[Paper Review] AI and Non AI Assessments for Dementia

Mahboobeh, Hamed Ghodrati+5 more|arXiv (Cornell University)|Jun 30, 2023

Dementia and Cognitive Impairment ResearchMedicine3 citations

TL;DR

This paper provides a comprehensive, interdisciplinary review of AI and non-AI assessments for early dementia detection, comparing their validity, reliability, accuracy, and practicality. It bridges the gap between clinicians and AI developers by analyzing existing methodologies, datasets, and evaluation frameworks, highlighting key research directions and the current maturity of dementia assessment technologies.

ABSTRACT

Current progress in the artificial intelligence domain has led to the development of various types of AI-powered dementia assessments, which can be employed to identify patients at the early stage of dementia. It can revolutionize the dementia care settings. It is essential that the medical community be aware of various AI assessments and choose them considering their degrees of validity, efficiency, practicality, reliability, and accuracy concerning the early identification of patients with dementia (PwD). On the other hand, AI developers should be informed about various non-AI assessments as well as recently developed AI assessments. Thus, this paper, which can be readable by both clinicians and AI engineers, fills the gap in the literature in explaining the existing solutions for the recognition of dementia to clinicians, as well as the techniques used and the most widespread dementia datasets to AI engineers. It follows a review of papers on AI and non-AI assessments for dementia to provide valuable information about various dementia assessments for both the AI and medical communities. The discussion and conclusion highlight the most prominent research directions and the maturity of existing solutions.

Motivation & Objective

• To address the growing need for early dementia detection through advanced assessment tools.
• To bridge the gap between clinical practitioners and AI developers by providing a shared understanding of dementia assessment technologies.
• To evaluate the validity, reliability, accuracy, and practicality of both AI and non-AI dementia assessment methods.
• To guide clinicians in selecting evidence-based assessments and inform AI developers about clinical requirements and existing benchmarks.
• To identify the most widely used dementia datasets and standard evaluation techniques in current research.

Proposed method

• Conducted a systematic review of peer-reviewed literature on AI and non-AI dementia assessments.
• Classified and compared various assessment modalities, including cognitive testing, neuroimaging, speech analysis, and wearable sensor data.
• Evaluated the performance of AI models using standard metrics such as accuracy, sensitivity, and specificity.
• Mapped the most frequently used dementia datasets, including ADNI, AIBL, and NACC, for model training and validation.
• Analyzed the technical and clinical challenges in deploying AI tools in real-world dementia care settings.
• Provided a comparative framework for assessing the maturity and clinical readiness of different assessment tools.

Experimental results

Research questions

• RQ1 What are the most effective AI and non-AI methods for early dementia detection, and how do they compare in terms of accuracy and reliability?
• RQ2 Which datasets are most commonly used in AI-based dementia assessment research, and what are their key characteristics?
• RQ3 How do clinical validation metrics such as sensitivity and specificity vary across different assessment modalities?
• RQ4 What are the main technical and practical barriers to integrating AI tools into routine dementia screening?
• RQ5 What are the most promising research directions for improving the clinical utility of AI-assisted dementia assessments?

Key findings

• AI-powered assessments show strong potential for early dementia detection, particularly when leveraging multimodal data such as speech, imaging, and behavioral patterns.
• Non-AI cognitive screening tools like the MMSE and MoCA remain widely used due to their simplicity and clinical validation, though they may lack sensitivity in early-stage detection.
• The ADNI dataset is the most frequently used resource in AI dementia research, enabling reproducible model training and benchmarking.
• AI models trained on speech and language data achieve accuracy rates above 85% in distinguishing mild cognitive impairment from healthy controls in controlled studies.
• Despite high performance in research settings, many AI tools face challenges in real-world deployment due to issues in generalizability, interpretability, and clinical integration.
• The paper identifies a clear need for standardized evaluation protocols and longitudinal validation to advance AI tools toward clinical adoption.