Abstract
A timely diagnosis of the prodromal stages of dementia remains a big challenge for the healthcare system. Of late, many tools for the automatic assessment of dementia have been proposed, but they are unreliable for detecting subtle changes in cognition. The scientific literature contains a number of reports about language disturbances at the earliest stages of dementia (the clinical syndrome known as Mild Cognitive Impairment, MCI), but the clinically-controlled collection of language data at scale remains a significant bottleneck for progress in this area. Building upon previous findings, ReMind aims to develop an innovative prototype platform that uses Natural Language Processing (NLP) technology for non-intrusive, cost-effective screening of dementia, based on a large array of Digital Linguistic Biomarkers (DLBs). Using a common tablet as a non-invasive, portable front-end, the platform will automatically collect, transcribe, time-align, analyse, and classify the digitised speech productions of both adult and elderly people while they are being tested on a few language tasks. A group of 90 people aged 65-85 years old, including subjects diagnosed with MCI, subjects diagnosed with Dementia of Alzheimer’s type, and healthy controls, will be enrolled in a one-year screening trial, whereby they will be engaged in reading aloud short texts and undertaking other language elicitation tasks with the tablet, while concurrently taking other cognitive, biological and neurological tests. Collected data will then be used to build automatic classifiers and quantitative models of types of dementia, with the aim to provide a meticulous assessment of the selective impact of (types of) dementia on the subjects’ linguistic skills. In particular, we will ascertain the validity of combining traditional tests with the evidence provided by DLBs, and analyse the independent and interactive contribution of the latter as diagnostic tests of the gravity of the conditions affecting the brain, and as prognostic indicators of the ways conditions are likely to evolve with time. ReMind will build on past and on-going work of the two project partners and their long-lasting experience in data collection and modelling of both pathological and non-pathological language data. It will push progress in the understanding and treatment of MCI through a three-pronged approach, based on data modelling, computational validation and clinical testing. The Project will release a fully-operational screening platform for cost-effective clinical screening of types of dementia. The quality of the project results will be Ministero dell'Università e della Ricerca MUR - BANDO 2022 measured in terms of i) accuracy of the implemented classifiers and data models (WP2); ii) cost-effectiveness and usability of the screening platform (WP3); iii) quantity and quality of the open-access repository of post-processed behavioural data collected by the Project (WP4), iv) level and quantity of the scientific literature produced (WP5).
Project details
Unibo Team Leader: Gloria Gagliardi
Unibo involved Department/s:
Dipartimento di Filologia Classica e Italianistica
Coordinator:
CNR - Consiglio Nazionale delle Ricerche(Italy)
Total Unibo Contribution: Euro (EUR) 98.214,00
Project Duration in months: 24
Start Date:
05/10/2023
End Date:
28/02/2026
