TapTalkTest Project:Development of a Non-invasive Screening Test to Detect Risk of Alzheimer's Disease Pathology

Alzheimer Disease, Dementia, Movement Abnormalities, Speech Disorders

Alzheimer's disease, Dementia, preclinical, Artificial Intelligence, Machine learning

This project aims to produce a solution for the rising incidence of dementia. This is particularly pertinent in Tasmania, Australia, with a rapidly ageing population and the oldest demographics of all Australian states. The team will develop TapTalk, a new screening test that detects risk of Alzheimer's disease (AD) pathology. TapTalk, will record a person's hand movements and speech patterns with a smartphone. Computer algorithms will learn which patterns of data are associated with AD pathology. This innovative test is based on: (i) emerging research that fine motor control required for hand and speech movements is sensitive to early AD pathology and (ii) the investigators' new machine learning methods.

This project aims to produce a solution for the rising incidence of dementia. This is particularly pertinent in Tasmania, with a rapidly ageing population. The investigators' will develop TapTalk, a new screening test that detects risk of Alzheimer's disease (AD) pathology. Accounting for 70% of all dementias, the brain pathology of AD progresses silently for more than 10 years before cognitive symptoms emerge (preclinical AD). It is possible to prevent 40% of dementia by modifying risk factors such as physical inactivity and smoking. However, the lack of a cost-effective screening tool means researchers and clinicians cannot target interventions, or recruit to drug trials, in early AD. Currently, cognitive tests lack sensitivity in preclinical AD, and specialist AD biomarker tests are invasive or costly. The investigators will address the hypothesis: Hand-speech movement patterns will detect the risk of Alzheimer's disease pathology in research and clinical cohorts through three aims: 1. Develop and validate analytic algorithms for TapTalk by determining which combinations of hand-speech movement data most accurately detect preclinical AD 2. Develop smartphone capability for TapTalk and determine usability and validity 3. Prospectively validate TapTalk in people who have cognitive symptoms against gold-standard clinical diagnosis of Mild Cognitive Impairment (MCI) and AD dementia AIM 1 Problem: Identify which combination of hand-speech tests will be most discriminatory. Method: The investigators will develop software to video-record a 2-minute oral DDK (diadochokinesis) test, where participants make speech-like sounds repetitively e.g. pa-ta-ka. We already have software to collect hand movements (see TAS Test project). The research team will invite 500 ISLAND Project participants (\>50 years old) with normal cognition to compete the hand-speech tests. All participants have provided blood samples for p-tau181 levels. This new assay quantifies AD pathology (using our ultrasensitive Simoa analyser) but the practicalities and cost of accessing the highly-specialist analytic equipment limit wide accessibility. We use ptau-181 as this is a highly predictive blood biomarkers of AD risk. Analysis: The investigators will use deep neural networks to automatically track video key points (e.g. finger/thumb tips) and audio features (e.g. pa-ta-ka). A sliding window approach extracts measures (e.g. speed/rhythm) as input data for developing an algorithm that that classifies p-tau181 levels. Outcome: TapTalk protocol and algorithm. AIM 2 Problem: Develop smartphone capabilities and age/cognitive status cut-offs Method We will develop a smartphone app. ISLAND Project participants (CANTAB cognitive tests every 24 months in-kind) will be invited to complete TapTalk online every 12 months. Analysis: Multi-level regression models will measure within-subject variability, and group differences on TapTalk and CANTAB at baseline, 12 and 24 months. Outcome: An externally validated TapTalk algorithm that produces AD risk scores across age and cognitive ranges. AIM 3 Problem: Validate TapTalk in people with cognitive symptoms. Method: The clinician researchers working at the Royal Hobart Hospital (RHH) will recruit 100 patients with cognitive symptoms (\>3 months) from RHH acute medical/subacute units. The research assistant (RHHF funding requested) will complete a standard cognitive screening tool (MoCA) and smartphone TapTalk then invite patients to attend the new ISLAND cognitive clinic after discharge. This 'one-stop' interdisciplinary clinic provides bulk-billed neuropsychological and geriatrician, neurologist and physiotherapist assessments. The team will also recruit 100 consecutive patients referred to the clinic by their GPs and all patients will complete TapTalk. Analysis: The accuracy of TapTalk and MoCA will be compared to diagnosis using ROC analysis. Outcome: TapTalk prospectively clinically validated.

Australia