Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Machine learning-enhanced non-amnestic Alzheimer's disease diagnosis from MRI and clinical features

Megan A. Witherow, Michael L. Evans, Ahmed Temtam, Hamid R. Okhravi, Khan M. Iftekharuddin

TL;DR

This study tackles the challenge of diagnosing non-amnestic Alzheimer's disease (atAD), which often eludes hippocampus-focused assessments. It introduces a machine learning pipeline based on Random Forests that combines standard clinical tests with comprehensive MRI features, evaluated across VHS, NACC, and ADNI datasets using a 5x2 nested cross-validation framework. The results show that while clinical features plus hippocampal volume provide strong baseline discrimination, adding whole-brain MRI features significantly improves atAD recall (notably to 0.69 in NACC and 0.77 in ADNI) with high precision, aided by Boruta-based identification of important brain regions. The approach supports improved, non-invasive diagnosis of atAD in memory clinics, highlighting region-level MRI signals beyond hippocampal atrophy and suggesting avenues for handling disease heterogeneity and multi-etiology dementias in future work.

Abstract

Alzheimer's disease (AD), defined as an abnormal buildup of amyloid plaques and tau tangles in the brain can be diagnosed with high accuracy based on protein biomarkers via PET or CSF analysis. However, due to the invasive nature of biomarker collection, most AD diagnoses are made in memory clinics using cognitive tests and evaluation of hippocampal atrophy based on MRI. While clinical assessment and hippocampal volume show high diagnostic accuracy for amnestic or typical AD (tAD), a substantial subgroup of AD patients with atypical presentation (atAD) are routinely misdiagnosed. To improve diagnosis of atAD patients, we propose a machine learning approach to distinguish between atAD and non-AD cognitive impairment using clinical testing battery and MRI data collected as standard-of-care. We develop and evaluate our approach using 1410 subjects across four groups (273 tAD, 184 atAD, 235 non-AD, and 685 cognitively normal) collected from one private data set and two public data sets from the National Alzheimer's Coordinating Center (NACC) and the Alzheimer's Disease Neuroimaging Initiative (ADNI). We perform multiple atAD vs. non-AD classification experiments using clinical features and hippocampal volume as well as a comprehensive set of MRI features from across the brain. The best performance is achieved by incorporating additional important MRI features, which outperforms using hippocampal volume alone. Furthermore, we use the Boruta statistical approach to identify and visualize significant brain regions distinguishing between diagnostic groups. Our ML approach improves the percentage of correctly diagnosed atAD cases (the recall) from 52% to 69% for NACC and from 34% to 77% for ADNI, while achieving high precision. The proposed approach has important implications for improving diagnostic accuracy for non-amnestic atAD in clinical settings using only clinical testing battery and MRI.

Machine learning-enhanced non-amnestic Alzheimer's disease diagnosis from MRI and clinical features

TL;DR

This study tackles the challenge of diagnosing non-amnestic Alzheimer's disease (atAD), which often eludes hippocampus-focused assessments. It introduces a machine learning pipeline based on Random Forests that combines standard clinical tests with comprehensive MRI features, evaluated across VHS, NACC, and ADNI datasets using a 5x2 nested cross-validation framework. The results show that while clinical features plus hippocampal volume provide strong baseline discrimination, adding whole-brain MRI features significantly improves atAD recall (notably to 0.69 in NACC and 0.77 in ADNI) with high precision, aided by Boruta-based identification of important brain regions. The approach supports improved, non-invasive diagnosis of atAD in memory clinics, highlighting region-level MRI signals beyond hippocampal atrophy and suggesting avenues for handling disease heterogeneity and multi-etiology dementias in future work.

Abstract

Alzheimer's disease (AD), defined as an abnormal buildup of amyloid plaques and tau tangles in the brain can be diagnosed with high accuracy based on protein biomarkers via PET or CSF analysis. However, due to the invasive nature of biomarker collection, most AD diagnoses are made in memory clinics using cognitive tests and evaluation of hippocampal atrophy based on MRI. While clinical assessment and hippocampal volume show high diagnostic accuracy for amnestic or typical AD (tAD), a substantial subgroup of AD patients with atypical presentation (atAD) are routinely misdiagnosed. To improve diagnosis of atAD patients, we propose a machine learning approach to distinguish between atAD and non-AD cognitive impairment using clinical testing battery and MRI data collected as standard-of-care. We develop and evaluate our approach using 1410 subjects across four groups (273 tAD, 184 atAD, 235 non-AD, and 685 cognitively normal) collected from one private data set and two public data sets from the National Alzheimer's Coordinating Center (NACC) and the Alzheimer's Disease Neuroimaging Initiative (ADNI). We perform multiple atAD vs. non-AD classification experiments using clinical features and hippocampal volume as well as a comprehensive set of MRI features from across the brain. The best performance is achieved by incorporating additional important MRI features, which outperforms using hippocampal volume alone. Furthermore, we use the Boruta statistical approach to identify and visualize significant brain regions distinguishing between diagnostic groups. Our ML approach improves the percentage of correctly diagnosed atAD cases (the recall) from 52% to 69% for NACC and from 34% to 77% for ADNI, while achieving high precision. The proposed approach has important implications for improving diagnostic accuracy for non-amnestic atAD in clinical settings using only clinical testing battery and MRI.
Paper Structure (28 sections, 4 equations, 4 figures, 4 tables)

This paper contains 28 sections, 4 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Materials and methods
  3. Data
  4. VHS data set
  5. NACC data set
  6. ADNI data set
  7. Data selection criteria
  8. tAD and atAD
  9. tAD
  10. atAD
  11. non-AD
  12. CN
  13. Clinical features and battery of cognitive tests
  14. Montreal Cognitive Assessment (MoCA)
  15. Mini-Mental State Examination (MMSE)
  16. ...and 13 more sections

Figures (4)

  • Figure 1: Distribution of (A) subject ages, (B) z-score of hippocampal volume, (C) MoCA total scores, and (D) MMSE total scores for tAD, atAD, non-AD, and CN groups within the VHS (first column), NACC (second column), and ADNI (third column) data sets. Note that the VHS data set does not have a CN group.
  • Figure 2: ROC curves for 5-fold cross validation atAD vs. non-AD classification: (first row) clinical scores + hippocampal volume for (A) VHS data set, (B) NACC data set, and (C) ADNI data set; (second row) clinical scores + MRI features for (D) NACC data set and (E) ADNI data set.
  • Figure 3: Visualization of significant (Type I error rate $\alpha$=0.05) brain regions for (A/D) tAD, (B/E) atAD, and (C/F) non-AD groups compared to CN subjects within the NACC (A-C) and ADNI (D-F) data sets. Blue indicates regions of decreased volume, cortical thickness, or surface area compared to CN. Red indicates increased volume, cortical thickness, or surface area compared to CN.
  • Figure 4: Visualization of significant (Type I error rate $\alpha$=0.05) brain regions distinguishing between atAD and non-AD groups in the (A) NACC and (B) ADNI data sets. Blue indicates regions of decreased volume, cortical thickness, or surface area for atAD compared to non-AD. Red indicates increased volume, cortical thickness, or surface area for atAD compared to non-AD.