Emil Fristed - CMIC/WEISS Joint Seminar Series

Invited Speaker: Emil Fristed – CEO and CSO, Novoic

Title:  Evaluation of the ASRT system: an automated AI-enhanced speech-based screener for mild cognitive impairment and amyloid beta positivity

Abstract: 

Background: Prior research documents detectable changes in speech in Alzheimer’s Disease (AD). The current study assesses whether a fully automated speech-based Artificial Intelligence (AI) system can detect early clinical impairment and amyloid positivity, which characterise the earliest stages of AD.

Methods: 200 participants with established amyloid beta (Aβ) and clinical diagnostic status (97 Aβ+, 103 Aβ-; 94 cognitively unimpaired (CU), 106 with mild cognitive impairment (MCI) or mild AD) completed assessments for the AMYPRED-UK (NCT04828122) and AMYPRED-US (NCT04928976) studies. Participants completed an automatically administered and analysed story recall test, the Automatic Story Recall Task (ASRT). The AI text-pair evaluation model ParaBLEU produced vector-based outputs from the original story text and transcribed participant recordings. Logistic regression models were trained with tournament leave-pair-out cross-validation analysis to predict Aβ status in the full sample (primary endpoint), MCI in the full sample, and Aβ status in diagnostic subgroups (secondary endpoints). Simulation analysis examined potential benefits of the ASRT system for: (1) MCI screening in primary care compared to the Mini Mental State Exam, and (2) reducing PET scans required for research requiring a target amyloid positive sample.

Findings: The ASRT system predicted Aβ (area under the receiver operating characteristic curve, AUC=0·77) and MCI/mild AD (AUC=0·83) in the full sample, and predicted Aβ in subsamples (MCI/mild AD: AUC=0·82; CU: AUC=0·71). Simulation indicates that the ASRT system screening could modestly increase correct MCI referrals (+8·5%), whilst substantially reducing incorrect referrals (-59·1%). Screening is estimated to reduce the number of PET scans required by 35·3% and 35·5% in MCI and CU individuals, respectively.

Interpretation: The ASRT is a brief, speech-based test offering simple accessible and scalable screening for MCI and amyloid beta biomarker positivity, which could be used to identify appropriate patients for treatment, and amyloid targeting clinical trials.

Funding: Novoic

Bio: 

Emil Fristed is the CEO and CSO at Novoic, a London-based digital diagnostics company that develops medical algorithms to detect and monitor neurological disease by analysing behavioural data streams such as speech. Here he oversees clinical development and serves as the chief investigator on 5 clinical trials, and the company’s partnerships with the Alzheimer’s Drug Discovery Foundation and the NHS.

He's obtained several million pounds in funding, and international recognition in the areas of translating science and socially conscious entrepreneurship, including a Leading For Impact Fellowship, awarded by the Skoll Centre in Oxford; and have served as an entrepreneur in residence at both Entrepreneur First in London and at the ELEVATE accelerator in Oxford. He’s the inventor on multiple patents in natural language processing techniques. He obtained his MSc in computational neuroscience from Oxford University.

Chair: Cameron Shand