Novoic Ltd, a clinical stage digital medtech company developing speech- and language-based biomarkers for neurological disease, has presented a scientific symposium together with the Alzheimer’s Drug Discovery Foundation’s Diagnostic Accelerator at the 14th Annual Clinical Trials on Alzheimer’s Disease in Boston, MA, USA.
- Validation of a novel fully automated story recall task for repeated remote high-frequency administration
- How clinically informed deep learning can make better speech biomarkers
- A harmonized speech dataset for Alzheimer's disease biomarker development: study design of the Diagnostics Accelerator Speech Consortium Study
- Amyloid Prediction in early-stage Alzheimer's disease from acoustic and linguistic patterns of speech: design of the AMYPRED studies
- Evaluation of a remote speech-based AI system for detection of amyloid-confirmed prodromal Alzheimer's disease
BOSTON, Nov. 10, 2021 /PRNewswire/ -- Novoic Ltd, a clinical stage digital medtech company developing speech- and language-based biomarkers for neurological disease, has presented a scientific symposium together with the Alzheimer's Drug Discovery Foundation's Diagnostic Accelerator (DxA) at the 14th Annual Clinical Trials on Alzheimer's Disease (CTAD) in Boston, MA, USA.
Scientists from Novoic and the DxA presented research on how speech analysis using digital devices can automate audio-verbal cognitive testing, to facilitate development of early disease sensitive measures and biomarkers. The symposium panel discussed how to accelerate the translation of these models into impactful medical devices that can be deployed a scale, and how this is being advanced by new gold standard speech datasets in early-stage amyloid-confirmed individuals, including the AMYPRED studies (NCT04928976, NCT04828122) and the DxA's Speech Consortium Study. Breaking research presented at CTAD by Novoic showed that speech biomarkers could support detection of both mild cognitive impairment and Alzheimer's disease at-risk individuals with confirmed amyloid biomarker positivity. The results show that Alzheimer's disease can be detected early by analysing brief spoken responses with novel first-in-class clinically informed deep learning models.
"Research in speech biomarkers for Alzheimer's disease originated out of the aphasia community with research usually examining patients with dementia, such as the DementiaBank dataset – but there's been anecdotal evidence that speech might change earlier in the course of the disease, reflecting not simply aphasia, but multi-domain cognitive changes associated with the disease that present in how people speak. We set up the AMYPRED studies to test if there's a speech phenotype at the prodromal and preclinical stages of Alzheimer's disease, and whether there are speech differences in amyloid biomarker positive individuals, who may be at greater risk for developing cognitive problems or Alzheimer's Dementia in the future" said Emil Fristed, CEO of Novoic, "In the context of individual variability, traditional analysis approaches using audio and text feature engineering are of limited use. But using a novel class of clinically informed deep learning models we reveal a distinct speech phenotype at both the prodromal and preclinical stage."
Data from the AMYPRED studies presented during the symposium and in associated poster presentations show that these models can be used on remotely collected data from brief self-supervised testing to identify early cognitive impairment.
In the symposium presentation "Validation of a novel fully automated story recall task for repeated remote high-frequency administration" Dr. Caroline Skirrow, Senior Scientist at Novoic, described the design and implementation of the ASRT (Automated Story Recall Task). The ASRT task can be automatically administered in-clinic, via telemedicine visits or via remote self-administration, and is automatically scored using automated transcription and text similarity metrics. Task performance metrics show sensitivity to mild cognitive impairment and with excellent good psychometric properties, including high parallel form reliability and concurrent and convergent validity with well-established tests of cognitive function.
In the presentation "How clinically informed deep learning can make better speech biomarkers" Dr. Jack Weston, CTO of Novoic, described work on building a novel clinical AI model incorporating inductive biases: "inductive biases are ways to mathematically constrain how data moves around in a neural network, to punish or reward certain behaviours. The art of translating domain expertise into inductive biases is at the core of the field of deep learning. What are the structural symmetries of a story recall task? It's effectively paraphrase data." To exploit this symmetry between story recall and paraphrase evaluation, Dr. Weston's group went on to develop the state-of-the-art model for paraphrase evaluation (ParaBLEU), and applying this model on the ASRT revealed distinct speech phenotype for both prodromal and preclinical Alzheimer’s disease. Dr. Jack Weston concluded: "it's the combination of the right dataset, the right speech task, and the right machine learning model."
In the presentation "A harmonized speech dataset for Alzheimer's disease biomarker development: study design of the Diagnostics Accelerator Speech Consortium Study" Dr. Lampros Kourtis described a global effort to collect a harmonized speech dataset for Alzheimer's disease biomarker development, where Novoic's Automated Story Recall Task is now being implemented. The DxA Speech and Language Consortium Study is a longitudinal, observational, multi-site substudy, designed by global leaders in the field intending to harmonize speech collection in multiple ongoing global cohort studies/sites. Research subjects will include cognitively unimpaired, through preclinical and prodromal Alzheimer's, to Alzheimer's dementia, as well as people with Fronto-temporal Dementia, Lewy body Dementia, vascular Dementia, and Parkinson's disease. The harmonized dataset will bea data resource facilitating the next decade of research on speech biomarkers.
About the AMYPRED studies
The AMYPRED studies (NCT04928976, NCT04828122) are the first clinical studies to investigate deep speech phenotyping in a biomarker-confirmed, early stage Alzheimer's disease population. The objectives of the studies were to test if a novel first-in-class clinical deep learning model can detect speech patterns specific to mild cognitive impairment and to amyloid biomarkers. In both the UK and US-based study, participants with mild cognitive impairment and normal cognition, and with confirmed positive and negative amyloid biomarker status were recruited into four distinct cohorts. In the studies participants did a battery of speech tasks both under supervision of a clinician and remotely with unsupervised testing.
About Novoic
Novoic is an Oxford-founded clinical stage digital medtech company developing clinical deep learning algorithms to detect early-stage neurological disease and psychiatric disorders from how people speak. The company's AMYPRED clinical studies were the first to test deep speech phenotyping in biomarker-confirmed early-stage Alzheimer's disease, and its research team has developed the state-of-the-art clinical deep learning model for text evaluation. Novoic is working with organisations such as the Alzheimer's Drug Discovery Foundation and the UK's National Health Service to accelerate clinical translation of speech biomarkers into impactful medical devices.
View original content:SOURCE Novoic Ltd
