Nuclei Grading of Clear Cell Renal Cell Carcinoma in Histopathological Image by Composite High-Resolution Network

The grade of clear cell renal carcinoma (ccRCC) is a critical prognostic factor, making ccRCC nuclei grading crucial task in RCC pathology analysis. Computer-aided aims to improve pathologists’ work efficiency while reducing their misdiagnosis rate by automatically identifying the grades tumor within histopathological images. Such requires precisely segment and accurately classify nuclei. However, most existing segmentation classification methods can not handle inter-class similarity property grading, thus be directly applied task. In this paper, we propose Composite High-Resolution Network for grading. Specifically, network called W-Net that separate clustered Then, recast fine-grained into two cross-category tasks are leaned newly designed high-resolution feature extractors (HRFEs). HRFEs share same backbone encoder with composite connection so meaningful features inherited Last, head-fusion block generate predicted label each nucleus. Furthermore, introduce dataset containing 1000 image patches 70945 annotated We demonstrate our proposed method achieves state-of-the-art performance compared on large dataset.