Cognitive feature evaluation for disease progression in dementia and its precursors using feature selection

Thabtah, Fadi; Peebles, David · 2025 · Crossref

DOI: 10.1007/s12553-025-01006-1

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 challenge of identifying specific cognitive markers that indicate disease progression in dementia, particularly during early stages such as the transition from Cognitively Normal (CN) to Mild Cognitive Impairment (MCI) and from MCI to Alzheimer’s Disease (AD). Current neuropsychological assessments are often time-consuming, resource-intensive, or invasive. The authors aim to provide a cost-effective, data-driven approach to pinpoint influential cognitive features from the Alzheimer’s Disease Assessment Scale-Cognitive 13 (ADAS-Cog) that clinicians can use for early screening and monitoring. The researchers utilized data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) repository, specifically merging ADAS-Cog-13 scores with diagnostic records. They constructed two longitudinal sub-cohorts: participants progressing from CN to MCI and those progressing from MCI to AD within a 36-month window. To address class imbalance, the Synthetic Minority Over-Sampling Technique (SMOTE) was applied. The analysis employed feature-feature correlation matrices using Pearson’s r to identify redundant items, alongside Permutation Feature Importance (PFI) and Mutual Information (MI) using Decision Tree and Logistic Regression classifiers to rank feature relevance. The results revealed distinct patterns of cognitive decline across progression stages. In the CN-to-MCI cohort, memory-related tasks—specifically Word Recall, Delayed Word Recall, and Word Recognition—showed high correlations and emerged as the most predictive features, aligning with early episodic memory deficits. Conversely, features like Naming, Command, and Ideational Praxis showed low correlations, suggesting they tap into distinct DSM-5 domains and are valuable for early screening. In the MCI-to-AD cohort, the importance of Orientation increased significantly, reflecting a shift toward executive and attentional decline. While Delayed Word Recall remained a top marker, Command and Comprehension tasks gained predictive power, indicating broader cognitive deterioration. The study concludes that data-driven machine learning approaches effectively identify evolving cognitive signatures during dementia progression. It demonstrates that memory tasks are critical for detecting early transitions (CN to MCI), while executive and attentional features become more prominent in later stages (MCI to AD). These findings support the use of targeted, item-level cognitive assessments rather than aggregate scores, offering a more efficient and precise method for clinical screening and disease management.

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