.
ECG and respiratory voltage signals were collected with the two unused detection channels
allowing simultaneous recording during imaging. A series of 50–100 frames (1.7 to 3.3 s) per plane in z
were collected at ascanspeed of 30frame/sec. Assigningcardiac and respiratory phase toimage.
We found that with a heart rate of about 5 Hz and breathing at 2 Hz, ~1.5 seconds or about 50
frames was sufficient to generate images in most of the cardiac/respiratory cycle phase space.
Matlab was used for reconstruction and cardiac/respiratory phase-dependent analysis. Scripts
areavailable in Supplement Materials.
46. Gating incardiovascularmicroscopy
Multi-photonmicroscopyincardiovascularresearch
Wuetal.(2017)http://dx.doi.org/10.1016/j.ymeth.2017.04.013
Motional artifacts and loss of focus in un-triggered in vivo TPLSM imaging. The blood
pressure variation during systole and diastole causes vessel contraction and relaxation, resulting in
intra-frame and inter-frame (out-of-focus) artifacts in the images. Three subsequent optical
sections of left carotid artery obtained in vivo without application of external triggering.
Frame rate was 2.3 Hz (1200 lps; line scan rate 1X, image size 400 * 400 pixels). Cell nuclei are visible.
Bars indicate 50 µm. Images are disturbed by intra-frame motional artifacts, causing the arterial wall to
appear as a curved-like structure. Inter-frame artifact (out-of-focus images) due to respiratory
movement result in a different imaging depth of the blood vessel, depending on the phase in the
cardiaccycle.During un-triggeredinvivoimaging, in focusimagesarerarely acquired.
Examples of intravital atherosclerosis (A-E) imaging. A) Imaging of major arteries after endothelial injury (dashed
lines show theoutlineof theelastin layers,) showing cell debrison theluminal sideof theblood vessel (whitearrows) and
subendothelial expression of the inflammatory marker VCAM-1-AF568 (red) in comparison to B) a healthy blood vessel
with an intactendothelial layer (green), labeled using CD31- AF488. C) Both collagen and elastin can beimaged without
labeling, using autofluorescence (coded green) or SHG (coded red), repetitively. These structures can be visualized
better after the addition of dyes, e.g., D) sulfo-rhodamine B (red) for elastin (white arrow) or E) CNA35-FITC (green) for
collagen in plaque-containing carotid artery. Enhanced accumulation of collagen can be observed in the plaque
shoulderregion.
48. Gating incardiovascularMRIimaging#1
Real-TimeGatingSystemforMouseCardiovascularMRImaging
MaherSabbah,HasanAlsaid,LatifaFakri-Bouchet,CedricPasquier,Andre
Briguet,EmmanuelleCanet-Soulas,andOdetteFokapu
MagneticResonanceinMedicine57:29–39(2007)
https://doi.org/10.1002/mrm.21096 |Citedby13
High-resolution MR images of mouse hearts and aortic arches were
acquired using a chain consisting of ECG signal detection, digital signal
processing, and gating signal generation modeled using Simulink (The
MathWorks,Inc.,Natick,MA,USA).
The signal-processing algorithmsusedwererespectivelylow-passfiltering,
nonlinear passband, and wavelet decomposition. Both updated and
nonupdated gating signal generation methods were tested. Noise
reduction was assessed by comparison of the ECG signal-to-noise ratio
(SNR) before and after each processing step. Gating performance was
assessed by measuring QRS detection accuracy before and after online
trigger-leveladjustments.
Low-pass filtering with trigger-level adjustment gave the best
performance for mouse cardiovascular imaging using gradient-echo
(GE), spin-echo (SE), and fast SE (FSE) sequences with minimum induced
delay and maximum gating efficiency (99% sensitivity and R-peak
detection).
This simple digital gating interface will allow various gating strategies to be
optimizedfor cardiovascularMRexplorationsinmice.
Further studies
willseekto
validate
cardiorespiratory
gating withreal-
timeextractionof
therespiratory
signalfromthe
respiration-
modulatedECG
signal.
49. “Intelligent”Gating incardiovascularMRIimaging
Prospectivegatingcontrolforhighlyefficientcardio-
respiratorysynchronisedshortandconstantTRMRIinthe
mouse
PaulKincheshetal.
MagneticResonanceImagingVolume53,November 2018,Pages20-27
https://doi.org/10.1016/j.mri.2018.06.017
Where steady state imaging techniques are required in small animals,
synchronisation is most commonly performed using retrospective
gatingtechniquesbuttheseinvokeaninherenttimepenalty.
Prospective gating incorporating the automatic reacquisition of
data corrupted by motion at the entry to each breath was implemented in
short TR 3D spoiled gradient echo imaging. Motion sensitivity was
examined over the whole mouse body for scans performed without
gating, with respiratory gating, and with cardio-respiratory gating. The
gating methods were performed with and without automatic reacquisition
ofmotioncorrupteddataimmediatelyaftercompletionofthesamebreath.
