Engineering:Project:Non

Project: Non-rigid EPI registration using ITK

Use case: 'I'd like to be able to register an EPI image to a conventional structural image. This will help correct for the distortions in EPI images, and help me use tools which have been developed for conventional structural images' - Saykin

Impact: Medium (Saykin)

Difficulty: Medium (Tuch)

Project summary slide

---

Vision: Software tool which can perform robust non-rigid, intrasubject registration between an EPI image and a conventional structural image.

Goal: Identify optimal ITK method and parameter settings for non-rigid intrasubject registration of T2 EPI to T1 conventional. Provide software devliverable.

Test Data: We'll look at all 6 healthy individuals from the Dartmouth data set. [https://portal.nbirn.net/BIRN/cgi-bin/DataGrid/browse.cgi?browseloc=/home/Projects/NAMIC__0003/Files/Dartmouth/HealthyControls6/DTIdata

DTI data] and [https://portal.nbirn.net/BIRN/cgi-bin/DataGrid/browse.cgi?browseloc=/home/Projects/NAMIC__0003/Files/Dartmouth/HealthyControls6/SPGRdata

conventional data].

Team members: Snyder, Jen, Tuch, Ibanez

Win conditions:

Kitware: Demonstration of advantages of ITK solution. Practical implementation towards DBP.

MGH: Guidance on optimal use of ITK non-rigid registration functions

Dartmouth: Software deliverable

Most probable failure modes:

• Loss of traction

• Insufficient accuracy

• Ambiguous accuracy results (E.g., registration performance inconsistent across subjects or regions)

---

Steps:

1. Requirements extraction (Tuch/Saykin): done
2. Create Wiki project page (Tuch): done
3. Create directory on NAMIC sandbox (Ibanez): done
4. Transfer test images (Snyder): done
5. Implement rigid registration method (Snyder/Ibanez): done
6. Implement deformable method (Snyder/Jen): done
7. Write test engine (Snyder/Jen): in progress
8. Evaluate registration performance (Snyder/Jen/Ibanez/Saykin): queued
9. Software deliverable (Snyder/Jen): queued