Deep neural network-based automated assessment of skin volume in laser-damaged mice

The purpose of this paper is to propose an automated deep neural network (DNN) based method for the assessment of skin volume in mice with laser injury. By constructing a laser-damaged mouse skin model, images of the damaged area were acquired using optical coherence tomography (OCT). The necessary dataset for deep learning was created by preprocessing these images. On this basis, U-Net and DeepLabV3+ models were used to segment OCT images of laser-damaged mouse skin. Algorithms were developed to quantify the volume of skin damage. The MIoU values of U-Net in the damage area (DA), (epi-)dermis layer (EDL), subcutaneous fat layer (SFL) and muscle fascia layer (FML) were 92.03%, 89.39%, 84.86% and 82.05%, respectively. The volume evaluation errors in the DA, EDL, SFL and FML were 8.926×10⁻² mm³, 4.086×10⁻² mm³, 3.947×10-2 mm³ and 5.475×10⁻²mm³, respectively.