Artificial intelligence has wide applications in medical imaging for analyzing images. Deep learning is commonly used for tasks like automated detection, segmentation, and classification of lesions or abnormalities from medical images. While AI systems have achieved high accuracy in some narrow tasks like nodule detection, integrating AI safely into clinical practice poses challenges regarding data privacy, system transparency, and regulatory approval. Overall AI has potential to improve healthcare by making imaging interpretation more efficient and accurate, but careful management of technology and change is needed.
2. 2
MI: deep learning activity
Widely employed methodology for analyzing medical images
The significant frequency and clinical impact of human errors.
To improve the reliability and efficiency of imaging interpretation.
Litjens et al, A Survey on Deep Learning in Medical Image Analysis, June 2017
4. 4
AI: definition
Intelligence
overt ability to
solve specific
problems and an
innate ability to
learn solutions for
new problems
AI
artificial entity
capable of solving
problems and
learning solutions
for new problems
AI seeks to improve computer systems until their
function is equal to, or greater than, that of a
human performing the same task
A.B. Simmons, S.G. Chappell, Artificial intelligence-definition and practice, IEEE J. Ocean. Eng. 13 (2) (1988) 14โ42.
M I Fazal et al. The past, present and future role of artificial intelligence in imaging, European Journal of Radiology 105 (2018) 246โ250
https://www.maxpixel.net/Brain-Artificial-Intelligence-
Control-Think-3382507
5. 5
โ
Artificial entity to be able:
โ To perceive its environment
โ
To detect input data and its
parameters
โ To search and perform pattern
recognition
โ
To identify and recognise features
of the problem
โ To plan and execute an
appropriate course of action
and perform inductive
reasoning to derive general
principles
โ
To learn from experience
AI: definition
M. Minsky, Steps toward artificial intelligence, Proc. IRE 49 (1) (1961) 8โ30.
https://cdn-images-1.medium.com/max/1600/1*1MDX25tDkQaoF7vmKOKxrg.png
6. 6
Medical imaging & AI: the roots
โ
1956. First AI conference at Durtmouth College
โ John McCarthy coined the term โArtificial Intelligenceโ
โ
1956-1974. First golden years, Cold War era
โ Translation Russian into English
โ
1972. MYCIN expert system (LISP)
โ to assist physicians in the diagnosis of infectious diseases
โ
1974 โ 1980. First AI winter. Disillusionment. Could not solve practical problems
โ ALPAC and Lighthill reports, loss of research funding in USA and UK
โ
1980 โ 1987. Designed by humans.
โ Sequential application of low-level pixel processing and mathematical modeling
โ
1987 - 1993. Second AI winter. Could not teach expertโs knowledge
โ Transition between specific LISP machine (failure) to general purpose computers
โ
1996. DeepBlue. Man vs Machine, Chess (512 Core)
โ the fastest computer to face a world chess champion
โ
At the end of the 1990s. Trained by computer. Crucial step-handcrafted features
โ Supervised techniques, where training data is used to develop MIA systems.
โ The basis of many successful commercially available medical image analysis systems.
โ Segmentation based on vector extraction; Atlas as a optimal decision boudaries
โ
2011. DeepQA. Man vs Machine, Jeopardy(2 880 Core โ Millions books of data)
โ computer system capable of answering questions posed in natural language
โ
2012, AlexNet won ImageNet challenge (16 000 Core -10 Milliions images)
โ Recognition of Convolutional Neural Network as technique of choice
Litjens et al, A Survey on Deep Learning in Medical Image Analysis, June 2017
N Mehta, M. V. Devarakonda, Machine learning, natural language programming, and electronic health records: The next step in the artificial intelligence journey?, J Allergy Clin
Immunol, June 2018
Fei Jiang et al. Stroke Vasc Neurol 2017;2:230-243
7. 7
Can a robot get into the University of Tokyo?
Todai robot project 2011 - 2021
A challenge perfectly
suited for Japan - using
small data to improve
accuracy
8. 8
MIA & AI: Augmented Medical Imaging
Image
processing
(IP)
Computer
vision
(CV)
Natural
language
processing
(NLP)
Machine learning
Diagnosis more efficient and accurate
Ch Liew, The future of radiology augmented with Artificial Intelligence: A strategy for success, European
Journal of Radiology 102 (2018) 152โ156
9. 9
MI & AI: actual context
Ch Liew, The future of radiology augmented with Artificial Intelligence: A strategy for success, European Journal of Radiology 102 (2018) 152โ156
Fei Jiang et al. Stroke Vasc Neurol 2017;2:230-243
โ
Radiology pioneering work
โ Medical imaging perception since 1980s.
โ
Radiologists are domain experts
โ Medical imaging, medical physics and radiation safety.
โ
Image substancial innovations
โ In the past 5โ10 years, new innovations in imaging from deep learning methods of image classification.