Diagrammatic representation of respiration gated (R-gated) and cardio-respiratory gated (CR-
gated) MRIschemes. Threshold levels are set on the amplified and filtered ECG and respiration (Resp)
analogue voltages to generate the C-logic and R-logic control signals respectively. The R-logic control
signal is evaluated for R-gated scanning. A user-variable post breath delay ( )τ) is used to ensure that
motion artefact is minimised from the trailing portion of the breath. Only the C-logic signalsthat occur during
the R-logic high level gate are selected to generate the CR-logic control signal which is evaluated for CR-
gated scanning. In the diagram a single respiration corrupted data acquisition block
(marked CD) is automatically reacquired as soon as each breath completes
(markedRD)toreduceartefactfrommotionduringtheonsetofeachbreath.
50. Gating forhumanMRI
Physiorack:AnintegratedMRIsafe/conditional,Gasdelivery,
respiratorygating,andsubjectmonitoringsolutionfor
structuralandfunctionalassessmentsof pulmonaryfunction
J.Magn.Reson.Imaging2014;39:735–741 TechnicalNote
AhmedF.HalaweishPhD H.CecilCharlesPhD
https://doi.org/10.1002/jmri.24219
Actual setup of Physiorack components both inside the scanner room (a) and in the
control room (b), as would be implemented during any given imaging session. (Not in
picture: Oro-nasalfacemask,filtersandPulseoximetrysystem.)
The signals recorded from the pneumotach transducers are amplified by means of
transducer amplifier modules (Biopac, Model DA 100C). All sampled signals
(respiratory, gaseous concentrations, pulse-oximetry, etc.) are recorded and digitized
using a pair of digitizing acquisition modules (DAQ, Windaq, Model DI-158, DataQ
Instruments,Akron,OH).
To evaluate the use of a modular MRI conditional respiratory
monitoring and gating solution, designed to facilitate proper
monitoring of subjects' vital signals and their respiratory efforts, during free‐
breathing and breathheld 19F, oxygen enhanced, and Fourier‐ ‐
decompositionMRI basedacquisitions.‐
We demonstrate an inexpensive,off the shelfsolutionfor monitoring these‐ ‐
signals, facilitating assessments of lung function. Monitoring of
respiratory efforts and exhaled gas concentrations assists in
understanding the heterogeneity of lung function visualized by gas
imaging.
51. Active motion measurement
Motioncharacterizationschemetominimizemotionartifactsin
intravitalmicroscopy
Leeetal.(2017)https://doi.org/10.1117/1.JBO.22.3.036005
METHODS: During intravital imaging sessions,
mice were anesthetized with 2% isoflurane and
2 l/min oxygen, and the body temperature of the
mice was kept constant at 37°C during all
procedures (surgery and imaging). For mice
ventilation, an animal ventilator (Harvard
Apparatus INSPIRAASV55-7058) was used.
The ECG signal, recorded using three needle
electrodes subcutaneously placed in the two
front legs and the right hind leg, was filtered and
amplified using a differential preamplifier (
ADInstrumentsDP-301, output ±10 V). Both
ECG and ventilator traces were recorded using
a data acquisition card (DAQ) (National
Instruments, NI PCI-6229, 1600 SGD, needs the
BNC block NIBNC-2110, 600 SGD) and
Labviewsoftware.
For sensing, a submicron-precision laser
displacement sensor unit (KeyenceLG-030,
~2000SGD) was mounted onto an objective
holder sliding nosepiece, allowing to easily
switch between the imaging objective and the
sensor,withoutrepositioningtheimagedanimal.
In vivo motion characterization. (a) A typical example
of tissue movement as measured by the custom-made
motion characterization system. Two dominant
repetitive motions are observed. The one with big
amplitude is due to respiration and the small one due
to cardiac activity. The ECG measurement in green
color confirms that the small movement is caused by the
heartbeat, and it issynchronized withmotion.
52. Real-time operatingsystems withDAQ
Real-TimeLinuxDynamicClamp:AFastandFlexibleWayto
ConstructVirtualIonChannelsinLivingCells
AlanD.DorvalDavidJ.ChristiniJohnA.White
AnnalsofBiomedicalEngineeringOctober 2001,Volume29,Issue10,
https://doi.org/10.1114/1.1408929
“The dynamic clamp require a high frequency current
clamp amplifier. The amplifier must connect to a
personal computer (PC) controlled, data acquisition
board (DAQ). Our amplifier was connected to a National
Instruments, PCI-MIO-16XE-50 data acquisition
board. This DAQ boasts 16 channel, 16 bit analog-to-
digital input (A/D) and 2 channel, 12 bit digital-to-analog
output(D/A),bothrunningatamaximumof20kHz.”
“The PC runs a free, open source extension to the Linux
operating system, known as Real Time Linux (RTL).
