Assessing Cognitive Features of Dementia Progression for Different Dementia Levels using Feature Selection

Thabtah, Fadi; Peebles, David · 2023 · Crossref

DOI: 10.21203/rs.3.rs-2976507/v1

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the need for cost-effective, data-driven methods to identify specific cognitive features that signal the progression of Alzheimer’s disease (AD). Current neuropsychological assessments are often time-consuming and resource-intensive. The research aims to determine whether specific neuropsychological items change as patients move between dementia stages, specifically from Cognitively Normal (CN) to Mild Cognitive Impairment (MCI) and from MCI to AD. By identifying influential cognitive subsets, the study seeks to assist clinicians in early screening and intervention. The researchers utilized data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), focusing on the Alzheimer’s Disease Assessment Scale-Cognitive 13 (ADAS-Cog) assessments. They analyzed two specific subgroups over a 36-month timeline from baseline diagnosis: participants who progressed from CN to MCI ('Cog-CN-MCI') and those who progressed from MCI to AD ('Cog-MCI-AD'). To handle class imbalance in the datasets, Synthetic Minority Over-sampling Technique (SMOTE) was applied. The primary method involved feature-feature assessment using Pearson Correlation coefficients to evaluate linear correlations among cognitive items, excluding the target class attribute. This approach aimed to identify redundant features and pinpoint influential cognitive activities associated with disease progression. The results revealed distinct patterns in feature correlation between the two progression groups. For the 'Cog-CN-MCI' group, three non-overlapping features—'command,' 'naming of objects,' and 'ideational praxis'—were identified as uniquely influential. These items showed negative correlations with other cognitive features and covered five DSM-5 cognitive domains, making them strong candidates for assessing early MCI traits. In contrast, the 'Cog-MCI-AD' group exhibited significant overlapping among cognitive features, with all correlations above 14%. This overlap suggests that distinguishing between patients who remain in MCI and those who progress to mild dementia is difficult using these features alone. However, 'spoken language' and 'word recognition' showed the least correlation in this group, indicating they may warrant further investigation for MCI assessment. Additionally, 'word recall' and 'delayed word recall' maintained high correlations across both groups. The study concludes that influential cognitive subsets are uniquely associated with specific dementia stages. The identification of non-overlapping features for CN-to-MCI progression provides a potential basis for more targeted neuropsychological assessments. Conversely, the high feature overlap in MCI-to-AD progression highlights the complexity of detecting early AD onset. The authors note limitations regarding the low frequency of positive progression cases and suggest future work integrating classification and feature assessment using deep neural networks. These findings offer a pathway for developing quicker, more affordable screening tools by focusing on specific, high-impact cognitive indicators.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• mci dementia driving