SlideShare a Scribd company logo
1 of 1
Introduction
Chemical tissue staining is currently used for
studying cancer pathology: tissue cores
extracted from a patient are thinly sliced and
stained using chemical dyes to highlight
different cell types. The stained tissue samples
are examined under a microscope to diagnose
disease.
The Bioimaging Science and Technology group
at the Beckman Institute uses infrared
spectroscopy to directly collect chemical
information. The goal is to use this quantitative
information to improve disease diagnosis by
providing more accurate information to
pathologists.
Methods and results
The first step in our program is loading the
ENVI header information. This includes all of
the properties of the image, such as the x and y
resolution, the resolution of the infrared spectra
(number of bands), and the units of
measurement used. These parameters define
how the random forest is constructed. Our
program stores them as variables for later use.
Due to the large file size of the ENVI images,
an entire image cannot be stored in memory,
and must rather be streamed from the hard
drive and classified sequentially. This is known
as out-of-core processing. Our program feeds
individual chunks into the classifier one by one,
building up a fully classified image.
Because four collaborators all worked on the
same program concurrently, utilizing a
distributed version control system called Git
was extremely important. This allowed us to
keep track of and comment on any edits made
in our code. Git also allows for branching, so
separate users can work independently, then
merge all the branches back into one main
master branch.
Our code was managed using Cmake, which
allows us to easily link our programs with
external libraries, such as Qt and ALGLIB,
which were used for user-interface design and
classification.
Spectroscopic Images
This resulting infrared signal produces a digital
image, in which each pixel corresponds to a
frequency in a spectrum graph where the point
will vibrate. This data is stored using the ENVI
(Environment for Visualizing Images) format.
This large, specialized file is then run through a
classification program using the Random-
Forest algorithm (Figure 5). This classifier
analyzes each pixel based on its spectrum and
surroundings.
David Bergvelt and Max Li, under the supervision of David Mayerich
Led by Professor Bhargava at Bioimaging Science and Technology group, Beckman Institute, University of Illinois at Urbana-Champaign
Digitizing cancer pathology research
Acknowledgments
We would like to thank Professor Bhargava for
giving us the opportunity to work alongside his
Bioimaging group, and also we would like to
thank David Mayerich for giving us the chance
to try our hands at programming, and guiding
us along our way.
We would also like to thank David Bergandine
for sponsoring us and advising us throughout
the I-STEM program, and Ms. Williams and
Mrs. Destefano for organizing the I-STEM
program.
Experience
We found that the hands-on experience of
working with programming in C++ helped a lot
in learning the language. Learning how to use
Git and Cmake, two industry standard
applications, will be very useful in the future. It
was also a very interesting experience
connecting the topics of cancer pathology and
statistical analysis together using programming
languages.
Working along with Dr. Mayerich in Professor
Bhargava’s group gave us the experience of
working in a research group doing cutting edge
research related how the future of cancer
research and diagnosis will look like. Overall,
the I-STEM experience was excellent, and we
hope to continue working with Dr. Mayerich and
the Bhargava group in the future.
Aim
Our project focuses on building a C++ program
which allows the user to load a spectroscopic
image of a tissue sample, feed it into the
Random-Forest algorithm, and output a
classified image showing differences in tissue
type.
The final program will incorporate a graphical
user interface, making it user-friendly and
useful to a wide range of researchers. We hope
that this will encourage researchers to adopt
these quantitative methods in their own
research and diagnostic practices. In
particular, we expect that this type of
technology will be useful for accurate disease
diagnosis in hospitals.
Fig. 1: Image of a breast tumor biopsy stained using various chemical methods
(left) and an image of the same biopsy stained digitally after spectroscopic
imaging and classification (right).
Fig. 5: A Random
Forest is composed of
hundreds of decision
trees, where each tree
selects a cell type
based on a random
subset of features of a
single spectrum.
Each tree then “votes”
for its selected class.
Fig. 4: Each pixel making up the image of the digitally stained cell carries its
own infrared spectrum (a and b). The image is the classified into various cell
types, such as epithelium, fibroblasts, etc. The accuracy of the classifier is
given using the precision: the ratio of cell types that are correctly classified.
1200 1800 2400 3000 3600
0.0
0.1
0.2
0.3
0.4
Absorbance(a.u.)
Wavenumber (cm-1
)
Fig. 2: Schematic of a mid-infrared spectroscopy setup. A detector is used to
measure the intensity of the infrared light as it passes through the specimen,
the independent variable being the position of a movable mirror.
Fig. 3:The position of the mirror is plotted along with the associated light
intensity, and a Fourier Transformation if applied, transforming it into a function
of wavelength and intensity.
1.00
0.00
Fig. 6: Final result of our classification. The color scale represents the
probability of tissue being epithelium tissue, where dark red = very strong
probability of epithelium tissue, dark blue = very weak probability of epithelium.
(a) (b)
(c)

