Automated segmentation of kidneys and kidney tumors is an important step in the tumor's morphometrical details to monitor the progression of disease and accurately compare decisions regarding the kidney tumor. Manual delineation techniques are often tedious, error-prone and expert knowledge for creating unambiguous representation of kidneys and tumors segmentation. In this work, we propose an end-to-end boundary fully Convolutional Neural Networks (CNNs) for reliable kidney and kidney semantic segmentation from arterial phase abdominal 3D CT scans. We a segmentation network consisting of an encoder-decoder architecture specifically accounts for organ and tumor edge information by devising a boundary branch supervised by edge-aware loss terms. We have our model on 2019 MICCAI KiTS Kidney Tumor Segmentation Challenge and our method has achieved dice scores of 0.9742 and 0.8103 for kidney tumor repetitively and an overall composite dice score of 0.8923.
No takes yet. Share an insight, caveat, or question.
Myronenko et al. (2019) studied this question.