Algorithm:MGH:Development

• Optimal path calculator
• Use case: 'Specify 2 points in a diffusion image and tell how connected they are.'
• Difficulty: High
• Impact: High

See AHM 2006:ProjectsDTIPathOfInterest

• Statistical power benefit of ITK nonlinear registration
• Use case: 'Evaluate benefit of using ITK nonlinear registration for group FA comparisons'
• Difficulty: Low-Medium
• Impact: Medium

See: Engineering:Project:Non-rigid_EPI_registration

• Automatic segmentation of DTI using Spectral Clustering
• Use case: 'Automatically segment DTI image.'
• Difficulty: Medium-High
• Impact: Medium-High

1. Implement algorithm in Matlab (Ziyan) : done
2. Test/optimize on group data (Ziyan): in progress
3. Port to diffusion toolkit (Tuch) : queued

• Replacing Numerical Recipes in FreeSurfer (for open sourcing)
• Use case: 'Unit tests pass with all replacements.'
• Difficulty: Medium-High
• Impact: High

1. Write test cases for each algorithm (Snyder, Jen): done
2. Identify replacements (Snyder, Jen): done
3. Integrate required libraries into FreeSurfer build process (Snyder, Jen): done
4. Iteratively replace recipes with substitutes and run tests (Snyder, Jen): in progress

• Adding NRRD I/O to Freesurfer
• Use case: 'Open a NRRD volume in FreeSurfer.'
• Use case: 'Convert an MGH volume to a NRRD volume with Freesurfer.'
• Difficulty: Low
• Impact: Medium

1. Write unit tests for new IO functions (Snyder): in progress
2. Add NrrdIO libraries from Teem to FS source tree, build with autoconf (Snyder): done
3. Write and test FS NRRD IO functions (Snyder, Kindlmann): in progress
4. Develop approriate headers for MGH DWI data (Snyder): queued
5. Automate header generation when possible (Snyder): queued

• Replacing FreeDiffusion components with NA-MIC equivalents
• Difficulty: Medium-High
• Impact: Low

1. Write regression tests: done
2. Identify componenets to replace: queued
3. Replace and test: queued

• queued - step identified/specd
• in progress - step in progress
• done - step complete

---

DONE

• QBALL visualization
• Use case: 'Visualize q-ball data in Slicer.'
• Difficulty: Low
• Impact: Medium

1. Implement ODF polygon decimation algorithm (Tuch) : done
2. Port decimation fileformat into FreeDiffusion Visualizer (Snyder) : done
3. Port QBALL/ODF visualization into Slicer (Estepar/Snyder/Kindlmann/Tuch/Westin): done
   1. Implement (Estepar): done
   2. Test on mock data set (Estepar): done
   3. Demo for real data set (Estepar/Snyder/Kindlmann): done

DONE

• Tensor-based group comparison (Cramer test)
• Use case: 'Compare DTI images between groups using the full tensor information.'
• Difficulty: Medium
• Impact: Medium-High

1. Implement in R (Whitcher/Tuch) : done
2. Power analysis (Whitcher) : done
3. Port to Matlab (Whitcher) : done
4. Validate Matlab version against R (Whitcher) : done
5. Test on group data : done
6. Release bootstrap-only version to test group: done
7. Port FFT method from R to matlab (Whitcher): done
8. Implement FFT method in diffusion development environment (Tuch): done

See Algorithm:MGH:Development:GroupComp