Speech-Based Artificial Intelligence for Detection of Dementia in Danish Patients (DetectAI)

Официальное название

Development of Deep Learning Models for Detection of Neurodegenerative Diseases Using Speech - a Danish Language-based Artificial Intelligence Study (DetectAI)

Обзор

The goal of this observational study is to learn if an artificial intelligence (AI)-based speech analysis tool can identify which patients with memory problems need specialist evaluation at a memory clinic. The main questions it aims to answer are: Can the AI model accurately distinguish between patients who need referral to a memory clinic (those with dementia or Mild Cognitive Impairment) and patients who don't (those with normal cognition or memory problems from other causes like depression)? Which speech patterns and cognitive test features are most useful for making this distinction? Researchers will compare speech recordings and cognitive test results from patients diagnosed with dementia or MCI to those from patients with normal cognition or non-neurodegenerative cognitive impairment to see if the AI model can reliably predict who needs specialist dementia care. Participants will: Complete standard cognitive tests at the memory clinic Perform structured speech tasks while being audio-recorded Receive their usual clinical evaluation and diagnosis from memory clinic specialists The results of this study will help develop a tool that can assist doctors in making faster, more accurate decisions about which patients need specialist dementia evaluation, potentially leading to earlier diagnosis and better patient outcomes.

Подробное описание

Background Dementia is a growing public health challenge, and early and accurate diagnosis is essential for effective care and potential future disease-modifying treatments. Current diagnostic pathways are resource-intensive and associated with long waiting times. Speech reflects cognitive functioning, and recent international studies have shown that machine learning models can detect dementia-related patterns in speech recordings with promising accuracy. This study aims to develop a speech-based deep learning model in a Danish setting, providing a non-invasive and scalable screening tool for use in primary care.

Study Design and Sampling Methods

This is an observational, cross-sectional study. Participants are recruited using two different sampling strategies corresponding to two artificial intelligence (AI) model development tracks:

Track A (Model A) - Retrospective case-control sampling:

This track addresses a focused diagnostic task: identification of Mild Cognitive Impairment (MCI). Participants are patients with a recent diagnosis from the memory clinic at Region Zealand University Hospital (ZUH). Sampling uses convenience sampling prioritizing patients who live close to the hospital, as data collection occurs during home visits. Patients with more recent diagnoses are prioritized to minimize the risk that participants have progressed to a new disease stage since diagnosis (e.g., from MCI to dementia).

Track B (Model B) - Prospective consecutive sampling:

This track uses prospective inclusion of newly referred patients to the memory clinic without pre-selection by diagnosis, reflecting a real-world clinical screening population. All eligible, consenting patients are included consecutively at their first clinic visit, before final diagnosis is established.

Model Development Following Best Practice Guidelines The study follows TRIPOD-AI (Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis - Artificial Intelligence) and PROBAST-AI (Prediction model Risk Of Bias ASsessment Tool - Artificial Intelligence) guidelines for developing and validating clinical prediction models.

Key methodological features include:

Transparent model development: All preprocessing steps, feature extraction methods, model architectures, and hyperparameters will be documented Robust validation strategy: Data will be split into training, validation, and hold-out test sets for in-depth internal validation.

Minimizing bias: Participant selection, predictor measurement, outcome determination, and statistical analysis are designed to minimize bias according to PROBAST-AI domains Clinically relevant performance metrics: Sensitivity, specificity, area under the receiver operating characteristic curve (AUC-ROC), positive and negative predictive values, and calibration Interpretability: Feature importance analysis to understand which speech characteristics contribute to predictions

Data Collection Speech data is collected through structured tasks including picture description, verbal fluency tests, story recall, and spontaneous speech. Audio is recorded using standardized equipment with quality control checks. Clinical diagnoses are established by experienced clinicians at the memory clinic following international diagnostic criteria.

Вмешательства

• Диагностический тест Mini-mental State Examination
  Participants will be recorded during the test in order til allow the AI to learn and analyze speech patterns.
• Диагностический тест Addenbrooke's Cognitive Examination
  Participants will be recorded during the test in order til allow the AI to learn and analyze speech patterns.
• Другое Speech Task - Picture Description
  Participants will be asked to describe the Cookie Theft Picture from the Boston Diagnostic Aphasia Examination. The task will take 2 minutes. Participants will be recorded during the speech task in order to allow the AI to learn and analyze the speech patterns.
• Другое Speech Task - Picture Recall
  Participants will be asked to recall the picture shown in the previous speech task "Picture Narrative". This task will take 2 minutes. Participants will be recorded during the test in order til allow the AI to learn and analyze speech patterns.
• Диагностический тест MRI
  For healthy controls an MRI will be conducted to provide comparable imaging and as part of screening to ensure they do not meet exclusion criteria (neuroradiological findings that could affect cognitive functions). For patient participants, imaging will be performed as part of the standard diagnostic battery and results will be obtained from the electronic journal.
• Диагностический тест blood sampling
  Healthy control participants will undergo a standard blood test panel commonly used in dementia diagnostics. The panel includes complete blood counts, inflammatory markers, kidney- and liver function markers, thyroid-stimulating hormone (TSH), vitamine B12 and folate. These tests are performed to exclude underlying medical conditions that could mimic cognitive impairment. For patient participants, blood sampling will be performed as part of the standard diagnostic battery and results will be ob
• Диагностический тест Depression screening
  Performed on healthy controls to rule out depression using either the geriatric depression scale (GDS) for patients \> 65 year of age or the Major Depression Index (MDI) for patiens \<65 year of age. For patient participants, depression screening will be performed as part of the standard diagnostic battery and results will be obtained from the electronic journal.
• Другое Somatic- and neurological examination
  Healthy controls will undergo a standard somatic and neurological examination to exclude conditions that may affect cognition. This includes basic neurological assessment and clinical evaluation of general health status. For patient participants, a somatic and neurological examination will be performed as part of the standard diagnostic battery and results will be obtained from the electronic journal
• Другое Speech Task - Picture Narrative
  The participant is asked to tell a brief story based on a culturally neutral picture. This task will take approximately 2 minutes. Participants will be recorded during the speech task in order to allow the AI to learn and analyze the speech patterns

