[論文レビュー] Attention-Based Deep Learning for Early Parkinson's Disease Detection with Tabular Biomedical Data

Olamide Samuel Oseni, Ibraheem Omotolani Obanla+1 more|arXiv (Cornell University)|Feb 8, 2026

Voice and Speech Disorders被引用数 0

ひとこと要約

tldr: SAINT, an attention-based Transformer, outperforms MLP, TabNet, and Gradient Boosting in early Parkinson's detection on UCI tabular voice data, achieving top weighted precision, recall, F1, MCC, and AUC-ROC.

ABSTRACT

Early and accurate detection of Parkinson's disease (PD) remains a critical challenge in medical diagnostics due to the subtlety of early-stage symptoms and the complex, non-linear relationships inherent in biomedical data. Traditional machine learning (ML) models, though widely applied to PD detection, often rely on extensive feature engineering and struggle to capture complex feature interactions. This study investigates the effectiveness of attention-based deep learning models for early PD detection using tabular biomedical data. We present a comparative evaluation of four classification models: Multi-Layer Perceptron (MLP), Gradient Boosting, TabNet, and SAINT, using a benchmark dataset from the UCI Machine Learning Repository consisting of biomedical voice measurements from PD patients and healthy controls. Experimental results show that SAINT consistently outperformed all baseline models across multiple evaluation metrics, achieving a weighted precision of 0.98, weighted recall of 0.97, weighted F1-score of 0.97, a Matthews Correlation Coefficient (MCC) of 0.9990, and the highest Area Under the ROC Curve (AUC-ROC). TabNet and MLP demonstrated competitive performance, while Gradient Boosting yielded the lowest overall scores. The superior performance of SAINT is attributed to its dual attention mechanism, which effectively models feature interactions within and across samples. These findings demonstrate the diagnostic potential of attention-based deep learning architectures for early Parkinson's disease detection and highlight the importance of dynamic feature representation in clinical prediction tasks.

研究の動機と目的

• Motivate and evaluate attention-based deep learning models for early Parkinson's disease detection using tabular biomedical data.
• Compare SAINT with established models (MLP, TabNet, Gradient Boosting) to assess performance gains.
• Demonstrate the potential of dynamic feature representation in clinical prediction tasks.

提案手法

• Empirically compare four classifiers on the UCI Parkinson's Telemonitoring dataset: MLP, Gradient Boosting, TabNet, SAINT.
• Implement TabNet with attention-based feature weighting and cross-entropy loss.
• Implement SAINT with intra-sample and inter-sample attention to model feature interactions.
• Evaluate using weighted precision, recall, F1-score, MCC, and AUC-ROC on an 80/20 train/test split (random state 42).
• Describe SAINT's embedding of input features and the attention-based scoring α_ij over embeddings, followed by sigmoid output for binary classification.

Figure 1 : Proposed PD early detection modelling framework

実験結果

リサーチクエスチョン

• RQ1 Can attention-based deep learning models like SAINT improve early PD detection from tabular biomedical data compared to traditional ML and other DL models?
• RQ2 Do intra-sample and inter-sample attention mechanisms enhance the modeling of feature interactions for PD detection?
• RQ3 How do SAINT, TabNet, MLP, and Gradient Boosting rank in terms of weighted precision, recall, F1-score, MCC, and AUC-ROC on the Parkinson’s Telemonitoring dataset?

主な発見

• SAINT achieves the highest weighted precision of 0.98 and weighted recall of 0.97.
• SAINT achieves the highest weighted F1-score of 0.97.
• SAINT attains a MCC of 0.9990 and the highest AUC-ROC among the models.
• TabNet follows with weighted precision 0.96, recall 0.95, F1 0.95, and MCC 0.9990.
• MLP reports weighted precision 0.95, recall 0.95, F1 0.95, and MCC 0.9995.
• Gradient Boosting performs lowest with weighted precision 0.90, recall 0.90, F1 0.90, and MCC 0.7310.

(a) Distribution of features with respect to status (PD or No PD).