Liver lesion segmentation informed by joint liver segmentation
This work addresses medical image segmentation for liver lesions, which is incremental as it builds on existing methods but simplifies the model and post-processing.
The authors tackled liver and liver lesion segmentation in CT volumes by proposing a joint segmentation model using two fully convolutional networks trained end-to-end, achieving competitive scores on the 2017 MICCAI challenge, including second highest precision for lesion detection with high recall.
We propose a model for the joint segmentation of the liver and liver lesions in computed tomography (CT) volumes. We build the model from two fully convolutional networks, connected in tandem and trained together end-to-end. We evaluate our approach on the 2017 MICCAI Liver Tumour Segmentation Challenge, attaining competitive liver and liver lesion detection and segmentation scores across a wide range of metrics. Unlike other top performing methods, our model output post-processing is trivial, we do not use data external to the challenge, and we propose a simple single-stage model that is trained end-to-end. However, our method nearly matches the top lesion segmentation performance and achieves the second highest precision for lesion detection while maintaining high recall.