AttentionAnatomy: A unified framework for whole-body organs at risk segmentation using multiple partially annotated datasets

TL;DR

This paper introduces AttentionAnatomy, a novel end-to-end deep learning framework that segments whole-body organs-at-risk in CT scans by effectively utilizing partially annotated datasets through attention modules, re-calibration, and hybrid loss functions.

Contribution

The paper proposes a unified model with attention guidance, re-calibration, and a hybrid loss to enable end-to-end whole-body organ segmentation from partially annotated datasets.

Findings

Significant improvements in Dice coefficient and Hausdorff distance.

Effective handling of partial annotations in whole-body segmentation.

Enhanced segmentation accuracy over baseline models.

Abstract

Organs-at-risk (OAR) delineation in computed tomography (CT) is an important step in Radiation Therapy (RT) planning. Recently, deep learning based methods for OAR delineation have been proposed and applied in clinical practice for separate regions of the human body (head and neck, thorax, and abdomen). However, there are few researches regarding the end-to-end whole-body OARs delineation because the existing datasets are mostly partially or incompletely annotated for such task. In this paper, our proposed end-to-end convolutional neural network model, called \textbf{AttentionAnatomy}, can be jointly trained with three partially annotated datasets, segmenting OARs from whole body. Our main contributions are: 1) an attention module implicitly guided by body region label to modulate the segmentation branch output; 2) a prediction re-calibration operation, exploiting prior information of the input images, to handle partial-annotation(HPA) problem; 3) a new hybrid loss function combining batch Dice loss and spatially balanced focal loss to alleviate the organ size imbalance problem. Experimental results of our proposed framework presented significant improvements in both S{\o}rensen-Dice coefficient (DSC) and 95\% Hausdorff distance compared to the baseline model.