More Related Content

What's hot

E-book Thesis Sara Carvalho
E-book Thesis  Sara CarvalhoE-book Thesis  Sara Carvalho
E-book Thesis Sara CarvalhoSara Carvalho
 
2008 SPIE Photonics West
2008 SPIE Photonics West2008 SPIE Photonics West
2008 SPIE Photonics WestDanthu
 
Multimodal Medical Image Fusion Based On SVD
Multimodal Medical Image Fusion Based On SVDMultimodal Medical Image Fusion Based On SVD
Multimodal Medical Image Fusion Based On SVDIOSR Journals
 
Twenty Years of Whole Slide Imaging - the Coming Phase Change
Twenty Years of Whole Slide Imaging - the Coming Phase ChangeTwenty Years of Whole Slide Imaging - the Coming Phase Change
Twenty Years of Whole Slide Imaging - the Coming Phase ChangeJoel Saltz
 
Detection of medical instruments project- PART 2
Detection of medical instruments project- PART 2Detection of medical instruments project- PART 2
Detection of medical instruments project- PART 2Sairam Adithya
 
Text Mining Radiology Reports for Deep Learning Radiology Images
Text Mining Radiology Reports for Deep Learning Radiology Images Text Mining Radiology Reports for Deep Learning Radiology Images
Text Mining Radiology Reports for Deep Learning Radiology Images Yifan Peng
 
Brain tumor detection using image segmentation ppt
Brain tumor detection using image segmentation pptBrain tumor detection using image segmentation ppt
Brain tumor detection using image segmentation pptRoshini Vijayakumar
 
IRJET- Breast Cancer Prediction using Support Vector Machine
IRJET-  	  Breast Cancer Prediction using Support Vector MachineIRJET-  	  Breast Cancer Prediction using Support Vector Machine
IRJET- Breast Cancer Prediction using Support Vector MachineIRJET Journal
 
Common Sense Based Joint Training of Human Activity Recognizers
Common Sense Based Joint Training of Human Activity RecognizersCommon Sense Based Joint Training of Human Activity Recognizers
Common Sense Based Joint Training of Human Activity Recognizerswacerone
 
Integrative Multi-Scale Analysis in Biomedical Data Science: Tools, Methods a...
Integrative Multi-Scale Analysis in Biomedical Data Science: Tools, Methods a...Integrative Multi-Scale Analysis in Biomedical Data Science: Tools, Methods a...
Integrative Multi-Scale Analysis in Biomedical Data Science: Tools, Methods a...Joel Saltz
 
NegBio: a high-performance tool for negation and uncertainty detection in rad...
NegBio: a high-performance tool for negation and uncertainty detection in rad...NegBio: a high-performance tool for negation and uncertainty detection in rad...
NegBio: a high-performance tool for negation and uncertainty detection in rad...Yifan Peng
 
PPT on BRAIN TUMOR detection in MRI images based on IMAGE SEGMENTATION
PPT on BRAIN TUMOR detection in MRI images based on  IMAGE SEGMENTATION PPT on BRAIN TUMOR detection in MRI images based on  IMAGE SEGMENTATION
PPT on BRAIN TUMOR detection in MRI images based on IMAGE SEGMENTATION khanam22
 
