Our primary objective is to formulate and test an AI algorithm that can accurately assess the severity of dystonia. Secondarily we want to collect data on the kinematic properties of patients with dystonia, classifying patient movements into characteristics such as joint angles and velocities.

Precise assessment of dystonia severity is essential for effective treatment and drug development, but current evaluation methods, such as BFM, TWSTRS and CDIP-58, rely heavily on subjective scales. These tools highlight the need for more objective and efficient evaluation methods.

Various technologies have been investigated to achieve objective assessments, although these methods offer accuracy, they are time-intensive and may influence symptoms, as physical contact can alter dystonia manifestations.

X-linked dystonia-parkinsonism (XDP) presents with a mix of dystonic and parkinsonian symptoms that progress and regress over time, making it vital to monitor the disease’s course accurately.

Our model was able to ingest the patient's movements into 3 characteristics for every joint - angle, velocity and frequency. Torticollis was the most severe feature with mean joint angle 31.228 +- 15.25 degrees, followed by Trunk and arm.  When our calculated angles were used to grade scales, our metrics showed the highest correlation when compared with the TWSTRS-2 Rating Scale (rho = 0.85-0.91, p <0.001).