Fig. 2: Data flow diagram of the collection and curation process of the training, tuning, internal validation, and external validation datasets.
From: Emergency triage of brain computed tomography via anomaly detection with a deep generative model
The training dataset was collected and curated to include brain CT scans from healthy individuals by reviewing and applying NLP algorithms to radiological reports. In addition, consecutive brain CT scans from individuals who underwent emergency screening for suspected neurological conditions in the EDs of the internal and external institutions were independently and retrospectively collected. The internal dataset was randomly divided into two parts: a tuning dataset and an internal validation dataset. The external validation dataset included brain CT scans from 1795 consecutive individuals who had visited the ED of the external institution for five months.
