1. Diagnosing heart diseases with deep neural networks
Recommending picking out he most interesting and important parts from your documentation.

2. Freelancer software / machine learning
Julian de Wit
Freelancer software / machine learning
Technical University Delft / TNO
Software engineering
Love biologically inspired computing
Last few years neural net “revolution”
Turn academic ideas into practical apps
Medical, documents, radar, plant grading
My background

3. Diagnose heart disease challenge Deep learning Solution discussion
Agenda
Diagnose heart disease challenge
Deep learning
Solution discussion
Results
Feel free to ask questions during talk !

4. Second national data science bowl Kaggle.com / Booz Allen Hamilton
Challenge
Second national data science bowl
Kaggle.com / Booz Allen Hamilton
This year’s challenge
Automate manual 30min clinical procedure
Ca patients/year in USA
Estimate heart volume based on MRI’s
Ratio systole/diastole is ‘health’ predictor
750 teams
$ prize money

5. Competition platform for ‘data scientists’
Challenge
Kaggle.com
Competition platform for ‘data scientists’
Challenges hosted for companies
Prize money and exposure
registered ‘competitors’
Learn: Always someone smarter than you !
Today’s state of the art is tomorrow’s baseline!

6. Given: MRI’s, metadata, train-volumes
Train 700, Test: 1000 patients, imgs
Estimate volume of left ventricle
Challenge

7. Image data → Deep Learning (CNN) Neural networks 2.0
Don’t believe ALL the hype
Structured data → feature engineering + Tree/Lin
Great when “perception” data is involved
Spectacular results with image analysis
My take: “Super human” with a twist
Deep learning

8. Deep learning

9. ‘Vanilla’ architecture.
Approach used by many teams (ie. #2 Ghent university)
Input slices, regress on provided volumes
Solution
123ml

10. Less publicized approach: Segment image.
Integrate estimated areas into volume using metadata.
Solution
Problem: ‘No annotations provided.’ Sunnybrook/hand

11. Segmentation : Traditional architecture bad fit
Every layer is higher level features less spatial info (BOW)
Per pixel classification possible coarse due to spatial loss
Cumbersome! 256 x 256 x classifications.
Solution

12. Segmentation : Fully convolutional architecture + upscale
Efficient. Classify all pixels at once
Still problem spatial bottleneck at bottom : coarse
Solution

13. Segmentation : U-net architecture
Skip connection give more detail in segmentation output
Author works at Deepmind health now
Solution

14. Segmentation results impressive. Machine did exactly what it was told.
Solution
Segmentation results impressive.
Machine did exactly what it was told.
Confused with uncommon examples < 1%.
Remedy : Active learning
Nice property : brightness == (un)certainty

15. Dirty secret: MUCH data cleaning Slice order Missing slices
Solution
Dirty secret: MUCH data cleaning
Slice order
Missing slices
Out of bound slices
Wrong orientation
Missing frames
BAD ground truth volumes
Gradient boosting “calibration” procedure
Not relevant in real setting. Just rescan MRI.

16. Sub 10ml MAE → clinically significant Many improvements possible :
Results
Result:
3rd place
Only 1 model. No ensemble.
Sub 10ml MAE → clinically significant
Many improvements possible :
More, cleaner train data
Expert annotations
Active learning

17. Many other deep learning medical successes
Results
Many other deep learning medical successes
Example: Retinopathy challenge
For bulk as good as expert doctors
Solution in use by companies already

18. Deep learning for medical imaging
Summary
Deep learning for medical imaging

20. EINDE....

21. Diagnosing heart diseases
with deep neural networks

22. Competition platform for ‘data scientists’
Kaggle.com
Competition platform for ‘data scientists’
Challenges hosted for companies
Prize money and exposure
registered competitors
Learn. Always someone smarter than you !
Today’s state of the art is tomorrow’s baseline!
Competition

23. Freelancer software / machine learning Technical University Delft : SE
Julian de Wit
Freelancer software / machine learning
Technical University Delft : SE
Biologically inspired computing / AI
Since 2006 heavily re-interested in neural nets
Looking for opportunities to test and bring in practice
My background

24. Approach

25. Use provided volumes to calibrate Remove systematic errors
Calibration
Use provided volumes to calibrate
Remove systematic errors
Use Gradient Booster on residuals
Top 5 -> top 3
Beware of overfitting

26. Every pixel: Left ventricle Yes/No Use convolutional neural network
Sunnybrook too simplistic
Train with hand-labeled segmentations
Reverse engineer how to label
Fix systematic errors with calibration against provided volumes.
Approach

27. Competition

28. Use DICOM info to make images uniform Crop around heart 180x180
Contrast stretch
Preprocessing

29. Hand labeling with own tool Big performance limiting factor
Could not find how to do it exactly

30. Cat!

31. Cat !

32. Grass

35. Uncertainty based on stdev in error as a function of size.
Submission
CRPS
Uncertainty based on stdev in error as a function of size.
Model provided uncertainty.
However does not account for uncertainty in labels
Example: patient 429. Error of 89ml !!!
Provided label was wrong…