Online Signature Analysis for Characterizing Early Stage Alzheimer’s Disease: A Feasibility Study

Abstract

We aimed to explore the online signature modality for characterizing early-stage Alzheimer’s disease (AD). A few studies have explored this modality, whereas many on online handwriting have been published. We focused on the analysis of raw temporal functions acquired by the digitizer on signatures produced during a simulated check-filling task. Sample entropy was exploited to measure the information content in raw time sequences. We show that signatures of early-stage AD patients have lower information content than those of healthy persons, especially in the time sequences of pen pressure and pen altitude angle with respect to the tablet. The combination of entropy values on two signatures for each person was classified with two linear classifiers often used in the literature: support vector machine and linear discriminant analysis. The improvements in sensitivity and specificity were significant with respect to the a priori group probabilities in our population of AD patients and healthy subjects. We show that altitude angle, when combined with pen pressure, conveys crucial information on the wrist-hand-finger system during signature production for pathology detection.

On Extracting Digitized Spiral Dynamics’ Representations: A Study on Transfer Learning for Early Alzheimer’s Detection

“…The highest average accuracies are obtained with pressure images (77% in Conv3 and 79.5% in Conv5 ) and altitude (77.5% in Conv3 ), followed by velocity images (73.5% in Conv3 ). This confirms our previous result in [ 28 ], obtained on the same population, but on signature samples: pressure and altitude convey a high discriminant power for early AD detection. More precisely, the altitude angle leads to the highest sensitivity (81%), highlighting that the way the pen is held by the writer during the spiral gesture, is of significant importance for AD detection.…”

An Automated Method for Biometric Handwritten Signature Authentication Employing Neural Networks

“…Many systems help to analyze a handwritten text [10][11][12][13][14], which utilizes such experts' features. There are also methods, based on signature analysis, that help diagnose a variety of illnesses, such as Parkinson's and Alzheimer's disease [15][16][17][18]. However, the last-mentioned topics are beyond the scope of this paper.…”