eTETRAS: An Extension To The Essential Tremor Rating Assessment Scale

Abstract

Essential tremor (ET) is a widespread neurological condition characterized by tremors in an individual’s limbs, significantly impacting one’s quality of life. It stands as the most prevalent type of movement disorder, often leading to debilitating effects. Various scales, such as the Fahn-Tolosa-Marin (FTM) tremor rating scale and The Essential Tremor Rating Assessment Scale (TETRAS), have been developed and used by physicians to classify the severity of ET. While the FTM scale is highly utilized in ET severity diagnosis, it relies on subjective assessments of the tremor. TETRAS, on the other hand, provides a more quantitative analysis of ET severity by ranking the severity of the tremor based on tremor magnitude and addresses the shortcomings of FTM. Though TETRAS is seen as an improvement to FTM, both scales often lack the necessary resolution and granularity for optimal diagnosis and treatment. Moreover, ET severity rating scales like TETRAS require a trained professional (such as a neurologist) to be present, and even in such cases, physicians use TETRAS as a metric baseline to visually approximate the severity of the tremor. This reliance on manual approximations by clinical personnel may introduce inconsistencies and bias due to variability in technique and accuracy, diminishing proper assessment and subsequent treatment. To address the limitations of TETRAS, an extended version of TETRAS (eTETRAS) was developed that leverages machine learning and wearable sensors to quantify ET severity across the entire joint range of motion. Specifically, the proposed eTETRAS utilizes a deep neural network (DNN) to classify ET severity over the range of motion of the affected limb using joint angle measurements, eliminating the need for a clinician diagnosis. To enable a rapid DNN response (i.e., nearly real-time ET classification) with enhanced resolution, the DNN assesses the severity of ET every 0.5 second increment. After the second iteration of eTETRAS, the DNN-based classifier achieved a training accuracy of 99.59% and 100% for the simulated and experimentally obtained datasets, respectively. Moreover, eTETRAS attained a training accuracy of 99.65% when trained using the combined simulation and experimental dataset.

In Chapter 1, an overview of TETRAS and its key issues are presented. This discussion serves as a foundation for exploring the motivation, clinical efficacy, and intellectual merit of the proposed eTETRAS classifier, which incorporates a comprehensive literature review. Chapter 2 discusses the ET signal generation and the experimental protocol. Chapter 3 proposes and tests a preliminary version of eTETRAS. This includes a single-dimensional DNN that uses knee encoder data as inputs into the DNN. Chapter 4 advances the work in Chapter 3 by presenting and testing the latest iteration of eTETRAS, which includes joint stiffness data concurrently with knee encoder measurements as inputs into the DNN to create a two-dimensional DNN. In Chapter 5, the thesis is concluded by highlighting the clinical and intellectual contributions of eTETRAS, and its future developments.