Meeting Banner
Abstract #0951

Evaluation of Three Automated Methods of Identifying the Hippocampus on T1 Weighted Images

Jian Lin1, Mingyi Li1, Katherine A. Koenig1, Mark J. Lowe1, Micheal Phillips1

1Radiology, Cleveland Clinic, Cleveland, OH, United States

In the abstract, we compare the two most popular and full automated methods of identifying hippocampus from T1 weighted whole-brain images (T1W), FSL/first and FreeSurfer, to a third candidate, a template registration method developed by our group based on Advanced Normalizaiton Tools and symmetric image normalization method(ANTS/SyN). After quantative matching analysis and qualitative visual inspection, we conclude that the ANTS method produced ROIs that are closest to the hand traced hippocampal.

Keywords

accuracy adding additional adequate advanced affect anatomic anatomical another ants appearance applied applying approaches approved audience auto automated automatic become best better bilateral body brain candidate clinic closest coefficient common comparing complex control controls correlation cross described developed dice disease disorders distortions drawn effort elderly evaluating evaluation evidence expert feature forward frequently full future gold grant greater grown hand healthy human identification identified identifying implicated includes individual input inspected inspection inverse invertible involved labeling library made male manual manually mark matching matrix measure medical metric model national nature neuron neuropsychiatric normalization often option perfect performance popular popularity populations potentially produced prohibitive propagate protocol qualitative quantifying radiology rating referred refinement registered registration reproducible researchers routine scanned scanner sclerosis scope score segmentation selection shape similarity smith society software space structure structures subject subjects symmetric systematic target targets template third tools traced tracing trained transformation trio unsymmetrical utilizing variance visual visually volume volumes whole