1. CSC 578 Neural Networks and Deep Learning
6-3. Convolutional Neural Networks (3)
Noriko Tomuro

2. Learning in CNN
Cross-correlation
Convolution

3. Learning in CNN
Convolution Output
Convolution Output Simplified

4. Learning in CNN formulation

6. Forward Propagation

8. CNN BackPropagation

9. CNN BackPropagation

14. CNN BackPropagation

15. References Useful links:

16. CNN Applications
Image Classification

17. CNN Applications

18. CNN Applications

19. CNN Applications

20. CNN Applications

21. CNN Applications

22. CNN Applications

23. CNN in Medical Imaging Biomedical Data: Examples of biomedical data:
The omics: Genome, Transcriptome
Imaging: Pathology, Radiology

24. CNN in Medical Imaging Volumetric data => 3D images
MRI is multimodal
Complementary information
Size, location, morphology

25. CNN in Medical Imaging Two types of image features Semantic features
Image features manually annotated by radiologist
Radiologist-dependent interpretation of an image
Most likely qualitative
Computational features
(semi)-automatically extracted from an image
Most likely quantitative

26. CNN in Medical Imaging
Semantic features:
Edge shape
Smooth
Lobulated

27. CNN in Medical Imaging Computational Features: Texture Features
Shape Features
Edge Sharpness Features

28. CNN in Medical Imaging
Given CT scan of a patient can we predict survival?
Survival score?
Conventional
Approach

29. CNN in Medical Imaging

30. CNN in Medical Imaging

31. CNN in Medical Imaging