AUTOMATED ANALYSIS OF INFANT FACIAL DYNAMICS FOR PRE-SYMPTOMATIC DIAGNOSIS OF RETT SYNDROME

Rett syndrome (RTT) is a rare neurological disorder caused by X-linked methyl-CpG-binding protein 2 (MECP2) gene mutations. It is characterized by severe cognitive, motor and social impairments with extensive clinical variability. Clinical progression involves apparently normal development followed by regression of previously acquired skills, typically around 2 years of age, precluding early diagnosis. Evidence from retrospective home videos and preclinical investigations suggests that atypical facial and body dynamics can be observed in RTT before overt symptoms, which could be exploited to enable earlier diagnosis; yet, these alterations are often too subtle to be reliably detected. Computer vision approaches may enable automatic and fine-tuned identification of early alterations, allowing an accurate discrimination of RTT infants from typically developing (TD) peers. To explore these tools, retrospective amateur videos of newborns aged 0–6 months who were later diagnosed with RTT (n=12), alongside videos of TD infants (n=21) were analyzed with Baby FaceReader (Noldus), a commercially available software that automatically classifies discrete facial muscle movements, allowing standardized comparison between RTT and TD infants’ facial dynamics. We demonstrate that subtle alterations in durations and frequency of lower-face movements can be detected in RTT infants compared to TD controls, particularly in the mouth and jaws, indicative of early disruptions in facial motor control. These results provide preliminary evidence that early RTT identification may benefit from automated approaches. Digital tools could enable timely intervention, improving clinical outcomes.
Funded by the European Union-Next Generation EU-NRRP M6C2-Investment 2.1 Enhancement and strengthening of biomedical research in the NHS (PNRR-MR1-2022-12376808).