Multi-Atlas Image Soft Segmentation via Computation of the Expected Label Value.

Iman Aganj,Bruce Fischl

doi:10.1109/tmi.2021.3064661

Abstract

The use of multiple atlases is common in medical image segmentation. This typically requires deformable registration of the atlases (or the average atlas) to the new image, which is computationally expensive and susceptible to entrapment in local optima. We propose to instead consider the probability of all possible atlas-to-image transformations and compute the expected label value (ELV), thereby not relying merely on the transformation deemed “optimal” by the registration method. Moreover, we do so without actually performing deformable registration, thus avoiding the associated computational costs. We evaluate our ELV computation approach by applying it to brain, liver, and pancreas segmentation on datasets of magnetic resonance and computed tomography images.

Highlights

A UTOMATIC image segmentation is often a central step in medical imaging studies, enabling the analysis of specific regions of interest (ROIs)
We evaluated our expected label value (ELV) computation method on several medical image databases via leave-one-out cross validation
The algorithm chose a single connected component (CC) from the mask in all experiments, except for 3% and 0.4% of the pancreas and hippocampus segmentation cases, respectively, where two CCs were chosen

Summary

Introduction

A UTOMATIC image segmentation is often a central step in medical imaging studies, enabling the analysis of specific regions of interest (ROIs). An algorithm segments a new image using the information derived from a training dataset of images that are accompanied with gold-standard (e.g. manually delineated) ROI labels. Manuscript received January 24, 2021; accepted March 4, 2021. Date of publication March 9, 2021; date of current version June 1, 2021.

Methods

Results

Conclusion