BRAINSia / SINAPSE

SINAPSE is an interdisciplinary team of computer scientists, software engineers, and medical investigators.

0 stars 1 forks source link

Mining Whole Brain Segmentations #24

Open hjmjohnson opened 5 years ago

hjmjohnson commented 5 years ago

User Story

As DataCurator, I want to identify whole brain size for a large number of data set from well defined atlases so that further analysis can occur

Detailed Description

[ ] Use atlases from /Shared/johnsonhj/2017Projects/20150416_MiniPigData/2017_ManualApprovedSegmentations directory as reference atlas
[ ] Use module load BRAINSTools/20190415 and then BRAINSLandmarkInitializer -h to create initial transforms from each atlas brain to the new subject brain.
[ ] Run AntsJointFusion to transfer labels from the 5 atlases to the other data sets.

Acceptance Criteria

[x] A single new brain is segmented
[x] A procedure to repeat for many datasets is identified

KKnoernschild commented 5 years ago

I am slightly confused on what I should do for the process here.

Do I need to create the landmark fcsv files for the 5 atlas files, or do those already exist?

hjmjohnson commented 5 years ago

You need to make the fcsv files for the 5 atlases as well. You are the only person ever to define or identify landmarks in these datasets.

Hans

From: "notifications@github.com" notifications@github.com Reply-To: BRAINSia/SINAPSE reply@reply.github.com Date: Thursday, May 9, 2019 at 9:11 AM To: BRAINSia/SINAPSE SINAPSE@noreply.github.com Cc: Hans Johnson hans.j.johnson@gmail.com, Author author@noreply.github.com Subject: Re: [BRAINSia/SINAPSE] Mining Whole Brain Segmentations (#24)

I am slightly confused on what I should do for the process here.

Do I need to create the landmark fcsv files for the 5 atlas files, or do those already exist?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.

KKnoernschild commented 5 years ago

Shell script created that outputs resulting files to a subject-session file, that contains sub folders antsJointFusionResults (final result segmentationlabelmap), ImageFilesLinear , ImageFilesNearestNeighbor , and TransformFiles

Currently find all files and iterate through 2 of the ~120 scans, but this loop could be removed and then replaced with a file location, and make it transferrable for batch processing

The search radius is what dramatically effects the run time of each loop during the antsJointFusion .

a search radius of 3 results in a run time average of ~220 (3-4 minutes) seconds per subject, and a search radius of 5 had an average run time of ~850 seconds (14 minutes)

Search Radius of 3 took 520 seconds to run for 2 subjects.

KKnoernschild commented 5 years ago

[ ] create a document with label/folder key to make files and results easily recognizable