Comparitive study of brain tumor detection using morphological operators
Comparitive study of brain tumor detection using morphological operatorsComparitive study of brain tumor detection using morphological operators
Comparitive study of brain tumor detection using morphological operatorseSAT Journals
 

What's hot (15)

E-book Thesis Sara Carvalho
E-book Thesis  Sara CarvalhoE-book Thesis  Sara Carvalho
E-book Thesis Sara Carvalho
 
La2418611866
La2418611866La2418611866
La2418611866
 
2008 SPIE Photonics West
2008 SPIE Photonics West2008 SPIE Photonics West
2008 SPIE Photonics West
 
Multimodal Medical Image Fusion Based On SVD
Multimodal Medical Image Fusion Based On SVDMultimodal Medical Image Fusion Based On SVD
Multimodal Medical Image Fusion Based On SVD
 
Twenty Years of Whole Slide Imaging - the Coming Phase Change
Twenty Years of Whole Slide Imaging - the Coming Phase ChangeTwenty Years of Whole Slide Imaging - the Coming Phase Change
Twenty Years of Whole Slide Imaging - the Coming Phase Change
 
Detection of medical instruments project- PART 2
Detection of medical instruments project- PART 2Detection of medical instruments project- PART 2
Detection of medical instruments project- PART 2
 
Text Mining Radiology Reports for Deep Learning Radiology Images
Text Mining Radiology Reports for Deep Learning Radiology Images Text Mining Radiology Reports for Deep Learning Radiology Images
Text Mining Radiology Reports for Deep Learning Radiology Images
 
Brain tumor detection using image segmentation ppt
Brain tumor detection using image segmentation pptBrain tumor detection using image segmentation ppt
Brain tumor detection using image segmentation ppt
 
IRJET- Breast Cancer Prediction using Support Vector Machine
IRJET-  	  Breast Cancer Prediction using Support Vector MachineIRJET-  	  Breast Cancer Prediction using Support Vector Machine
IRJET- Breast Cancer Prediction using Support Vector Machine
 
Common Sense Based Joint Training of Human Activity Recognizers
Common Sense Based Joint Training of Human Activity RecognizersCommon Sense Based Joint Training of Human Activity Recognizers
Common Sense Based Joint Training of Human Activity Recognizers
 
Integrative Multi-Scale Analysis in Biomedical Data Science: Tools, Methods a...
Integrative Multi-Scale Analysis in Biomedical Data Science: Tools, Methods a...Integrative Multi-Scale Analysis in Biomedical Data Science: Tools, Methods a...
Integrative Multi-Scale Analysis in Biomedical Data Science: Tools, Methods a...
 
NegBio: a high-performance tool for negation and uncertainty detection in rad...
NegBio: a high-performance tool for negation and uncertainty detection in rad...NegBio: a high-performance tool for negation and uncertainty detection in rad...
NegBio: a high-performance tool for negation and uncertainty detection in rad...
 
Final Poster
Final PosterFinal Poster
Final Poster
 
PPT on BRAIN TUMOR detection in MRI images based on IMAGE SEGMENTATION
PPT on BRAIN TUMOR detection in MRI images based on  IMAGE SEGMENTATION PPT on BRAIN TUMOR detection in MRI images based on  IMAGE SEGMENTATION
PPT on BRAIN TUMOR detection in MRI images based on IMAGE SEGMENTATION
 
Comparitive study of brain tumor detection using morphological operators
Comparitive study of brain tumor detection using morphological operatorsComparitive study of brain tumor detection using morphological operators
Comparitive study of brain tumor detection using morphological operators
 

Viewers also liked

Revista Valorarte de la Asociación S.O.L.
Revista Valorarte de la Asociación S.O.L.Revista Valorarte de la Asociación S.O.L.
Revista Valorarte de la Asociación S.O.L.Asociación Sol
 
Technical qo s and kpi lebanon
Technical qo s  and kpi lebanonTechnical qo s  and kpi lebanon
Technical qo s and kpi lebanonharoudaki
 
