Fig. 1: Our proposed anomaly detection framework based on a deep generative model— closest normal-style-based generative adversarial network (CN-StyleGAN)—trained on normal brain CT images.
From: Emergency triage of brain computed tomography via anomaly detection with a deep generative model
The model reconstructs a brain CT image as its closest normal-style brain CT image. Based on the density error, the anomaly score of the scan is determined and used to identify emergency cases, followed by visualization of the detected lesion upon overlay.
