Micrographia in Parkinson's Disease: Automatic Recognition through Artificial Intelligence

Background: Parkinson's disease (PD) leads to handwriting abnormalities primarily characterized by micrographia. Whether micrographia manifests early in PD, worsens throughout the disease, and lastly responds to L-Dopa is still under scientific debate. Objectives: We investigated the onset, progression and L-Dopa responsiveness of micrographia in PD, by applying a non-invasive and cheap tool of artificial intelligence- (AI)-based pen-and-paper handwriting analysis. Methods: Fifty-seven PD undergoing chronic L-Dopa treatment were enrolled, including 30 early-stage (H&Y ≤ 2) and 27 mid-advanced stage (H&Y > 2) patients, alongside 25 age- and sex-matched controls. Participants completed two standardized pen-and-paper handwriting tasks in an ecological scenario. Handwriting samples were examined through clinically-based (ie, perceptual) and AI-based (ie, automatic) procedures. Both consistent (ie, average stroke size) and progressive (ie, sequential changes in stroke size) micrographia were evaluated. Receiver operating characteristic (ROC) curves were used to evaluate the accuracy of the convolutional neural network (CNN) in classifying handwriting in PD and controls. Results: Clinically- and AI-based analysis revealed a general reduction in stroke size in PD supporting the concept of parkinsonian micrographia. Compared with perceptual analysis, AI-based analysis clarified that micrographia manifests early during the disease, progressively worsens and poorly responds to L-Dopa. The AI models achieved high accuracy in distinguishing PD patients from controls (91%), and moderate accuracy in differentiating early from mid-advanced PD (77%). Lastly, the AI model was not able to detect patients in OFF and ON states. Conclusions: AI-based handwriting analysis is a valuable non-invasive and cheap tool for detecting and quantifying micrographia in PD, for telemedicine purposes.