Facial phenotype information from ordinary photos - algorithm to aid diagnoses of rare diseases
Dr Chris Nellåker
Friday, 24 October 2014, 1pm to 1.30pm
Sherrington Large Lecture Theatre, Sherrington Building, off South Parks Road OX1 3PT
Facial gestalt is highly informative for clinical geneticists when diagnosing genetic diseases. To aid high-throughput diagnosis of ultra-rare developmental diseases we develop an automatic approach that implements recent developments in computer vision. To extract phenotypic information from ordinary photographs we apply machine learning and model human facial dysmorphisms in a multidimensional ‘Clinical Face Phenotype Space’ (CFPS). The CPFS locates patients in the context of known syndromes and thereby enables the generation of diagnostic hypotheses. The approach will aid clinicians by greatly narrowing (by 27.6-fold) the search space of potential diagnoses for patients with suspected developmental disorders. Furthermore, CFPS allows the clustering of patients by phenotype even when no known syndrome diagnosis exists, thereby aiding disease identification. We demonstrate that this approach provides a novel method for inferring causative genetic variants from clinical sequencing data through functional genetic pathway comparisons. With further development the CFPS approach can be the basis of a valuable tool for clinical investigation of rare diseases.