Первичные конечные точки

• Model A: Primary measure is the AUC-ROC of the model in distinguishing between MCI and AD as well as between MCI and cognitively healthy control participants. [Срок оценки: At baseline (speech recording)]

Вторичные конечные точки (5)

• Accuracy for dementia vs. depression [Срок оценки: At baseline (speech recording)]
• Sub-classification of Mild Cognitive Impairment (MCI) into progressive vs. non-progressive [Срок оценки: At baseline (speech recording) and up to 12 months after enrollment (to determine progression)]
• Classification of dementia subtypes (AD, VaD, LBD, FTD) [Срок оценки: At baseline (speech recording)]
• Comparison with established biomarkers [Срок оценки: At baseline, or at time of biomarker testing if performed after baseline]
• Feature importance analysis [Срок оценки: At baseline (speech recording)]

Критерии участия

Критерии включения

Model A (patient participants)

• Age > 50 years
• Fluent in Danish
• Minimum of 7 years of schooling
• A diagnosis of either MCI or AD, given at the SUH memory clinic within 6 months before enrollment

Model A (cognitively healthy controls)

• Age > 50 years
• Fluent in Danish
• Minimum of 7 years of schooling

Model B:

• Age > 50 years
• Fluent in Danish
• Minimum of 7 years of schooling

Критерии исключения

Model A:

Patients:

• Significantly impaired vision or hearing (to the extent that the patient cannot participate in the AI analysis)
• MMSE score < 16
• Concomitant diagnoses which are expected to influence cognitive impairment (eg. depression)
• Patients unable to give consent
• Patients with alcohol consumption >21 standard alcohol units per week
• Any history of speech or language impairment predating the current condition

Cognitively healthy controls:

• Significantly impaired vision or hearing (to the extent that the patient cannot participate in the AI analysis)
• MMSE < 26 and ACE < 90
• Clinical, laboratory, or neuroradiological findings that could affect cognitive functions
• Known diseases which are expected to impair cognitive functions
• Any history of speech or language impairment predating the current condition
• Patients with alcohol consumption >21 standard alcohol units per week.

Model B:

• Significantly impaired vision or hearing (to the extent that the patient cannot participate in the AI analysis)
• MMSE score < 16
• Patients unable to give consent
• Patients with concomitant psychosis or severe psychiatric comorbidities other than depression
• Any history of speech or language impairment predating the current condition

Критерии приведены из реестра в оригинале (на английском). Окончательную оценку соответствия проводит исследовательский центр.

Здоровые добровольцы: Да

Дизайн исследования

Модель наблюдения

Другое

Центры проведения

Дания · 1 центр

• Zealand University Hospital — Roskilde

Публикации

• Dargaud L, Partal A, Birn A, & Detlefsen S. N. (2023). Developing a Spontaneous Speech-based Artificial Intelligence for Alzheimer's Disease Detection. Transatlantic Telehealth Research Network (TTRN) International Scientific Conference 2023, Journal of the International Society for Telemedicine and eHealth.
• Lanzi AM, Saylor AK, Fromm D, Liu H, MacWhinney B, Cohen ML. DementiaBank: Theoretical Rationale, Protocol, and Illustrative Analyses. Am J Speech Lang Pathol. 2023 Mar 9;32(2):426-438. doi: 10.1044/2022_AJSLP-22-00281. Epub 2023 Feb 15. PMID 36791255
• Li J, Song K, Zheng B, Li D, Wu X, Meng H. Leveraging Pretrained Representations with Task-related Keywords for Alzheimer's Disease Detection. arXiv preprint. 2023.
• Luz S, Haider F, de la Fuente Garcia S, Fromm D, MacWhinney B. Detecting cognitive decline using speech only: The ADReSSo challenge. arXiv preprint 2021.
• Luz S, Haider F, Fromm D, Lazarou I, Kompatsiaris I, Macwhinney B. An Overview of the ADReSS-M Signal Processing Grand Challenge on Multilingual Alzheimer's Dementia Recognition Through Spontaneous Speech. IEEE Open J Signal Process. 2024;5:738-749. doi: 10.1109/ojsp.2024.3378595. Epub 2024 Mar 18. PMID 38957540
• Bex T. Comprehensive Guide to Multiclass Classification With Sklearn. Towards Data Science. 2021.
• Nicholas LE, Brookshire RH. A system for quantifying the informativeness and efficiency of the connected speech of adults with aphasia. J Speech Hear Res. 1993 Apr;36(2):338-50. doi: 10.1044/jshr.3602.338. PMID 8487525
• Buderer NM. Statistical methodology: I. Incorporating the prevalence of disease into the sample size calculation for sensitivity and specificity. Acad Emerg Med. 1996 Sep;3(9):895-900. doi: 10.1111/j.1553-2712.1996.tb03538.x. PMID 8870764

Идентификаторы

NCT: NCT07200739 · SJ-1107

Первоисточники (государственные реестры)

• ClinicalTrials.gov (NIH, США)

Открыть это исследование на ClinicalTrials.gov ↗