Enhanced Computer Vision Technique for Differentiating Tremor Types

Tremors, involuntary rhythmic oscillations of body parts, can significantly impact individuals' quality of life and pose diagnostic challenges. This study focuses on differentiating among rest tremor, essential tremor, and cerebellar tremor, each associated with distinct neurological pathways and clinical characteristics. Clinicians face considerable challenges due to the similar symptoms exhibited by these tremor types. This paper aims to distinguish the characteristics of these tremors using an advanced algorithm developed with the CVZone library, based on the Mediapipe framework. The developed algorithm differentiates tremors by considering pose variations with 90% accuracy on PT data, 87.5% accuracy on ET data and 85.7% on CT data taken from multiple sources. The binomial test on the results demonstrated the algorithm's capability to differentiate tremors with a statistically significant p-value of 0.00039571, indicating robust performance in correctly identifying tremor types.