Is the increasing availability of automated image analysis a possibility to strengthen the application of diffusion-MRI as a biometric parameter, and to enhance the future of image biobanks? Or is this evolution threatening the position of radiologists as medical doctors. Is a redefinition of radiologist as computer technicians inevitable?
2. Automated Image Analysis: what is it about?
Automatic reading of
images by A.I.
Big Data & Deep Learning
Radiological &
non-radiological images
Imaging Biomarkers
Radiomics
Radiogenomics
3. Personalised Disease Evolution
• What do we want to know for each patient?
– Is a tumor present?
– Is it aggressive?
– Is it focally treatable?
– Is it radiosensitive?
– Will it metastasize?
• Evidence based medicine
– has succeeded in defining effective therapeutics for large populations
– is lacking when applied to small subpopulations (“precision medicine”)
– is insufficient when applied to the individual level (“personalized medicine”).Krishnaraj A et al. The future of imaging biomarkers in radiologic practice.
J Am Coll Radiol 2014;11:20-23
5. Biobanks are crucial
Long-term storage and retrieval of tissue samples is needed in biobanks.
Human biobanks include biological material of healthy subjects and patients with specific
pathologies, most cancer-related.
Data from radiological imaging are not included in the human biobanks.
Association between phenotype (imaging) and genotype will become possible by means of
imaging biomarkers.
Quantitative medical imaging with identification of imaging biomarkers represents a crucial part
of personalised medicine.
Several imaging biobank projects have been started.
6. Radiomics vs. Radiogenomics
RADIOMICS
Automated extraction, storage and
analysis of a large amount of imaging
features (morphology and imaging
biomarkers)
Cloud-based deep-learning techniques
Conversion of images to mineable data, in
order to create accessible databases
(biobanks)
...and to reveal quantitative predictive or
prognostic associations between images
and medical outcomes.
RADIOGENOMICS
Imaging findings can be considered as the
phenotypic expression of a patient,
which can be correlated to the
genotype.
Radiogenomics is the extension of
radiomics, aiming to identify a link
between genotype and phenotype
imaging.
7. “To explore the full
potential of radiomics, we
have to enter the era of
big data, team science and,
most of all, the new age of
imaging bioinformatics”
Dr. Hricak
8. Imaging Biobanks
QIBA
Founded in 2007 by RSNA
Mission: to improve the value and practicality of
quantitative imaging biomarkers by reducing
variability across devices, patients, and time.
Support from volunteer committee members from
academia, medical device, pharmaceutical and other
business sectors, and government.
4 Modality-based committees: Q-CT, Q-MR, Q-NM, Q-US
10 Biomarker committees
EIBALL
Founded by ESR in March 2015
Coordination of ESR activities concerning
imaging biomarkers
Merging of activities of
ESR Subcommittee on Imaging
Biomarkers
ESR Working Group on Personalised
Medicine
ESR-EORTC Working Group
10. DWI as biomarker of cancer
Toronto 2008: Consensus and Recommendations on use of DW-MRI as
cancer imaging biomarker
DWI-MRI should be tested as imaging biomarker in clinical trials
DWI-MRI measurements should be compared with histologic indices
Standards for measurement, analysis and display are needed
Annotated data should be made available
MRI vendors should be engaged in processPadhani AR, Liu G, Mu-Koh D, et al. Diffusion-Weighted Magnetic Resonance Imaging as a Cancer Biomarker: Consensus and Recommendations.
Neoplasia (New York, NY). 2009;11(2):102-125.
11. Advantages of DWI-MRI
Improved tissue characterisation (malignant vs. benign)
Monitoring of treatment after chemotherapy or radiation
DD of post-therapeutic changes from residual active tumor
Detection of recurrent cancer
Prediction of treatment outcome
Tumour staging
Detection of lymph node involvement
12. Remaining challenges for DWI to assess cancer
Divergence among and between vendors on data measurements/analysis
No accepted standards for measurements and analysis
Multiple data acquisition protocols depending on body part and usage of data
Qualitative to quantitative assessments
Lack of understanding of DW-MRI at a microscopic level
Incomplete validation and documentation of reproducibility
Divergent nomenclature and symbols
Lack of multicenter working methodologies, accepted quality assurance (QA)
standards, and physiologically realistic phantoms
14. Hybrid imaging: PET/MRI vs. PET/CT
Current Status of Hybrid PET/MRI in Oncologic Imaging
Andrew B. Rosenkrantz et al., American Journal of Roentgenology 2016 206:1, 162-172
15. 85-year-old man with prostate cancer who underwent initial staging workup that showed metastases to bone.
Standard bone scan shows one metastatic lesion in left acetabulum (solid arrow) and small subtle lesion in upper
thoracic spine (dashed arrow) which was attributed to degenerative spine disease. ANT = anterior view, POS =
posterior view.
16. NaF PET/MR image
obtained 3 weeks later
reveals nine metastatic
lesions (circles), showing
higher sensitivity of
PET/MRI.
Lesion in T2 spinous process on PET/MRI (dashed
arrows) corresponds to small subtle lesion on
bone scan
17. Radiologists of the future
Medisch Contact, 5 dec 2016
Healthcare in Europe, 28 nov 2016
“Machine learning can discover whether certain
image data point towards certain diseases; it can
discover correlations as yet unknown, or confirm
suspected correlations respectively by analysing the
large amounts of data.”
18. Are biomarkers and A.I. threatening radiology?
“Automatic reading of images by A.I. is not developed enough to replace
the trained and experienced observer with his/her ability to interpret and
judge during image reading sessions”
“Nevertheless, subjective, and therefore, qualitative interpretations are
observer dependent and highly variable, and variability inevitably
degrades outcomes in healthcare in general”
“Extracting objective, quantitative results from medical images is one way
to reduce the variability...and thus will improve patient outcomes”.
Siegfried Trattnig, Chair of EIBALL
19. 12 Opinion Leaders’ ideas
Paul M. Parizel
Geraldine McGinty
Lluis Donoso Bach
Luis Marti-Bonmati
Nicola Strickland
Koenraad Mortele
Wiro Niessen
Charles Kahn
Marion Smits
Peter Mildenberger
Mario Maas
Vasileios Katsaros
20. Redefinition of radiology
Copyright Dr. E. R. Ranschaert
•Multidisciplinary integration, expert consultancy, therapy guidance
•Disease-focused approach, gatekeeping, lean approach
Workflow
management
•Deal with errors, reduce failure and mistake
•Measure outcomes & improve performanceQuality management
•Embrace power of digital networks, cloud services, big data, deep
learning & Artificial IntelligenceImaging Informatics
•Structured reporting (+ coding), multimedia, actionable reports
•Patient-oriented approach, lay-language, open notesCommunication
Precision medicine
• Functional imaging & imaging biomarkers, radiomics & radiogenomics,
integrated diagnostics
• Image-guided interventions