Figure 1
From: Automated deep learning model for estimating intraoperative blood loss using gauze images
Overall scheme of the gauze EBL prediction includes the gauze detection and EBL regression processes. In the gauze detection stage, the masked gauze image and gauze patches are produced based on the detection result of the input gauze image. Features extracted from these masking images and the crumpled state value (CSV), calculated using local binary pattern (LBP), are concatenated for EBL regression. Abbreviations: CNN, Convolutional neural network; MLP, Multilayer perceptron; FC, Fully connected; CS, Crumpled state; P-W CS, Patch-wise crumpled state.
