Gold standard validation, robust algorithm performance, 124 findings.
The Annalise CXR model was trained on over 520,000 CXR studies, comprising over 820,000 individual CXR images. Annalise CXR is trained and validated on datasets labelled by a group of 148 radiologists. Through an intensive process, the group confirmed that each label met standardised definitions set by globally recognised thoracic specialist radiologists.
The Annalise CXR artificial intelligence algorithm is trained on a broad dataset covering a spectrum of CXR machine equipment and manufacturers, departmental and portable imaging, patient demographics, and inpatient and outpatient groups.