RTL is a ‘‘hard’’ real time operating system, which
means that commands will always be executed in a
known amount of time. RTL provides high temporal
precision on a PC, while maintaining the full functionality
of the now widely supported parent operating system,
Linux.”
53. DAQA/D Resolution importance
Performancecomparisonbetween8-and14-bit-depthimaginginpolarization-
sensitiveswept-sourceopticalcoherencetomography
ZenghaiLu,DeepaK.Kasaragod,andStephenJ.Matcher (2011)
https://doi.org/10.1364/BOE.2.000794
AQplusreceivernoise
measurementsatdifferentset
fullanaloginputvoltageranges
(FIVR) for 14-bit(a)and8-bit
DAQ(b),respectively.(c):
standarddeviationofthe
measuredDAQplusreceiver
noisealongwiththecalculated
noisestandarddeviationof
quantizationnoiseoftheDAQ
https://spectrum-instrumentation.com/en/m2i4022
We compare true 8- and 14-bit-depth
imaging of SS-OCT and polarization-
sensitive SS-OCT (PS-SS-OCT) by
using two hardware-synchronized high-
speeddataacquisition (DAQ)boards.
The two signals are sampled at 20MS/s
simultaneously with 14-bit (M2i.4022,
Spectrum GmbH, Germany) and 8-bit (
M2i.2031, Spectrum GmbH, Germany)
resolution.
55. Startwithopen-sourceplatforms suchasRTXI
Hardreal-timeclosed-loopelectrophysiologywiththeReal-
TimeeXperimentInterface(RTXI)
YogiA.Patel,AnselGeorge,AlanD.Dorval,JohnA.White,DavidJ.Christini,
RobertJ.ButeraPLOSComputationalBiology13(7):e1005656
https://doi.org/10.1371/journal.pcbi.1005430
https://doi.org/10.1371/journal.pcbi.1005656
http://rtxi.org/
https://github.com/rtxi
On-going RTXI development efforts are also
focused on providing API calls for distributing
computational loads across dedicated
processor cores and GPUs, with the goal of
requiring little to no technical know-how on the
user’send.
RTXI uses the open source Xenomai framework to implement
communication with a variety of commercially available multifunction
DAQ cards with both analog and digital input and output channels. This
makes RTXI essentially hardware-agnostic and able to
communicate with multiple actuators and sensors that may span different
modalities.
ListofDAQssupportedbytheanalogydriver
Driverslist ni_pcimio
This drivers suppors a long list of NationalInstrumentsPCI /PXI
cards:
PCI-MIO-16XE-50, PCI-MIO-16XE-10, PCI-MIO-16E-1, PCI-MIO-16E-4,
PCI-6014
PCI-6023E, PCI-6024E, PCI-6025E, PXI-6025E
PCI-6030E, PXI-6030E, PCI-6031E, PCI-6032E, PCI-6033E, PCI-6034E,
PCI-6035E, PCI-6036E
PCI-6040E, PXI-6040E
PCI-6052E, PXI-6052E
PCI-6070E, PXI-6070E, PCI-6071E, PXI-6071E
PCI-6110, PCI-6111
PCI-6220, PCI-6221
PCI-6143,PXI-6143
PCI-6224, PCI-6225, PCI-6229
PCI-6250, PCI-6251, PCIe-6251,PCI-6254, PCI-6259, PCIe-6259
PCI-6280, PCI-6281, PXI-6281, PCI-6284, PCI-6289,
PCI-6711, PXI-6711,PCI-6713,PXI-6713,
PCI-6731,PCI-6733, PXI-6733,
56. GPUswithDAQs convergingwithreal-timedeeplearning
DataAcquisitionwithGPUs:TheDAQfortheMuong-2
ExperimentatFermilab
W.Gohn(Submittedon15Nov2016)
https://arxiv.org/abs/1611.04959
The muon g-2 experiment at Fermilab is heavily
relying on GPUs to process its data. The data
acquisition system for this experiment must have
the ability to create deadtime-free records from
700 µs muon spills at a raw data rate 18 GB
per second. Data will be collected using 1296
channels of µTCA-based 800 MSPS, 12 bit
waveform digitizers and processed in a layered
array of networked commodity processors with
24 GPUs working in parallel (26 Nvidia Tesla
K40 GPUs housed by pairs in 13 front-end
computers) to perform a fast recording of the
muondecaysduring the spill.
