GestaltMatcher

Next-Generation Phenotyping with AI and FAIR data

The term next-generation phenotyping (NGP) is often used to refer to pattern recognition with advanced methods of computer vision such as deep convolutional neural networks. GestaltMatcher is a suite of algorithms that were trained to detect characteristic features in medical imaging data that can be used to diagnose rare genetic disorders. GestaltMatcher can therefore support clinicians in assessing dysmorphism and lab personnel in classifying variants from genetic testing.

GestaltMatcher services are a collection of applications that can be used to improve the quality of healthcare for patients with rare disorders. By default, the web services are operated on secure servers located in Germany. If you'd rather be interested in deploying the services locally at your clinic or in a private cloud, feel free to contact us. A team of experienced software engineers, computer scientists, and bioinformaticians would be happy to assist you in providing NGP in accordance with your data security requirements.

GestaltMatcher database (GMDB) is an online repository of curated case reports of patients with rare disorders. Each entry consists of clinical features annotated with HPO terminology, a molecular diagnosis if available, and medical imaging data. GMDB is accessible to registered clinicians for reference, and computer scientists for training, and testing. GMDB is compliant with the FAIR principles and maintained by the Arbeitsgemeinschaft für Gen-Diagnostik e.V. (AGD). AGD is a non-profit entity that is funded by membership fees and donations from Eva-Luise und Horst Köhler Stiftung and Wirtgen Stiftung.

GestaltMatcher algorithms are open-source and published in peer-reviewed scientific journals. The source code can be found in these GitHub repositories. The work is funded by public research funding unless otherwise indicated. Details can be found in the funding section of the papers.

Scientific soundness of next-generation phenotyping and technical performance: Hsieh, 2019 (PEDIA), Hsieh, 2022, Hustinx 2023,

Clinical utility and compliance with FAIR principles: Schmidt 2024, Lesmann (2024)