In Bundesgesundheitsblatt, Gesundheitsforschung, Gesundheitsschutz
Rare diseases can often be diagnosed by carefully assessing the phenotype of the patient, as characteristic deviations (dysmorphisms) occur in many genetic diseases. These affect, for example, the features of the face - the "facial gestalt."This paper highlights an area of artificial intelligence (AI) in which there has been great progress in recent years: the recognition of characteristic patterns in medical image data using deep, convolutional neural networks (next-generation phenotyping - NGP). The technical basis of the method is briefly described and the high relevance of FAIR data for the scientific community to develop AI is discussed. Furthermore, it is explained why decisions made by AI should always remain comprehensible and how it can overcome the challenges with regard to data protection and transparency.In the future, software applications with AI will support medical professionals in the diagnosis of rare diseases. AI will be trustworthy if patients retain their data sovereignty and can understand how the diagnosis was made.
Krawitz Peter M
Artificial intelligence, Dysmorphology, Facial gestalt, Machine learning, Rare disorders