An AI-Based Model for the Prediction of a Newborn’s Sickle Cell Disease Status

Abstract

Sickle cell disease remains a global public health problem. In Senegal, a neonatal screening and early follow-up program is conducted at the CERPAD. Such a program, started in April 2017, implements the strategy of systematic screening at birth and concerns children born in the maternity wards of the CHRSL as well as from the reference health center of the city of Saint-Louis. However, out of 18 257 newborns screened since the beginning of the program, only 49 (less than 0.5%) are pathological (SS, SC, etc.) which is extremely low compared to the cost in terms of human resources, working time and use of laboratory consumables. To mitigate the impacts of these limitations of the actual early detection and follow-up approach, we therefore propose in this paper a new approach to targeted screening based on artificial intelligence. We tested and compared the performances of five machine learning algorithms for the prediction of sickle cell status. The preliminary results are promising for the task of whether or not a given newborn has a potentially pathological profile, with the majority of the models showing a high prediction accuracy.