Skip to main content

Medical-image Analysis and Statistical Interpretation Lab

‘STAPLER’

Empirical experiment using axial cross-section of cerebellar data to assess the performance of STAPLE (on a slice-by-slice basis) and STAPLER (volumetric fusion). The representative slices shown in (A) – (C) present an example truth model, and observations by minimally trained undergraduate students, respectively. The slices seen in (D) and (E) are the estimated labels by STAPLE and STAPLER, respectively. The plot on the top of (F) shows the accuracy on a per slice basis of the observations (box plots) STAPLER (green), STAPLE (blue), and majority vote (red). The histogram on the bottom of (F) shows the number of observations per slice. Lastly, the plot seen in (G) shows the difference in DSC between STAPLER, STAPLE, and majority vote on a per label basis. The legend for these label numbers can be seen at the bottom of (G).

Robust Statistical Fusion of Image Labels

Feb. 1, 2012—Bennett A. Landman, Andrew J. Asman, Drew Scoggins, John A. Bogovic, Fangxu Xing, and Jerry L. Prince. “Robust Statistical Fusion of Image Labels”, IEEE Transactions on Medical Imaging. 2012 Feb;31(2):512-22. PMC3262958 Full text:  https://www.ncbi.nlm.nih.gov/pubmed/22010145 Abstract Image labeling and parcellation (i.e., assigning structure to a collection of voxels) are critical tasks for the assessment of volumetric...

Read more