In addition to numerous models of GPUs, there are also coprocessor systems such
as the Intel Xeon Phi, which utilize fewer but faster cores than the GPUs, as well as
FPGAsor ASDQs,which require significantlymore programming overhead than do
theGPUssystems
18 GB per second
→ 144 Gbit/s
Comparetohigh-speedcameras
with PCIExpressGen.3x8 with8
GBperseconddatarates
59. GPUacceleratedDAQswithOCTimagingaswell#1
DevelopingtheWorld’sFirstReal-Time3DOCT Medical
ImagingSystem WithLabVIEWandNI FlexRIO
Dr.KohjiOhbayashi 大林 康二 ,KitasatoUniversity,GraduateSchoolofMedicalScience
http://sine.ni.com/cs/app/doc/p/id/cs-13387
“Using optical coherence tomography (OCT) and a 320-
channel data acquisition system combining NI FlexRIO field-
programmable gate array (FPGA) hardware and GPU
(NVIDIA Quadro FX 3800) processing to create the world’s
firstreal-time 3D OCT imaging system”
For high-speed acquisition, we use the NI 5751 adapter module,
which has a 50 MS/s sample rate on 16 simultaneous channels with 14-
bit resolution. The adapter module interfaces to the NI PXIe-7962R
FPGA module, which we use to perform the first stage of processing –
subtraction of the sample-cut noise and multiplication of a window
function. In total, we have 20 modules across two PXI Express chassis,
so we use two NI PXIe-6674T timing and synchronization
modules to distribute clocks for the system and assure precise phase
synchronizationacrossallthechannelsinthesystem.
61. GPUacceleratedDAQswithOCTimagingaswell#2
High-speedFPGA-GPUprocessingfor3D-OCT imaging
Kyung-ChanJin; Kye-SungLee; Geun-HeeKim(March 2018)
https://doi.org/10.1109/CompComm.2017.8322904
In thispaper, we propose the designofa real-time image acquisition andpre-
processing FPGA(NI PCIe-1473R)viaLabVIEW(NationalInstruments(NI))with
GPU-basedaccelerationthatiscapableofsustainingtherateofdataacquisition.
Results showed that, by applying GPU acceleration to the tomographic
inspectionofbiologicalsamples,SD-OCTimaginginexcessof40frames/s(FPS)
for the NVIDIA M6000 (7 Tflops at fp32) GPU-accelerated SD-OCT with frame
size 4096 (axial) × 512 (lateral) becomes feasible, and more than 512 × 512 × 500
volumes can be reconstructed with a speed increase of at least 7x that of a
non-GPU.
Linux-based
systemwithFPGA-
GPUmodule
we utilized the Spimagine Python packageto interactively visualize
and process the 3D tomographic image (via OpenCL)
65. You mightwanttoadjustyourstimulusbasedonresponse
Neurofeedback paradigms with brain stimulation (tACS, rTMS), steady-state visual evoked responses (SSVEPs),
individualized alphafrequency (IAF) driving, etc.
D. Reatoet al. . Effectsof weak
transcranial alternatingcurrent
stimulation on brain activity—areview of
knownmechanismsfrom animal studies.
FrontiersinHumanNeuroscience,7,Oct.2013.
http://dx.doi.org/10.3389/fnhum.2013.00687
Gets even tricker ifyou need to read neuron firing
from an image(calcium dye) in real-time as you need
somedeep learning imageanalysis for this.
Frequency dependenceof optogenetic
slicemodeloftACSfrom Kukietal.2013
LeChasseuretal.(2011)
Electrophysiology withoptical
electrocorticography
67. SpatiotemporalNeurovascular Coupling
Functionalopticalcoherencetomographyofneurovascular
couplinginteractionsintheretina
Sonetal.(2018)https://doi.org/10.1002/jbio.201800089
Here, we report a multimodal functional optical coherence
tomography (OCT) imaging methodology to enable
concurrent intrinsic optical signal (IOS) imaging of stimulus‐
evoked neural activity and hemodynamic responses at
capillaryresolution.
OCT angiography guided IOS analysis was usedto separate
neural IOS‐IOS and hemodynamic IOS changes‐IOS in the
same retinal image sequence. Frequency flicker stimuli
evoked neural IOS changes in the outer retina‐IOS ; that is,
photoreceptor layer, first and then in the inner retina, including
outer plexus layer (OPL), inner plexiform layer (IPL), and
ganglion cell layer (GCL), which were followed by
hemodynamic IOS changes primarily in the inner‐IOS
retina;thatis, OPL,IPL,andGCL.
Different time courses and signal magnitudes of
hemodynamic IOS responses were observed in blood‐
vesselswith variousdiameters
Alteredneurovascularcouplingasmeasuredbyoptical
imaging: abiomarkerfor Alzheimer'sdisease
https://doi.org/10.1038/s41598-017-13349-5
Spatiotemporal
patterns analyzed
now afterthe
experiment, but what
if you would want to
do this in real-time?
69. “BigData” enteringvolumetricfunctionalimaging
StudyingAxon-AstrocyteFunctionalInteractionsby3DTwo-
PhotonCa2+Imaging:APracticalGuidetoExperimentsand
“BigData”Analysis
IaroslavSavtchouk,GiovanniCarrieroandAndreaVolterra
Front.Cell.Neurosci.,13April2018
https://doi.org/10.3389/fncel.2018.00098
Recent advances in fast volumetric imaging have enabled
rapid generation of large amounts of multi-dimensional
functional data. While many computer frameworks exist for
data storage and analysis of the multi-gigabyte Ca2+
imaging experiments in neurons, they are less useful for
analyzing Ca2+ dynamics in astrocytes, where
transients do not follow a predictable spatio-temporal
distributionpattern.
