Mask Mining for Improved Liver Lesion Segmentation

TitleMask Mining for Improved Liver Lesion Segmentation
Publication TypeConference Paper
Year of Publication2020
AuthorsRoth, K, Hesser, J, KonopczyƄski, T
Conference Name2020 IEEE 17th International Symposium on Biomedical Imaging (ISBI)
Date Publishedapr
KeywordsBiomedical imaging, combined learning setups, computerised tomography, CT scans, Data Mining, false-positive predictions, image segmentation, improved Liver lesion Segmentation, learning (artificial intelligence), learning process, lesion segmentation data, Lesions, Liver, Liver Lesion Segmentation, Liver Tumor Segmentation challenge, mask mining, medical image processing, Medical Imaging, multitask segmentation, Pipelines, segmentation errors, segmentation performance, standard segmentation pipelines, Three-dimensional displays, Training, tumours, U-Net, U-Net based models

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.

Citation Keyroth_mask_2020