Telemedicine: an algorithm analyses patients' handwriting and provides valuable information on their health status
Handwriting is an acquired cognitive and motor task of particular complexity, which offers an interesting observation window into the brain's functions. For this reason, handwriting monitoring provides useful biological information, especially in neurological patients: handwriting disorders are indeed frequently observed in patients suffering from neurodegenerative diseases, including Parkinson's disease (micrographia) and Alzheimer's disease (agraphia).
An interdisciplinary research team, coordinated by Antonio Suppa of the Department of Human Neuroscience at Sapienza University of Rome, has proposed handwriting analysis through artificial intelligence as an innovative system for the remote monitoring, in telemedicine, of neurological patients. The system, based on the accuracy of machine learning algorithms in detecting certain handwriting 'patterns' attributable to physiological ageing in healthy subjects, is an alternative to the usual outpatient clinical assessment.
The study, carried out in collaboration with the departments of Information Engineering, Electronics and Telecommunications of Sapienza University of Rome, IRCCS Neuromed and the Department of Neurology at the University of Cincinnati in Ohio, was published in the journal Frontiers in Aging Neuroscience.
The researchers recruited 156 healthy, right-handed subjects. They divided them into three age groups: 51 young people between the ages of 18 and 32, 40 adults between the ages of 37 and 57, and, finally, 63 subjects in advanced adulthood, i.e. between the ages of 62 and 90. Each of them was asked to write their name and surname 10 times on a sheet of white paper with a black biro and then take a photograph of their writing sample with a smartphone and send it to the researchers.
"The main scientific achievement of our study," says Antonio Suppa, "consists in the accuracy of the automatic analysis of handwriting with artificial intelligence algorithms, capable of objectifying the progressive reduction in character width due to physiological ageing and, therefore, of attributing each writing sample to a specific age group of the author".
"Although previous research had already demonstrated age-related changes in handwriting dexterity, approaches based on more complex analysis techniques such as machine learning were required to analyse large amounts of data in telemedicine".
'The analysis of handwriting with artificial intelligence algorithms,' adds Simone Scardapane, co-author of the study, 'was carried out using a convolutional neural network - i.e. an artificial network specialised in processing images and digital signals - capable of automatically converting characters into parameters of interest.
It is a simple, ecological, low-cost method that is easy to use in various fields. Indeed, in addition to its considerable implications in the neurological field, it can contribute, for instance, to the historical dating of a given document by automatically assessing the age of the person who wrote it. In particular, in the medico-legal field, it could facilitate the dating of a will when it was written or signed.
'Our hope,' concludes Francesco Asci, co-author of the study, 'is that the analysis of handwriting remotely and through artificial intelligence algorithms may constitute an innovative biomarker of ageing in the future, with a relevant impact in the field of diagnosing neurodegenerative diseases and following telemedicine methods.
Handwriting Declines With Human Aging: A Machine Learning Study – Francesco Asci, Simone Scardapane, Alessandro Zampogna, Valentina D’Onofrio, Lucia Testa, Martina Patera, Marco Falletti, Luca Marsili, Antonio Suppa - Frontiers in Aging Neuroscience (2022) https://doi.org/10.3389/fnagi.2022.889930
Department of Human Neuroscience