In this manuscript, we provide a detailed protocol and
commentary for recording and analyzing three-dimensional
(3D) Ca2+ transients through time in GCaMP6f-expressing
astrocytes of adult brain slices in response to axonal
stimulation, using our recently developed tools to perform
interactive exploration, filtering, and time-correlation analysis
ofthetransients.
https://wwwfbm.unil.ch/dnf/group/glia-an-active-synaptic-partner/member/v
olterra-andrea-volterra
form ofsoftwarepluginsfor ImageJ (NIH)
Synchronization of imaging and electrical stimulation via simultaneous capture of timing
information from the two systems. By concurrently recording the image frame counts and the
electrical inputs, one can later link analytically the imaging and electrophysiology data. (A) Schematic
diagram of connections. An electrophysiology computer is also simultaneously recording two signals
for synchronization: a Y-galvanometer position feedback pin, and a split-off of a stimulator TTL
trigger. (B) Low-zoom overview of the captured synchronization signals (light-blue highlight is
magnified in next panel, not to scale). Vertical deflections in the Y-galvo trace correspond to individual
Y-frame scans, whereas spikes on the Stim TTL trace indicate the relative timings of axonal
stimulation (C). A high-zoom version of the highlighted stretch in (B), showing the relationship
between the imaging frame position and the stimulation signal timing. Each Z-stack consumes an
entire Y-scan frame per focal plane (here 21) plus any additional overhead, depending on whether bi-
directionalz-scanningisimplemented,etc.
70. Closed-LoopElectrophysiology Example #1
Hardreal-timeclosed-loopelectrophysiologywiththeReal-
TimeeXperimentInterface(RTXI)
YogiA.Patel,AnselGeorge,AlanD.Dorval,JohnA.White,DavidJ.Christini,
RobertJ.ButeraPLOSComputationalBiology13(7):e1005656
https://doi.org/10.1371/journal.pcbi.1005430
https://doi.org/10.1371/journal.pcbi.1005656
Real-time control applications for biological research are
available; however, these systems are costly and often restrict
the flexibility and customization of experimental protocols. The
Real-Time eXperiment Interface (RTXI) is based on Xenomai, a
Real-Time Linux framework, and is an open source software
platform for achieving hard real-time data acquisition and
closed-loop control in biological experiments while retaining the
flexibilityneededfor experimental settings.
RTXI has enabled users to implement complex custom closed-loop
protocols in single cell, cell network, animal, and human
electrophysiology studies. RTXI is also used as a free and open
source, customizable electrophysiology platform in open-loop
studies requiring online data acquisition, processing, and
visualization.
RTXI interfaces with experiments through a variety of hardware interfaces,
including PCI/PCIebasedDAQsfromNational Instrumentsand Sensoray, Ethernet
based devices such as cameras and commercial amplifiers, as well as USB-based
acquisitiondevice
On-going development efforts within RTXI are focused on incorporating new measurement
modalities (e.g., imaging) and acquiring from high channel count interfaces—all with hard RT
closed-loop performance. For example, a image acquisition and processing module (GenICam)
andaEthernet-based dataacquisition module(EthernetAcq)arenowavailableforusewithin RTXI.In
many cases, hard RT closed-loop control with image or high channel count data processing can
require more computation time than is available per cycle. On-going RTXI development efforts
are also focused on providing API calls for distributing computational loads across
dedicated processor cores and GPUs, with the goal of requiring little to no technical know-how
on theuser’send.
Experiment data and metadata saved by the
Data Recorder are stored in the Hierarchical
Data Format (HDF5). HDF5 file read and write
operations are supported by many common
analysis frameworks and languages (e.g.,
MATLAB, Python, Julia, R, etc).
72. Closed-LoopElectrophysiology Example #3
Multimed:AnIntegrated,Multi-ApplicationPlatformforthe
Real-TimeRecordingandSub-MillisecondProcessingof
Biosignals
AntoinePirog,YannickBornat,RomainPerrier,MatthieuRaoux,Manon
Jaffredo,AdamQuotb,JochenLang,NoëlleLewis,andSylvieRenaud
Sensors(Basel).2018Jul;18(7):2099.
https://dx.doi.org/10.3390%2Fs18072099
We designed Multimed, which is a versatile hardware platform
for the real-time recording and processing of biosignals. Digital
processing in Multimed is an arrangement of generic processing
units from a custom library. These can freely be rearranged to
matchthe needsof the application.
