A camera combined with AI could help further diagnosis of movement disorders in newborns

Detecting impaired motor function in newborns using a standard camera and AI can aid in diagnosing infantile conditions like hypotonia, which can limit active movement. A research team is investigating this approach, drawing from work on babies with spinal muscular atrophy (SMA), a rare genetic neuromuscular disorder. In 2019, the expensive treatment Zolgensma became headline news, costing nearly 2 million euros per injection.

Early diagnosis is crucial for effective therapies, as delays significantly reduce their effectiveness. The team aimed to tackle the challenge of delayed diagnosis by using an ordinary camera combined with artificial intelligence. Identifying SMA as early as possible is essential, as it can prevent neurons from dying and improve patient outcomes.

SMA is a rare genetic condition that causes progressive degeneration of motor neurons controlling muscle function. Severe cases, known as type 1 SMA, lead to rapid loss of movement, sitting up, and eventually breathing within two years without treatment. France included SMA in routine neonatal screening since 2025, but complementary tools for rapid clinical assessment remain useful.

Hypotonia, reduced muscle tone, is a nonspecific symptom that can be associated with various childhood conditions but often serves as the first indicator for diagnosis. The study involved 25 infants in pediatric intensive care units, five with genetically confirmed SMA and 20 without neurological issues. Computer vision was used to analyze infants' spontaneous movements, reconstructing a digital skeleton of the infant based on anatomical joints and segments.

An artificial intelligence algorithm then analyzed the video footage, extracting 108 features to distinguish typical from altered motor skills characteristic of SMA. The algorithm correctly classified both groups with an accuracy of 97%, with depth of movements being a significant distinguishing parameter. The tool quantifies visual impressions into objective data, providing clinicians with an initial objective indicator to guide diagnosis using AI.