Multi-Atlas Segmentation Using Partially Annotated Data: Methods and Annotation Strategies

Lisa Margret Koch,Tong Tong,Paul Aljabar,Wenjia Bai,Daniel Rueckert,Jonathan Passerat-Palmbach,Christian Frederik Baumgartner,Martin Rajchl

doi:10.1109/tpami.2017.2711020

Abstract

Multi-atlas segmentation is a widely used tool in medical image analysis, providing robust and accurate results by learning from annotated atlas datasets. However, the availability of fully annotated atlas images for training is limited due to the time required for the labelling task. Segmentation methods requiring only a proportion of each atlas image to be labelled could therefore reduce the workload on expert raters tasked with annotating atlas images. To address this issue, we first re-examine the labelling problem common in many existing approaches and formulate its solution in terms of a Markov Random Field energy minimisation problem on a graph connecting atlases and the target image. This provides a unifying framework for multi-atlas segmentation. We then show how modifications in the graph configuration of the proposed framework enable the use of partially annotated atlas images and investigate different partial annotation strategies. The proposed method was evaluated on two Magnetic Resonance Imaging (MRI) datasets for hippocampal and cardiac segmentation. Experiments were performed aimed at (1) recreating existing segmentation techniques with the proposed framework and (2) demonstrating the potential of employing sparsely annotated atlas data for multi-atlas segmentation.

Highlights

I N recent years, major efforts have been undertaken towards building large medical image databases such as ADNI [1]
We propose a unifying framework for multi-atlas segmentation using a novel graphical representation of the labelling problem
The proposed graph configurations Partial Annotation Strategy A: Slicewise (PA-SW)-configuration 1 (CONF1) and PA-SWCONF2 are equivalent to multi-atlas segmentation with regularisation refinement (MASr-LW)

Summary

Introduction

I N recent years, major efforts have been undertaken towards building large medical image databases such as ADNI [1]. Automated segmentation approaches may face challenges in large databases due to large variability in shape and appearance of the structures of interest, the presence of pathologies, or different imaging protocols used to acquire the images. Suitable atlases are not always available for large image databases, especially if the images in the database exhibit large variabilities, e.g. due to the presence of disease or aging processes. This motivates the use of training data obtained with different annotation strategies, where atlas images are only partially annotated, drastically reducing the labelling effort per image and allowing expert raters to (partially) annotate more training images in the same time. We review relevant work in the field before identifying the main contributions of this paper

Results

Discussion

Conclusion