Embedded onto a Field Programmable Gate Array (FPGA), these
modules utilize full-hardware signal processing to lower processing
latency. It achieves constant latency, and sub-millisecond
processing and decision-making on 64 channels. The FPGA core
processing unit makes Multimed suitable as either a
reconfigurable electrophysiology system or a prototyping
platform for VLSI implantable medical devices. It is specifically
designed for open- and closed-loop experiments and
provides consistent feedback rules, well within biological
microsecondstimeframes.
Multimedsetupshave been installedin multiplesitesandare being usedbypartner laboratories
in collaborative projects, always aiming for closed-loop experiments (BRAINBOW (EU
project 284772, ICT- FET FP7/2007–2013, FET Young Explorers) [33], CENAVEX (ANR grant
2013-NEUC-0001-01 and NIH grant 5 R01 NS086088-02) [34,35,40], ISLET CHIP (ANR grant
2013-PRTS-0017)[25,41],HYRENE(ANRgrant2010-BLANC-0316-01)[21])
74. Deeplearningcanimprovethemicroscope imagequality
DeeplearningmicroscopyYairRivenson,ZoltánGöröcs,Harun Günaydin, Yibo Zhang,Hongda Wang,and Aydogan Ozcan.
Optica Vol.4, Issue11, pp. 1437-1443 (2017) https://doi.org/10.1364/OPTICA.4.001437
Thefirststepin thisdeep-learning-
basedmicroscopyframework
involveslearningthestatistical
transformationbetweenlow-
resolutionandhigh-resolution
microscopicimages,whichis used
totrainaCNN
Wehavechosenbright-field
microscopy withspatiallyand
temporallyincoherentbroadband
illuminationasanexample, the
samedeeplearningframework
mightbeapplicabletoother
microscopy modalities,including,
e.g.,holography,dark-field,
fluorescence,multi-photon,optical
coherence tomography,among
others.
After appropriate training, this framework and its derivatives might
be applicable to other forms of optical microscopy and imaging
techniques and can be used to transfer images that are
acquired under low-resolution systems into high-
resolution and wide-field images, significantly extending the
space bandwidth product of the output images. Furthermore,
using the same deep learning approach we have also
demonstrated the extension of the spatial frequency response of
the imaging system along with an extended DOF. In addition to
optical microscopy, this entire framework can also be applied to
other computational imaging approaches, also spanning
different parts of the electromagnetic spectrum, and can be used
to design computational imagers with improved resolution, FOV,
and DOF.
75. Youcouldhavea“goldstandard”systemwithhigh-endgatingand
motionmeasurement/compensationsystem
Andthentrain
deeplysupervised
deeplearning
networktoboth
detectandcorrect
motionartifacts?
Deep learning-based detectionofmotionartifactsin
probe-based confocal laserendomicroscopyimages
MarcAubreville,MaikeStoeve,NicolaiOetter,MiguelGoncalves,ChristianKnipfer,Helmut
Neumann,ChristopherBohr,FlorianStelzle,AndreasMaier
InternationalJournalofComputerAssistedRadiologyandSurgery(2018)
https://doi.org/10.1007/s11548-018-1836-1
Each of the images was manually assessed for motion artifacts by
two experts with background in biomedical engineering, while the second
expert was able to see the annotations of the first expert (non-blinded). The
annotation results have been validated by two medical experts with profound
experience in CLE diagnosis. All annotations (bounding boxes of
artifacts) werestored in arelational databaseandusedforbothtrainingand
evaluation.
Efficient DataLabellingfor Ocular Imaging
https://www.slideshare.net/PetteriTeikariPh
D/efficient-data-labelling-for-ocular-imagin
g-110540104
82. Measurethe PSFwithAdaptiveOptics andlearntheinverseproblem
Characterizationandadaptiveopticalcorrectionofaberrationsduringinvivoimagingin
themousecortex NaJi, Takashi R. Sato, and Eric Betzig (2012)https://doi.org/10.1073/pnas.1109202108
Ji et al. (2012): Lateral and axial images of GFP-expressing
dendritic processes (mouse cortex, 2-PM, 170 μm
Adaptiveopticsinmultiphotonmicroscopy:comparisonoftwo,
threeand fourphotonfluorescence David Sinefeld etal.(2015)
https://doi.org/10.1364/OE.23.031472
Phase correction for a 2-m-focal length cylindrical lens for 2-, 3- and 4- photon
excited fluorescence of Alexa Fluor 790, Sulforhodamine 101 and Fluorescein.
(a) Left – 4-photon fluorescence convergence curve showing a signal
improvement factor of × 320. Right – final phase applied on the SLM (b) left – 3-
photon fluorescence convergence curve showing a signal improvement factor of
× 40. Right – final phase applied on the SLM. (c) Left – 2-photon fluorescence
convergence curve showing a signal improvement factor of × 2.1. Right – final
phaseappliedon theSLM.Color-barsarein wavelengthunitscale.
