Statistical Analysis by Semiparametric Additive Regression and LSTM-FCN Based Hierarchical Classification for Computer Vision Quantification of Parkinsonian Bradykinesia

TL;DR

The paper addresses the need for objective quantification of Parkinson's disease bradykinesia, particularly capturing occasional arrest and decrement in amplitude. It proposes a hierarchical CV-based approach that combines MediaPipe hand pose features, a local-regression fatigue metric, and an LSTM-FCN for arrest classification (scores in $\{0,1,2,3\}$) with final bradykinesia scoring produced by XGBoost ($g: \mathcal{F} \to \{0,1,2,3\}$). Statistical significance of the proposed features is assessed via a semiparametric additive model, reinforcing their relevance to bradykinesia severity. On a dataset of 1396 motion videos from 310 PD patients, the method achieves an overall accuracy of $80.3\%$, with high AUC values across actions, and ablation studies confirm the contribution of arrest and fatigue features to performance, highlighting practical potential for clinical deployment.

Abstract

Bradykinesia, characterized by involuntary slowing or decrement of movement, is a fundamental symptom of Parkinson's Disease (PD) and is vital for its clinical diagnosis. Despite various methodologies explored to quantify bradykinesia, computer vision-based approaches have shown promising results. However, these methods often fall short in adequately addressing key bradykinesia characteristics in repetitive limb movements: "occasional arrest" and "decrement in amplitude." This research advances vision-based quantification of bradykinesia by introducing nuanced numerical analysis to capture decrement in amplitudes and employing a simple deep learning technique, LSTM-FCN, for precise classification of occasional arrests. Our approach structures the classification process hierarchically, tailoring it to the unique dynamics of bradykinesia in PD. Statistical analysis of the extracted features, including those representing arrest and fatigue, has demonstrated their statistical significance in most cases. This finding underscores the importance of considering "occasional arrest" and "decrement in amplitude" in bradykinesia quantification of limb movement. Our enhanced diagnostic tool has been rigorously tested on an extensive dataset comprising 1396 motion videos from 310 PD patients, achieving an accuracy of 80.3%. The results confirm the robustness and reliability of our method.

Figures (1)

• Figure 1: This figure illustrates the comprehensive pipeline of our approach. Utilizing Google MediaPipe[18], a state-of-the-art technology for human pose estimation, we extract a graph from the motion video. The video is then analyzed to create numerical features. Concurrently, the LSTM-FCN model is trained to quantify arrest. Based on the outputs of this stage, we perform ordinal classification to construct a score sheet.