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

TL;DR

This study investigates the limitations of recurrent neural networks in non-invasively detecting Alzheimer's disease through stroke-based handwriting analysis, revealing their poor performance compared to traditional methods due to architectural mismatches.

Contribution

It demonstrates that recurrent models fail on stroke-based handwriting data because of their assumptions about continuous temporal sequences, highlighting the need for alternative approaches.

Findings

Recurrent models show poor specificity and high variance.

Traditional ensemble methods outperform deep recurrent architectures.

Deep models struggle with feature-based, segmented stroke data.

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.