Evaluation Of Mixtures Efficiency In Refrigerating Systems
Evaluation Of Mixtures Efficiency In Refrigerating SystemsEvaluation Of Mixtures Efficiency In Refrigerating Systems
Evaluation Of Mixtures Efficiency In Refrigerating SystemsTOYJOHN
 
حقيقة الاسلام والايمان ومنزلة العمل في الاسلام
حقيقة الاسلام والايمان ومنزلة العمل في الاسلامحقيقة الاسلام والايمان ومنزلة العمل في الاسلام
حقيقة الاسلام والايمان ومنزلة العمل في الاسلاممبارك الدوسري
 
Aurafix Pratik Kol Askısı
Aurafix Pratik Kol AskısıAurafix Pratik Kol Askısı
Aurafix Pratik Kol Askısıkolaskilari
 
Crónicas: Realidad y Educación
Crónicas: Realidad y EducaciónCrónicas: Realidad y Educación
Crónicas: Realidad y EducaciónAsociación Sol
 
The-CMO-Shift-to-Gaining-Business-Lift-Executive-Perspectives
The-CMO-Shift-to-Gaining-Business-Lift-Executive-PerspectivesThe-CMO-Shift-to-Gaining-Business-Lift-Executive-Perspectives
The-CMO-Shift-to-Gaining-Business-Lift-Executive-PerspectivesSanjay Bhattacharji
 
HR - Recruitment and Selection process importance
HR - Recruitment and Selection process importanceHR - Recruitment and Selection process importance
HR - Recruitment and Selection process importanceNeha Nagulkar Ghorad
 
проектная деятельность музея
проектная деятельность музеяпроектная деятельность музея
проектная деятельность музеяURFU
 
Post-Merger Cultural Integration
Post-Merger Cultural IntegrationPost-Merger Cultural Integration
Post-Merger Cultural IntegrationVirgilijus Dadonas
 
Exploración de Cuello y Cadenas Ganglionares
Exploración de Cuello y Cadenas GanglionaresExploración de Cuello y Cadenas Ganglionares
Exploración de Cuello y Cadenas GanglionaresManuel Meléndez
 

Viewers also liked (15)

Tutorial test
Tutorial  testTutorial  test
Tutorial test
 
Revista Valorarte de la Asociación S.O.L.
Revista Valorarte de la Asociación S.O.L.Revista Valorarte de la Asociación S.O.L.
Revista Valorarte de la Asociación S.O.L.
 
Technical qo s and kpi lebanon
Technical qo s  and kpi lebanonTechnical qo s  and kpi lebanon
Technical qo s and kpi lebanon
 
Evaluation Of Mixtures Efficiency In Refrigerating Systems
Evaluation Of Mixtures Efficiency In Refrigerating SystemsEvaluation Of Mixtures Efficiency In Refrigerating Systems
Evaluation Of Mixtures Efficiency In Refrigerating Systems
 
كيف تثقل ميزانك ؟
كيف تثقل ميزانك ؟كيف تثقل ميزانك ؟
كيف تثقل ميزانك ؟
 
حقيقة الاسلام والايمان ومنزلة العمل في الاسلام
حقيقة الاسلام والايمان ومنزلة العمل في الاسلامحقيقة الاسلام والايمان ومنزلة العمل في الاسلام
حقيقة الاسلام والايمان ومنزلة العمل في الاسلام
 
Aurafix Pratik Kol Askısı
Aurafix Pratik Kol AskısıAurafix Pratik Kol Askısı
Aurafix Pratik Kol Askısı
 
Crónicas: Realidad y Educación
Crónicas: Realidad y EducaciónCrónicas: Realidad y Educación
Crónicas: Realidad y Educación
 
The-CMO-Shift-to-Gaining-Business-Lift-Executive-Perspectives
The-CMO-Shift-to-Gaining-Business-Lift-Executive-PerspectivesThe-CMO-Shift-to-Gaining-Business-Lift-Executive-Perspectives
The-CMO-Shift-to-Gaining-Business-Lift-Executive-Perspectives
 
