3D unsupervised anomaly detection through virtual multi-view projection and reconstruction: Clinical validation on low-dose chest computed tomography

Computer-aided diagnosis for low-dose computed tomography (CT) based on deep learning has recently attracted attention as a first-line automatic testing tool because of its high accuracy and low radiation exposure. We propose a method based on a deep neural network for computer-aided diagnosis called virtual multi-view projection and reconstruction for unsupervised anomaly detection (VMPR-UAD) in low-dose chest CT. Presumably, this is the novel method that only requires data from healthy patients for training to identify three-dimensional (3D) regions containing any anomalies. The method has three key components. Unlike existing computer-aided diagnosis tools that use conventional CT slices as the network input, our method (1) improves the recognition of 3D lung structures by virtually projecting an extracted 3D lung region to obtain two-dimensional (2D) images from diverse views to serve as network inputs, (2) accommodates the input diversity gain for accurate anomaly detection, and (3) achieves 3D anomaly/disease localization through a novel 3D map restoration method using multiple 2D anomaly maps. The proposed method based on unsupervised learning showed a high performance in pneumonia, tuberculosis, and both diseases with patient-level anomaly detection performance of 0.965 area under the curve (AUC) (95% confidence interval (CI); (0.955, 0.972)), 0.948 AUC (95% CI; (0.928, 0.966)), and 0.963 AUC (95% CI; (0.955, 0.970)), respectively. Additionally, our technology visualizes anomalous regions in the 3D perspective. This achieved 93% accuracy in visualizing the location of lung cancer lesions through external validation. These results highlight the potential of a new AI methodology without utilizing disease data learning; this can secure AI model prediction stability by reducing the false negative rate that occurs in various patterns of diseases.