Early detection of motor alterations in mouse models of Cdkl5 deficiency disorder and Rett Syndrome through automated analysis

CDKL5 deficiency disorder (CDD) and Rett syndrome (RTT) are rare neurological disorders characterized by cognitive, social, motor, and physiological impairments that emerge in infancy and predominantly affect females. Both conditions are caused by de novo mutations in X-linked genes. Their low prevalence and sporadic nature make diagnosis particularly challenging, often requiring experienced clinicians. Consequently, the path to clinical recognition is frequently long and often occurs after the closure of critical periods of brain development, when interventions become less effective. Emerging evidence revealed mild motor alterations during early postnatal life in both patients and mouse models, supporting the potential use of computer vision tools for automated action-based identification of at-risk newborns. Exploiting the standardization provided by animal models, we aimed at providing proof that the automated analysis of motor patterns can help recognizing mouse pups carrying CDD- and RTT-causing mutations before the onset of overt symptoms. To this end, we collected videos of spontaneously moving mouse pups during the first postnatal weeks of life, and trained action recognition models to distinguish mutant from wild type (wt) mouse pups based on their estimated poses. When evaluated on independent data, the model successfully differentiated between mutant and wt pups across postnatal ages, and stratifying the training data by age significantly improved prediction accuracy. Our findings highlight the potential of automated detection of CDD and RTT in animal models by leveraging early spontaneous movement alterations that precede full symptoms development and support the future application of this methodology in clinical settings to benefit patients.