Trigonometr 11
Trigonometr 11Trigonometr 11
Trigonometr 11
 
HR - Recruitment and Selection process importance
HR - Recruitment and Selection process importanceHR - Recruitment and Selection process importance
HR - Recruitment and Selection process importance
 
проектная деятельность музея
проектная деятельность музеяпроектная деятельность музея
проектная деятельность музея
 
New driving forces in Asia
New driving forces in Asia New driving forces in Asia
New driving forces in Asia
 
Post-Merger Cultural Integration
Post-Merger Cultural IntegrationPost-Merger Cultural Integration
Post-Merger Cultural Integration
 
Exploración de Cuello y Cadenas Ganglionares
Exploración de Cuello y Cadenas GanglionaresExploración de Cuello y Cadenas Ganglionares
Exploración de Cuello y Cadenas Ganglionares
 

Similar to Poster draft 6_17_13_Final_For_Printing

APPLICATION OF CNN MODEL ON MEDICAL IMAGE
APPLICATION OF CNN MODEL ON MEDICAL IMAGEAPPLICATION OF CNN MODEL ON MEDICAL IMAGE
APPLICATION OF CNN MODEL ON MEDICAL IMAGEIRJET Journal
 
Breast cancer detection using ensemble of convolutional neural networks
Breast cancer detection using ensemble of convolutional neural networksBreast cancer detection using ensemble of convolutional neural networks
Breast cancer detection using ensemble of convolutional neural networksIJECEIAES
 
Preprocessing Techniques for Image Mining on Biopsy Images
Preprocessing Techniques for Image Mining on Biopsy ImagesPreprocessing Techniques for Image Mining on Biopsy Images
Preprocessing Techniques for Image Mining on Biopsy ImagesIJERA Editor
 
LEAF DISEASE DETECTION USING IMAGE PROCESSING AND SUPPORT VECTOR MACHINE (SVM)
LEAF DISEASE DETECTION USING IMAGE PROCESSING AND SUPPORT VECTOR MACHINE (SVM)LEAF DISEASE DETECTION USING IMAGE PROCESSING AND SUPPORT VECTOR MACHINE (SVM)
LEAF DISEASE DETECTION USING IMAGE PROCESSING AND SUPPORT VECTOR MACHINE (SVM)Journal For Research
 
Detection of Cancer in Pap smear Cytological Images Using Bag of Texture Feat...
Detection of Cancer in Pap smear Cytological Images Using Bag of Texture Feat...Detection of Cancer in Pap smear Cytological Images Using Bag of Texture Feat...
Detection of Cancer in Pap smear Cytological Images Using Bag of Texture Feat...IOSR Journals
 
IRJET - Machine Learning Applications on Cancer Prognosis and Prediction
IRJET - Machine Learning Applications on Cancer Prognosis and PredictionIRJET - Machine Learning Applications on Cancer Prognosis and Prediction
IRJET - Machine Learning Applications on Cancer Prognosis and PredictionIRJET Journal
 
IRJET - Classification of Cancer Images using Deep Learning
IRJET -  	  Classification of Cancer Images using Deep LearningIRJET -  	  Classification of Cancer Images using Deep Learning
IRJET - Classification of Cancer Images using Deep LearningIRJET Journal
 
Brain Image Fusion using DWT and Laplacian Pyramid Approach and Tumor Detecti...
Brain Image Fusion using DWT and Laplacian Pyramid Approach and Tumor Detecti...Brain Image Fusion using DWT and Laplacian Pyramid Approach and Tumor Detecti...
Brain Image Fusion using DWT and Laplacian Pyramid Approach and Tumor Detecti...INFOGAIN PUBLICATION
 
A Dualistic Sub-Image Histogram Equalization Based Enhancement and Segmentati...
A Dualistic Sub-Image Histogram Equalization Based Enhancement and Segmentati...A Dualistic Sub-Image Histogram Equalization Based Enhancement and Segmentati...
A Dualistic Sub-Image Histogram Equalization Based Enhancement and Segmentati...inventy
 
