The Utility of Face2Gene App for Syndrome Recognition in Indian Children with Dysmorphism.

OBJECTIVES

To determine the accuracy of Face2Gene (F2G) app in the diagnosis of a genetic syndrome as a first correct response, after uploading the image of the patient in the app (top 1 accuracy), first 3 responses (top 3 accuracy), and first 10 responses (top 10 accuracy) out of 30 differential diagnoses given by the app. Also, to determine the accuracy of the app for rare and ultra-rare diagnoses given by the app.

METHODS

Frontal facial images of individuals with the diagnosis of a genetic syndrome (established clinically or molecularly) were analysed with and without additional clinical features.

RESULTS

In this study, a total of 118 children were recruited. Overall, the molecularly confirmed eventual diagnosis appeared in the "top 10" suggested syndromes by Face2Gene in 75/118 cases, providing a diagnostic yield of 63.6%. In this study, the top 1 accuracy for correct first diagnosis by the app and clinician's first diagnosis was 45.8% (n = 54). The Mcnemar test was examined for the clinician's accurate diagnosis as compared to top 1, top 3, and top 10 accuracy by the app and the p-value was statistically significant for top 10 accuracy (0.0005) and not for the top 1 and 3 diagnoses. The top 10 accuracy for the app in the rare cases was 21/30 cases (70%), and for ultra-rare cases was 28/64 (43.8%).

CONCLUSIONS

The Face2Gene app is useful as an assistant to clinicians in the diagnosis of rare and ultra-rare diseases. The top 10 accuracy is better than clinical diagnosis, and the yield is better for the ultra-rare cases and the single gene disease category, too.