Artificial Intelligence Software Changes Rare Disease Testing Strategy in Real Time: An International Case Series...

Genetic disorders commonly share features such as developmental delays, cognitive impairment, and behavioral challenges, yet many conditions also present unique dysmorphic features that distinguish them. Performing a thorough medical and family history and a detailed physical exam with attention to dysmorphic features is often the first step toward arriving at a genetic diagnosis. Synthesizing a differential diagnosis from the information gathered helps to guide the genetic testing strategy. Challenges to recognizing a disorder include the breadth of the clinician's prior experience, the lack of distinctive dysmorphology or overlapping dysmorphology in some conditions, atypical presentations, and difficulties identifying phenotypes across different ancestries. In cases where such challenges exist, advanced facial recognition technology can help the consulting expert by directing a more efficient test strategy. We present 17 cases involving 19 patients, including one pair of affected siblings and one case involving a child and her affected mother, for which DeepGestalt, the technology powering Face2Gene, changed medical geneticists' testing decisions. These cases illustrate how this form of artificial intelligence can provide clinical utility through influencing providers' genetic testing recommendations in real time.