Brain Tumor Classification using Support Vector Machine
Brain Tumor Classification using Support Vector MachineBrain Tumor Classification using Support Vector Machine
Brain Tumor Classification using Support Vector MachineIRJET Journal
 
Plant Leaf Diseases Identification in Deep Learning
Plant Leaf Diseases Identification in Deep LearningPlant Leaf Diseases Identification in Deep Learning
Plant Leaf Diseases Identification in Deep LearningCSEIJJournal
 
Simplified Knowledge Prediction: Application of Machine Learning in Real Life
Simplified Knowledge Prediction: Application of Machine Learning in Real LifeSimplified Knowledge Prediction: Application of Machine Learning in Real Life
Simplified Knowledge Prediction: Application of Machine Learning in Real LifePeea Bal Chakraborty
 
Brain Tumor Detection Using Artificial Neural Network Fuzzy Inference System ...
Brain Tumor Detection Using Artificial Neural Network Fuzzy Inference System ...Brain Tumor Detection Using Artificial Neural Network Fuzzy Inference System ...
Brain Tumor Detection Using Artificial Neural Network Fuzzy Inference System ...Editor IJCATR
 
Lung Cancer Detection using Machine Learning
Lung Cancer Detection using Machine LearningLung Cancer Detection using Machine Learning
Lung Cancer Detection using Machine Learningijtsrd
 
DIRECTVIEW EVP Plus Software
DIRECTVIEW EVP Plus SoftwareDIRECTVIEW EVP Plus Software
DIRECTVIEW EVP Plus SoftwareCarestream
 

Similar to Poster draft 6_17_13_Final_For_Printing (20)

APPLICATION OF CNN MODEL ON MEDICAL IMAGE
APPLICATION OF CNN MODEL ON MEDICAL IMAGEAPPLICATION OF CNN MODEL ON MEDICAL IMAGE
APPLICATION OF CNN MODEL ON MEDICAL IMAGE
 
Breast cancer detection using ensemble of convolutional neural networks
Breast cancer detection using ensemble of convolutional neural networksBreast cancer detection using ensemble of convolutional neural networks
Breast cancer detection using ensemble of convolutional neural networks
 
Prototype System to Detect Skin Cancer Through Images
Prototype System to Detect Skin Cancer Through ImagesPrototype System to Detect Skin Cancer Through Images
Prototype System to Detect Skin Cancer Through Images
 
Preprocessing Techniques for Image Mining on Biopsy Images
Preprocessing Techniques for Image Mining on Biopsy ImagesPreprocessing Techniques for Image Mining on Biopsy Images
Preprocessing Techniques for Image Mining on Biopsy Images
 
LEAF DISEASE DETECTION USING IMAGE PROCESSING AND SUPPORT VECTOR MACHINE (SVM)
LEAF DISEASE DETECTION USING IMAGE PROCESSING AND SUPPORT VECTOR MACHINE (SVM)LEAF DISEASE DETECTION USING IMAGE PROCESSING AND SUPPORT VECTOR MACHINE (SVM)
LEAF DISEASE DETECTION USING IMAGE PROCESSING AND SUPPORT VECTOR MACHINE (SVM)
 
Detection of Cancer in Pap smear Cytological Images Using Bag of Texture Feat...
Detection of Cancer in Pap smear Cytological Images Using Bag of Texture Feat...Detection of Cancer in Pap smear Cytological Images Using Bag of Texture Feat...
Detection of Cancer in Pap smear Cytological Images Using Bag of Texture Feat...
 
