AI points to life-saving therapy for newborn babies

Raymond-Poincaré Hospital
Located in Garches (France), the Raymond-Poincaré hospital specializes in the care of adults and children suffering from severe disabilities (of neuromuscular or traumatic origin) and infectious diseases.

It also has a developed orthopedic surgery service: shoulder, hip, knee, hand, foot and ankle surgery. Equipped with a high-performance technical platform with balneotherapy, it is also a “rare diseases” reference center in several areas such as neuromuscular diseases, Fabry disease and rare hypersomnias.

The Raymond-Poincaré Hospital is one of the 38 health establishments that make up L’Assistance publique – Hôpitaux de Paris (APHP).

The client

The Global Movements project aims to use AI in hospital rooms to early identify diseases and malformations currently affecting around 1 in 2000 infants by examining their reflex movements (Global Movements, GMs) within the first 42 weeks after birth.

The GMs project aims to use artificial intelligence applied to video recordings of newborn babies’ movements to detect suspected infantile spinal muscular atrophy (SMA) as early as possible. If detected early enough, this disease can be successfully treated using various gene therapy strategies. On the other hand, if it is not detected before 6 months, the child’s prognosis is life-threatening.

Challenge

Diagnosis by pediatric neurology experts alone is challenging,  as there are few specialist services in the regions across France. This detection assistance system, if widely deployed (some initial deployments are in progress), for example, among pediatricians, will become a major factor in reducing the lethality of this syndrome by enabling families to be referred to specialists at the right time. 

This project intends to employ two main methods: a classification approach using videos recorded by hospitals and parents, analyzed by a Sogeti-trained AI to detect abnormal movements; and an analytical model approach considering both medical and non-medical criteria to help in diagnosis.

In addition to the intrinsic interest of the method and its first application to ASI (Amyotrophie Spinale Infantile), the other aspect of this research project is the explicability of the learning model used, which leads to a rationalization of the analysis result. We are in the process of deploying this approach by applying it to other pathologies.

Challenges involve developing motion measurement technologies, establishing new health monitoring parameters, and ensuring the system’s effectiveness across a broad range of patients and conditions without being invasive. Societal benefits include promoting preventive over curative care, reducing costs, and allocating neurologists more time to focus on other complex cases.

Solution

The project is being carried out in partnership with the APHP (Paris region hospital group), the Raymond Poincaré University Hospital, the University of Versailles Saint-Quentin en Yvelines, and the University of Paris Saclay. Through these partnerships Sogeti has fostered a lasting collaboration – it has been the subject of several scientific publications, a patent application and 2 PhD theses.