Mask Mining for Improved Liver Lesion Segmentation
This work addresses segmentation accuracy for medical imaging applications, specifically liver lesion detection, and appears incremental as it builds upon existing U-Net based models.
The paper tackles the problem of improving liver and lesion segmentation from CT scans by proposing a novel procedure that incorporates segmentation errors into the learning process to boost performance, achieving an increase in dice score of up to 2 points on the LiTS dataset.
We propose a novel procedure to improve liver and lesion segmentation from CT scans for U-Net based models. Our method extends standard segmentation pipelines to focus on higher target recall or reduction of noisy false-positive predictions, boosting overall segmentation performance. To achieve this, we include segmentation errors into a new learning process appended to the main training setup, allowing the model to find features which explain away previous errors. We evaluate this on semantically distinct architectures: cascaded two- and three-dimensional as well as combined learning setups for multitask segmentation. Liver and lesion segmentation data are provided by the Liver Tumor Segmentation challenge (LiTS), with an increase in dice score of up to 2 points.