A01110107
A01110107A01110107
A01110107
 
R-PI BASED DETECTION OF LUNG CANCER USING MRI IMAGE
R-PI BASED DETECTION OF LUNG CANCER USING MRI IMAGER-PI BASED DETECTION OF LUNG CANCER USING MRI IMAGE
R-PI BASED DETECTION OF LUNG CANCER USING MRI IMAGE
 
IRJET - Machine Learning Applications on Cancer Prognosis and Prediction
IRJET - Machine Learning Applications on Cancer Prognosis and PredictionIRJET - Machine Learning Applications on Cancer Prognosis and Prediction
IRJET - Machine Learning Applications on Cancer Prognosis and Prediction
 
IRJET - Classification of Cancer Images using Deep Learning
IRJET -  	  Classification of Cancer Images using Deep LearningIRJET -  	  Classification of Cancer Images using Deep Learning
IRJET - Classification of Cancer Images using Deep Learning
 
Brain Image Fusion using DWT and Laplacian Pyramid Approach and Tumor Detecti...
Brain Image Fusion using DWT and Laplacian Pyramid Approach and Tumor Detecti...Brain Image Fusion using DWT and Laplacian Pyramid Approach and Tumor Detecti...
Brain Image Fusion using DWT and Laplacian Pyramid Approach and Tumor Detecti...
 
A Dualistic Sub-Image Histogram Equalization Based Enhancement and Segmentati...
A Dualistic Sub-Image Histogram Equalization Based Enhancement and Segmentati...A Dualistic Sub-Image Histogram Equalization Based Enhancement and Segmentati...
A Dualistic Sub-Image Histogram Equalization Based Enhancement and Segmentati...
 
Brain Tumor Classification using Support Vector Machine
Brain Tumor Classification using Support Vector MachineBrain Tumor Classification using Support Vector Machine
Brain Tumor Classification using Support Vector Machine
 
Plant Leaf Diseases Identification in Deep Learning
Plant Leaf Diseases Identification in Deep LearningPlant Leaf Diseases Identification in Deep Learning
Plant Leaf Diseases Identification in Deep Learning
 
Simplified Knowledge Prediction: Application of Machine Learning in Real Life
Simplified Knowledge Prediction: Application of Machine Learning in Real LifeSimplified Knowledge Prediction: Application of Machine Learning in Real Life
Simplified Knowledge Prediction: Application of Machine Learning in Real Life
 
Madhavi tippani
Madhavi tippaniMadhavi tippani
Madhavi tippani
 
COMPARATIVE ANALYSIS OF DIFFERENT MACHINE LEARNING ALGORITHMS FOR PLANT DISEA...
COMPARATIVE ANALYSIS OF DIFFERENT MACHINE LEARNING ALGORITHMS FOR PLANT DISEA...COMPARATIVE ANALYSIS OF DIFFERENT MACHINE LEARNING ALGORITHMS FOR PLANT DISEA...
COMPARATIVE ANALYSIS OF DIFFERENT MACHINE LEARNING ALGORITHMS FOR PLANT DISEA...
 
Brain Tumor Detection Using Artificial Neural Network Fuzzy Inference System ...
Brain Tumor Detection Using Artificial Neural Network Fuzzy Inference System ...Brain Tumor Detection Using Artificial Neural Network Fuzzy Inference System ...
Brain Tumor Detection Using Artificial Neural Network Fuzzy Inference System ...
 
Lung Cancer Detection using Machine Learning
Lung Cancer Detection using Machine LearningLung Cancer Detection using Machine Learning
Lung Cancer Detection using Machine Learning
 
DIRECTVIEW EVP Plus Software
DIRECTVIEW EVP Plus SoftwareDIRECTVIEW EVP Plus Software
DIRECTVIEW EVP Plus Software
 

