diff --git a/meeting-notes.md b/meeting-notes.md index f3c3e92..e112262 100644 --- a/meeting-notes.md +++ b/meeting-notes.md @@ -38,83 +38,82 @@ Additionally, the notes are [hosted publicly on the GitHub](https://github.com/J ### Short-Term Outcomes - +- Jacob intro's Phil and Jakub +- Jacob follows-up with Phil ### Long-Term Outcomes - +- Create a template repository for JuliaHealth ## Notes - New member introductions - - Phil Vernes - 3. Works at JuliaHub - 4. Developing platform for running Julia jobs - 5. Many people at JuliaHub using tools within epi - 6. Can solve many problems in DSL - - Jay Sanjay - 8. Started contributing to the JuliaHealth ecosystem - 9. Looking forward to collaborating + - Phil Vernes + - Works at JuliaHub + - Developing platform for running Julia jobs + - Many people at JuliaHub using tools within epi + - Can solve many problems in DSL + - Jay Sanjay + - Started contributing to the JuliaHealth ecosystem + - Looking forward to collaborating - Running tasks follow-ups: - * Short-term task follow-ups: - * Jacob shares info on waste water management + viral load information - * Long-term task follow-ups: - * Creating a template repository - * We need to have a data structure to hold metadata (DICOM, NIFTI, etc.) - * JuliaNeuro - * HDF5 for long-term storage - * Would be great to see everyone using this - * To work on this to bring this together - * Multiple packages could have same -5. Presentation by Jakub Mitura on sub-ecosystem he created for working with CT, PET, and other medical imaging types of data. - 6. Created three packages - 7. Mainly talking about MedEye3D - 8. Segment data and iterate to see what is going on - 9. Wanted to create tools for everything around model creation - 10. Wanted to make a viewer that is well-suited for the Julia ecosystem - 11. Most medical viewers are quite "old" - 12. Not really dynamic - 13. Hard to show changes within run-time - 14. Easy to get big increase in Julia - 15. Usually something like 10x's faster - 16. We do not yet standardize way to load data - 17. Metadata is saved to HDF5 format - 18. Can introduce dynamic annotations - 19. Can have layers and switch on and switch layers - 20. Can annotate for saying where is the problem in the viewer - 21. Viewer can dynamically update - 22. Questions - 23. Tested some semi-automatic algorithms - 24. Do evaluate repeat - 25. Makes it faster for evaluation and reviewing of medical images - 26. Depends on OpenGL and NVIDIA drivers - 27. Working on Docker container that keeps - 28. What segmentation algorithm? Approach? - 29. Based on Gaussian probability distributions - 30. Some relaxation applied - 31. Based mainly on the units and different kinds - 32. Becoming more interested in transformers - 33. Implemented in JAX but want to bring it into Julia - 34. Segmentation for bladder cancer in image analysis - 35. Restarted work recently in Julia - 36. Would be useful for others? - 37. New segmentation for other ecosystem within Julia -9. Upcoming and ongoing research opportunities - * Call for collaboration on using JuliaHealth observational health tools for multi-site study -10. Medical Imaging Extension for Real World Evidence exploration - 11. Idea was to implement package for medical imaging - 12. Pillars - 13. Computing statistics across medical imaging - 14. Complete datasets for experimetnign - 15. Feature segmentation and scanning - 16. Align probabilistic model between different scans - 17. Become easier for physicians - 18. ML model for complex models for image segmentation - 19. Thing to consider -- need more robustness for image alignment? - 20. SOme transformations are relatively easier to repair - 21. Elastic deformations - 22. If we want -12. Open discussion + - Short-term task follow-ups: + - Jacob shares info on waste water management + viral load information + - Long-term task follow-ups: + - Creating a template repository + - We need to have a data structure to hold metadata (DICOM, NIFTI, etc.) + - JuliaNeuro + - HDF5 for long-term storage + - Would be great to see everyone using this + - To work on this to bring this together + - Multiple packages could have same +- Presentation by Jakub Mitura on sub-ecosystem he created for working with CT, PET, and other medical imaging types of data. + - Created three packages + - Mainly talking about MedEye3D + - Segment data and iterate to see what is going on + - Wanted to create tools for everything around model creation + - Wanted to make a viewer that is well-suited for the Julia ecosystem + - Most medical viewers are quite "old" + - Not really dynamic + - Hard to show changes within run-time + - Easy to get big increase in Julia + - Usually something like 10x's faster + - We do not yet standardize way to load data + - Metadata is saved to HDF5 format + - Can introduce dynamic annotations + - Can have layers and switch on and switch layers + - Can annotate for saying where is the problem in the viewer + - Viewer can dynamically update + - Questions + - Tested some semi-automatic algorithms + - Do evaluate repeat + - Makes it faster for evaluation and reviewing of medical images + - Depends on OpenGL and NVIDIA drivers + - Working on Docker container that keeps + - What segmentation algorithm? Approach? + - Based on Gaussian probability distributions + - Some relaxation applied + - Based mainly on the units and different kinds + - Becoming more interested in transformers + - Implemented in JAX but want to bring it into Julia + - Segmentation for bladder cancer in image analysis + - Restarted work recently in Julia + - Would be useful for others? + - New segmentation for other ecosystem within Julia +- Upcoming and ongoing research opportunities +- Call for collaboration on using JuliaHealth observational health tools for multi-site study +- Medical Imaging Extension for Real World Evidence exploration + - Idea was to implement package for medical imaging + - Pillars + - Computing statistics across medical imaging + - Complete datasets for experimenting + - Feature segmentation and scanning + - Align probabilistic model between different scans + - Become easier for physicians + - ML model for complex models for image segmentation + - Thing to consider -- need more robustness for image alignment? + - Some transformations are relatively easier to repair + - Elastic deformations # September 29 2023