DeepCT

“DeepCT (Ver. 4.1.4) is a software-only device that uses two components: (1) Image Forwarding Software and (2) Image Processing and Analysis Server. DeepCT uses an artificial intelligence algorithm to analyze images for findings suggestive of a pre-specified clinical condition”

“A deep residual convolutional neural network (aka Residual Network or ResNet for short, see Kaiming He et al. https://arxiv.org/abs/1512.03385) was adopted as the core learning model. By repeatedly applying residual connection, the ResNet model can ease the training of networks that are substantially deeper, effectively help the convergence of the model and gain the accuracy with deep neural networks. The model was trained with a categorical cross-entropy loss with Adam optimizer. Data augmentation was introduced to motivate the model to learn the rotated and translated images. Our DeepCT system was trained with PyTorch, an open source deep learning software library (https://pytorch.org). Same model with three different layer size, 18, 34, 52, respectively, were trained to cross evaluate the model performance. 34-layer Residual Network was chose as the final model since it got the best tradeoff between performance and computation complexity. We saw little performance gain by extending from 34 to 52 layers.”