Poster draft 6_17_13_Final_For_Printing

  • 1. Introduction Chemical tissue staining is currently used for studying cancer pathology: tissue cores extracted from a patient are thinly sliced and stained using chemical dyes to highlight different cell types. The stained tissue samples are examined under a microscope to diagnose disease. The Bioimaging Science and Technology group at the Beckman Institute uses infrared spectroscopy to directly collect chemical information. The goal is to use this quantitative information to improve disease diagnosis by providing more accurate information to pathologists. Methods and results The first step in our program is loading the ENVI header information. This includes all of the properties of the image, such as the x and y resolution, the resolution of the infrared spectra (number of bands), and the units of measurement used. These parameters define how the random forest is constructed. Our program stores them as variables for later use. Due to the large file size of the ENVI images, an entire image cannot be stored in memory, and must rather be streamed from the hard drive and classified sequentially. This is known as out-of-core processing. Our program feeds individual chunks into the classifier one by one, building up a fully classified image. Because four collaborators all worked on the same program concurrently, utilizing a distributed version control system called Git was extremely important. This allowed us to keep track of and comment on any edits made in our code. Git also allows for branching, so separate users can work independently, then merge all the branches back into one main master branch. Our code was managed using Cmake, which allows us to easily link our programs with external libraries, such as Qt and ALGLIB, which were used for user-interface design and classification. Spectroscopic Images This resulting infrared signal produces a digital image, in which each pixel corresponds to a frequency in a spectrum graph where the point will vibrate. This data is stored using the ENVI (Environment for Visualizing Images) format. This large, specialized file is then run through a classification program using the Random- Forest algorithm (Figure 5). This classifier analyzes each pixel based on its spectrum and surroundings. David Bergvelt and Max Li, under the supervision of David Mayerich Led by Professor Bhargava at Bioimaging Science and Technology group, Beckman Institute, University of Illinois at Urbana-Champaign Digitizing cancer pathology research Acknowledgments We would like to thank Professor Bhargava for giving us the opportunity to work alongside his Bioimaging group, and also we would like to thank David Mayerich for giving us the chance to try our hands at programming, and guiding us along our way. We would also like to thank David Bergandine for sponsoring us and advising us throughout the I-STEM program, and Ms. Williams and Mrs. Destefano for organizing the I-STEM program. Experience We found that the hands-on experience of working with programming in C++ helped a lot in learning the language. Learning how to use Git and Cmake, two industry standard applications, will be very useful in the future. It was also a very interesting experience connecting the topics of cancer pathology and statistical analysis together using programming languages. Working along with Dr. Mayerich in Professor Bhargava’s group gave us the experience of working in a research group doing cutting edge research related how the future of cancer research and diagnosis will look like. Overall, the I-STEM experience was excellent, and we hope to continue working with Dr. Mayerich and the Bhargava group in the future. Aim Our project focuses on building a C++ program which allows the user to load a spectroscopic image of a tissue sample, feed it into the Random-Forest algorithm, and output a classified image showing differences in tissue type. The final program will incorporate a graphical user interface, making it user-friendly and useful to a wide range of researchers. We hope that this will encourage researchers to adopt these quantitative methods in their own research and diagnostic practices. In particular, we expect that this type of technology will be useful for accurate disease diagnosis in hospitals. Fig. 1: Image of a breast tumor biopsy stained using various chemical methods (left) and an image of the same biopsy stained digitally after spectroscopic imaging and classification (right). Fig. 5: A Random Forest is composed of hundreds of decision trees, where each tree selects a cell type based on a random subset of features of a single spectrum. Each tree then “votes” for its selected class. Fig. 4: Each pixel making up the image of the digitally stained cell carries its own infrared spectrum (a and b). The image is the classified into various cell types, such as epithelium, fibroblasts, etc. The accuracy of the classifier is given using the precision: the ratio of cell types that are correctly classified. 1200 1800 2400 3000 3600 0.0 0.1 0.2 0.3 0.4 Absorbance(a.u.) Wavenumber (cm-1 ) Fig. 2: Schematic of a mid-infrared spectroscopy setup. A detector is used to measure the intensity of the infrared light as it passes through the specimen, the independent variable being the position of a movable mirror. Fig. 3:The position of the mirror is plotted along with the associated light intensity, and a Fourier Transformation if applied, transforming it into a function of wavelength and intensity. 1.00 0.00 Fig. 6: Final result of our classification. The color scale represents the probability of tissue being epithelium tissue, where dark red = very strong probability of epithelium tissue, dark blue = very weak probability of epithelium. (a) (b) (c)

Editor's Notes

  1. 1