โ
Accuracy rates vs human
โ Current artificial neural networks have accuracy rates which surpass those of human radiologists in narrow-
basedtasks such as nodule detection.
Fei Jiang et al. Stroke Vasc Neurol 2017;2:230-243
10. 10
Radiology&AI:impacts
โ
Shortage of specialist: detection and
prediction automation
โ To automate detection nodules in CT scans and
pneumonia on chest x-rays
โ To predict the behavior of pre-cancerous lesions to reduce
the number of unnecessary invasive tests such biopsy
โ Screening for cancer
โ
Accuracy diagnosis: intelligence
augmentation
โ Hybrid intelligence: higher levels of accuracy in diagnosis
โ
Precision medicine: precision diagnostics and
big data
โ To mine huge data to link gene expression to imaging
features
โ Genetic and imaging biomaker correlation
โ
Assisted agent: radiological decision support
systems
โ Imaging studies have increased over the last two
decades, almost doubling every ten years
โ Reduce information overload and burnout (Radiologist
interpret one image 3-4s.)
โ Aid of rapid detection of emergency conditions
Ch Liew, The future of radiology augmented with Artificial Intelligence: A strategy for success, European
Journal of Radiology 102 (2018) 152โ156
Source: www.wired.com
11. 11
Radiology&AI: strategy
โ
To cut costs for patients and
insurance organizations
โ
Differentiation value
โ Accuracy
โ Increase access to specialist
services
โ Reduce time to process
interpretation
โ
Big data
โ Molecular imaging
โ โomicsโ: radiomics,
radiogenomics
โ Large population cancer screening
Ch Liew, The future of radiology augmented with Artificial Intelligence: A strategy for success, European
Journal of Radiology 102 (2018) 152โ156
12. 12
Medical imaging &AI: responsabilities
โ
Chief information Officer (CIO)
โ Safely and effective implementation
โ Patient safety and privacy
โ Systems interoperability (EHR,PACS, RIS ...)
โ Define frameworks and guidelines to
implement AI systems
โ Set standards to validate technology
โ
Chief data officer
โ safeguard the use of data for validation
and training of machine learning systems
โ data governance issues
โ
Medical imaging specialist
โ To generate valid data sets to training ML
models
โ To define ML use cases
โ To perform beta tests
โ To be trained
Ch Liew, The future of radiology augmented with Artificial Intelligence: A strategy for success, European
Journal of Radiology 102 (2018) 152โ156
13. 13
Medical Imaging &AI: technology
โ
Computational hardware
โ
Connectivity bandwith
โ
Cloud security
โ
Storage capacity
โ
Speech recognition
โ
Update and upgraded
systems
โ
Capital investment
Ch Liew, The future of radiology augmented with Artificial Intelligence: A strategy for success, European
Journal of Radiology 102 (2018) 152โ156
14. 14
Radiology&AI: implementation
Automated image
segmentation
โ Lesion detection
โ Measurement
โ Labelling
โ Historical
comparaison
Generating reports
โ NL processing
โ NL generation
โ Image
classification
Semantic error
detection
โ NLP agents as a
second reader to
warn about
semantic errors
Data mining for
research
โ NLP to build
searchables
databases
Business intelligence
โ Real-time
dashboards
โ Alert systems
โ Workflow analysis
โ Performance
assessment
Emergent
disciplines
โ omics
Ch Liew, The future of radiology augmented with Artificial Intelligence: A strategy for success, European
Journal of Radiology 102 (2018) 152โ156
15. 15
MI&AI: other considerations
โ
Medical imaging most
published subject
โ Deep learning in
healthcare
โ
New roles and
capabilites
โ Displacement of traditional
radiological tasks
โ Data scientist as early
carrier in medicine
โ
Regulatory
Ch Liew, The future of radiology augmented with Artificial Intelligence: A strategy for success, European
Journal of Radiology 102 (2018) 152โ156
16. 16
Medical Imaging&AI: challenges
โ
Security
โ AI system must be validated to be
safe, accurate and infallible before
it can be used on patients.
โ
Failure & Judicial
transparency
โ AI system must be auditable, not
black-box algorithms
โ
Privacy
โ AI systems should have access
only to relevant personal health
information
โ AI systems and companies must be
auditable and to ensure
compliance within the framework
of patient consent
Ch Liew, The future of radiology augmented with Artificial Intelligence: A strategy for success, European
Journal of Radiology 102 (2018) 152โ156
Source: fujitsu
17. 17
Global vision: go ahead
โ
To generate roadmaps for
the guided progress of AI in
medical imaging
โ
To define frameworks for
ethical, regulatory policy and
quality assurance
โ
To establish partnerships
with industry
โ
To facilitate global
collaboration between
medical, engineering and
data scientists
Ch Liew, The future of radiology augmented with Artificial Intelligence: A strategy for success, European
Journal of Radiology 102 (2018) 152โ156
Source: UM