Automatic assessment of the depth of laser-induced skin damage in mice by U-Net based on attention mechanism

Laser therapy, a prevalent treatment for skin injury, is imperative for the evaluation of the depth and extent of skin damage. In this study, a model of laser-induced skin damage in mice was constructed, and a dataset was created using Optical Coherence Tomography (OCT) to collect B-Scan image data of the damaged area. The damaged area in the OCT images of mouse skin was then segmented and compared by adding three attention mechanisms, respectively, on the basis of U-net. The depth of the damage was then computed by using an algorithm. The findings indicate that U-Net, when integrated with the Convolutional Block Attention Module (CBAM), attains the maximum segmentation evaluation index, with 88.41%, 94.44% and 93.09% for mIoU, mRecall and mPrecision, respectively. The depth error of the damaged region was found to be 2~19 μm. Accurate Assessment of Wound Depth Caused by Laser Damage to Skin Using an Attention Mechanism-Based U-Net Network Model.