86. Deeplearningfor coded-illumination sourcedesign
Physics-basedLearnedDesign:Optimized Coded-
IlluminationforQuantitativePhaseImaging
Michael R. Kellman, EmrahBostan,NicoleRepina, Michael Lustig,Laura Waller
https://arxiv.org/abs/1808.03571
Learning Coded-Illumination Design for Quantitative
Phase Imaging: (a) Schematic of the LED-illumination
microscope where multiple intensity measurements are captured
under unique coded-illumination patterns, (b) Computational
phase reconstruction of the sample’s optical phase with coded-
illumination measurements. (c) Optimization for learning of coded-
illuminationdesignbasedonthenon-linear iterativereconstruction.
InvitroQuantitativePhaseImaging (QPI)
enablesthestain-andlabel-free imagingof
transparentbiologicalsamples.
Thisveryrelevantwhen
tryingtoimageretinal
ganglioncells
EthanA.Rossi etal.(2017)
10.1073/pnas.1613445114
87. AdaptiveOptics withthreephotons ratherthantwo
Sinefeld D,PaudelHP,WangT,WangM,OuzounovDG,Bifano TG,XuC: Nonlinearadaptive
optics:aberrationcorrectioninthreephotonfluorescencemicroscopyformouse
brainimaging.Proc SPIE2017 https://doi.org/10.1117/12.2252686
Here, we present a 3PM AO microscopy system
for brain imaging. Soliton self-frequency shift is
used to create a femtosecond source at 1675
nm and a microelectromechanical (MEMS) SLM
serves as the wavefront shaping device. We
perturb the 1020 segment SLM using a modified
nonlinear version of three-point phase shifting
interferometry. The nonlinearity of the
fluorescence signal used for feedback ensures
that the signal is increasing when the spot size
decreases, allowing compensation of phase
errors in an iterative optimization process without
direct phase measurement. We compare the
performance for different orders of nonlinear
feedback, showing an exponential growth in
signal improvement as the nonlinear order
increases. We demonstrate the impact of the
method by applying the 3PM AO system for in-
vivo mouse brain imaging, showing
improvement in signal at 1-mm depth inside
the brain.
SECTIONING
Horton et al. (2013)
Three-photon microscopy
2PM, attenuation
z2
from focal plane
3PM, attenuation
z4
from focal plane
osa-opn.org, November 2013
3-PM 601um 2-PM 429 um
Wang et al.
(2015)
88. Deeplearningtakingoverthe roleofphysicalcomponents aswell
EranHershko,Lucien E.Weiss,TomerMichaeli, YoavShechtman. Technion(2018)
Multicolor localizationmicroscopy bydeeplearning. ProcSPIE2017
https://arxiv.org/abs/1807.01637
First, we experimentally demonstrate an algorithm for
determining an emitter’s color using a standard fluorescence
microscope equipped with a grayscale camera with no
additional hardware modification. This is enabled by the fact that
the PSF of any optical system is dependent on the
wavelength, even without PSF engineering. Second, we
developandexperimentallydemonstrate anadditionalneuralnet
that algorithmically optimizes a color-encoding PSF using
phasemodulation,for maximalcolor-distinguishability.
To test whether a neural net could discriminate between two
types of emitters, we prepared a thin sample containing green
and red quantum dots (Qdots) with emission peaks at 565
and705nm,andimageditusing anepifluorescencemicroscope.
Here, we have demonstrated how deep learning is capable of
performing roles traditionally accomplished with physical
components. Post-process, software tools can be
advantageous over hardware-based methods due to a lower
implementation cost, system adaptability, and further
optimization without the requirement of collecting new,
experimentaldatasets.
To optimally discriminate between PSFs, we have shown that
PSF-engineering can be done in coordination with net training
to maximize on the strengths of the reconstruction net, which do
notfollowthesameprocessasmost-likelihoodestimators.
89. Marryingdeeplearning withMonteCarlophysics-basedmodelling
AnalyzingInverseProblemswithInvertibleNeuralNetworks
Lynton Ardizzone, Jakob Kruse, Sebastian Wirkert,Daniel Rahner, EricW.Pellegrini, RalfS. Klessen, LenaMaier-Hein,Carsten Rother, Ullrich Köthe (Submitted on 14Aug 2018)
https://arxiv.org/abs/1808.04730
The results produced by the INN provide several new insights: First, we find that
the posteriors for layer thickness and anisotropy match the shape of their
priors, i.e. y holdsno information about these parameters– theyare unrecoverable.
Second, we find that the sampled distributions for the blood volume fraction and
scattering amplitude are strongly correlated. As blood volume fraction
increases, more light is absorbed inside the tissue. For the sensor to record
the same intensities y asbefore, scattering must be increasedaccordingly.
While the correspondence between simulations and real measurements remains to
be established, we share the excitement of the application experts to push INNs
towards a generic tool, helping scientists from many different disciplines to better
interpret theirdata and models, and to better plan their next experimental steps
– be it modeling,measuringorsimulation
In medical science, the functional state of biological tissue is
of interest for many applications, such as tumor detection or
verifying organ transplantation success. Tumors, for
example, are expected to exhibit changes in oxygen
saturation. Such changes influence the reflectance of
the tissue, which can be measured by multispectral
cameras.
