When Deep Learning Fails: Limitations of Recurrent Models on Stroke-Based Handwriting for Alzheimer's Disease Detection

2025-08-05

AI Summary

Early Alzheimer’s disease (AD) screening is hindered by the high cost and invasiveness of neuroimaging and cerebrospinal fluid biomarkers. This study proposes a non-invasive AD detection paradigm based on handwriting behavior, introducing a novel dataset comprising 34 diverse writing tasks. We systematically benchmark recurrent neural networks (RNNs)—including LSTM and GRU—against traditional ensemble learning models using pre-extracted kinematic and spatial features. Crucially, we find that forcing discrete stroke sequences into continuous time-series representations induces fundamental representation mismatch in RNNs, resulting in low specificity and high performance variance. In contrast, ensemble methods leveraging handcrafted features achieve significantly superior performance (accuracy +8.2%, F1-score +11.5%). This work is the first to empirically delineate the applicability boundary of deep temporal models for discontinuous biological behavioral signals. It provides both theoretical insight and empirical evidence to guide algorithm selection in wearable health monitoring systems.

Technology Category

• Machine Learning: Ensemble Methods
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Humans and AI: Brain-Sensing and Analysis

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Alzheimer's disease detection requires expensive neuroimaging or invasive procedures, limiting accessibility. This study explores whether deep learning can enable non-invasive Alzheimer's disease detection through handwriting analysis. Using a dataset of 34 distinct handwriting tasks collected from healthy controls and Alzheimer's disease patients, we evaluate and compare three recurrent neural architectures (LSTM, GRU, RNN) against traditional machine learning models. A crucial distinction of our approach is that the recurrent models process pre-extracted features from discrete strokes, not raw temporal signals. This violates the assumption of a continuous temporal flow that recurrent networks are designed to capture. Results reveal that they exhibit poor specificity and high variance. Traditional ensemble methods significantly outperform all deep architectures, achieving higher accuracy with balanced metrics. This demonstrates that recurrent architectures, designed for continuous temporal sequences, fail when applied to feature vectors extracted from ambiguously segmented strokes. Despite their complexity, deep learning models cannot overcome the fundamental disconnect between their architectural assumptions and the discrete, feature-based nature of stroke-level handwriting data. Although performance is limited, the study highlights several critical issues in data representation and model compatibility, pointing to valuable directions for future research.

Problem

• Non-invasive Alzheimer's detection via handwriting analysis
• Recurrent models fail on discrete stroke-based features
• Traditional methods outperform deep learning in this task

Innovation

• Uses stroke-based features not raw signals
• Compares recurrent models with traditional methods
• Highlights model-data representation mismatch issues