Nailfold Microhemorrhage Segmentation with Modified U-Shape Convolutional Neural Network

Nailfold capillaroscopy is a reliable way to detect and analyze microvascular abnormalities. It safe, simple, noninvasive, inexpensive. Among all the capillaroscopic abnormalities, nailfold microhemorrhages are closely associated with early vascular damages might be present in numerous diseases such as glaucoma, diabetes mellitus, systemic sclerosis. Segmentation of provides valuable pathological information that may lead further investigations. A novel deep learning architecture named DAFM-Net proposed for accurate segmentation this study. The network mainly consists U-shape backbone, dual attention fusion module, group normalization layer. backbone generates rich hierarchical representations while module utilizes captured features fine adjustment. Group introduced an effective method effectively improve convergence ability our neural network. effectiveness model validated through ablation studies experiments; achieves competitive performance microhemorrhage IOU score 78.03% Dice 87.34% compared ground truth.