We can simulate these measurements from a tissue model
involving oxygen saturation, blood volume fraction,
scattering magnitude, anisotropy, and tissue layer thickness
[Wirkertetal.2016]. While these simulations can determine
the reflectance spectrum (y) for a given tissue, inverting the
measurements to recover the underlying functional
properties(x) isanactivefieldofresearch.
We train an INN for this problem, along with two ablations
(only forward or only inverse training), as well as a regular
neural net using the method of Kendall andGal 2007, with
Monte-Carlo (MC) dropout and additional aleatoric error
termsforeach parameter.
90. TheMoreDataThemoreyoucansynthesizedataaswell
Fast3Dcelltrackingwithwide-fieldfluorescencemicroscopythroughdeeplearning
KanLiu, Hui Qiao, Jiamin Wu, Haoqian Wang, LuFang QionghaiDai (2018)
https://arxiv.org/abs/1805.05139
Framework of the proposed 3D localization microscopy. Lateral detection CNN, highlighted by the blue
dashed box, first determines whether there exist diffraction patterns at the central lateral position of the sliding window.
Axial localization CNN, highlighted by the orange dashed box, then estimates the axial positions of the predicted
positive samplesof lateral detection CNN.
Therefore, 3D positions of the fluorescent probes are finally acquired. Making use of the determined 3D localization
results, fast 3Dtracking can be realized with aKalmanfilter.
The large amounts of training data for our framework is obtained from the simulation of the
incoherent superposition of multiple objects with the prior knowledge of the z-stack of a single object.
While thez-stack oftheobjectcan be synthesized bythesimulated pointspread function (PSF)and the
shape of the object, we choose an experimental z-stack of a single object for training data synthesis
due to the diverse imaging environments in different experiments, such as the optical aberration, medium
inducedrefractiveindex mismatch, andnoisecondition.
Tracking blood cells (75 µm/s) at 100 fps of a one-day-old
live zebrafish restrained in agarose. (a) Captured wide-field
fluorescence images of the ROI at different time stamps and the
corresponding localization results reconstructed by our method
and MLE method. (b) 3D tracking results of the blood cells
reconstructedbyour methodandMLEmethod.
91. Manyfieldsofopticsbenefitfromlargedatasetstobeusedfororwithsynthesis
pipelines.Deeplearningfor ultrashortpulsereconstruction
Deeplearningreconstructionof ultrashortpulsesTom
Zahavy,AlexDikopoltsev,Daniel Moss, Gil Ilan Haham,Oren Cohen, ShieMannor, and
Mordechai SegevOpticaVol. 5,Issue5,pp. 666-673 (2018)
https://doi.org/10.1364/OPTICA.5.000666
Here, we propose and
demonstrate, theoretically and
experimentally, the reconstruction
of ultrashortoptical pulsesby
employingdeepneural networks
(DNNs),andshow (on simulated
data) that ourtrainednetwork
outperformsother state-of-the-art
techniquesfor low SNR
measurements.
We furtherdevelop our
methodology bymodifyingthe
network trainingstage to combine
bothsupervisedandunsupervised
learning, and showthat thisnew
network isable to reconstruct
ultrashort pulsesfrom low SNR
experimental data, while being
trained on simulated data.
To further enhance the performance of our
approach, we plan to investigate the sim-to-real
challenges in future work. First is increasing the
variety of the computer-generated dataset to include
asmany spectral amplitudesand phasesaspossible.
The second is to significantly enlarge the number of
measured pulses that train the network. Of
course, this suggestion has obvious disadvantages, but
in some experimental schemes, where the
measurements are embedded in noise, or when extreme
accuraciesare crucial,thiscould be practical.
The third is using generative models to generate
more data by learning the data distribution of measured
pulses. In particular, a recently developed network called
the generative adversarial network (GAN) [34]
can be used to create new data pulses on which the
DNN tends to make mistakes (poorly reconstruct the
pulses). These pulses will be new to the dataset on
purpose, and will increase the variety of the pulses in the
trainingdataset.
92. Synthesizingnewsamples asweknowprettywellwithallthetweaksthe
“imagemodel”/latentspace,creating “virtualanimals”
RobustHeartbeatDetectionfromMultimodalData viaCNN-basedGeneralizableInformation
Fusion BS Chandra, CS Sastry,S Jana (2018)https://arxiv.org/abs/1807.03232
Virtual patient generation with possibly different
paravalvular leakage (PVL) levels, for patients with transcatheter
aorticvalvereplacement(TAVR)
Combining a convolutional neural network (CNN) and
generative adversarial networks (GAN), we discover
the pathophysiologic meaning of the feature
space. This demonstrates l generative invertible
networks (GIN) can generate virtual patients not only
visually authentic but also pathophysiologically
interpretable