Automatic Segmentation of Kidney Volume Using Multi-Module Hybrid Based U-Shape in Polycystic Kidney Disease

Polycystic kidney disease (Autosomal Dominant Kidney Disease, ADPKD) is the most common genetic of kidney, and measurement Total Volume (TKV) in clinical research this essential to study progression ADPKD. At present, volume segmentation polycystic kidneys mainly relies on doctors manually outline boundary radiological image. This process time-consuming, labor-intensive, inefficient, subjective, difficult guarantee consistency. In paper, A multi-module hybrid U-shape method proposed (HUNet), which introduces wavelet pooling, cascade residual, efficient multi-head self-attention into structure. We use pooling instead traditional down-sampling reduce loss detailed features, cascaded residual modules can improve ability model feature reuse, effectively capture global multi-scale information. decoding U-shape, corresponding value each decoder will be calculated, finally, total obtained by weighted average. The was trained tested dataset provided Shanghai Changzheng Hospital. automatically segmented ADPKD MRI images using with a remarkably high Dice similarity coefficient relative manual (mean=0.915). percentage difference between values HUNet methods only 0.4%. approach enables fast accurate TKV measurement.