Am Mittwoch, 16. Januar 2019, wird Prof. Dr. Peter Krawitz einen Vortrag im Rahmen der ZBI “Distinguished Speaker Series” zum Thema “Deep learning on medical imaging data” halten.


Phenotype information is crucial for the interpretation of genomic variants. So far it has only been accessible for bioinformatics workflows after encoding into clinical terms by expert dysmorphologists. Facial analysis technologies have recently measured up to the capabilities of expert clinicians in syndrome identification.

We developed a facial analysis framework, DeepGestalt,using computer vision and deep learning algorithms, which quantifies similarities to hundreds of genetic syndromes, based on unconstrained 2D images. The DeepGestalt model presented in this study was trained on a subset of over 17,000 patient images out of a rapidly growing data set of tens of thousands of validated clinical cases, curated through Face2Gene, a community-driven platform. DeepGestalt currently achieves 91% top-10-accuracy in identifying over 215 different genetic syndromes and has outperformed clinical experts in three separate experiments.

We also introduce an approach, driven by DeepGestalt that uses portrait photographs for the interpretation of clinical exome data. We measured the value added by computer-assisted image analysis to the diagnostic yield on a cohort consisting of 679 individuals with 105 different monogenic disorders. For each case in the cohort we compiled frontal photos, clinical features and the disease-causing mutations and simulated multiple exomes of different ethnic backgrounds. With the additional use of similarity scores from computer-assisted analysis of frontal photos, we were able to achieve a significant improvement for the top-10-accuracy rate to 99 % for the disease-causing gene. 

We suggest that this form of artificial intelligence is ready to support medical genetics in clinical and laboratory practices, by quantifying the phenotypic similarity and will play a key role in the future of precision medicine.