Can T1-Weighted Magnetic Resonance Imaging Significantly Improve Mini-Mental State Examination-Based Distinguishing Between Mild Cognitive Impairment and Early-Stage Alzheimer’s Disease?

Abstract

Background: Detecting early-stage Alzheimer’s disease (AD) is still problematic in clinical practice. This work aimed to find T1-weighted MRI-based markers for AD and mild cognitive impairment (MCI) to improve the screening process. Objective: Our assumption was to build a screening model that would be accessible and easy to use for physicians in their daily clinical routine. Methods: The multinomial logistic regression was used to detect status: AD, MCI, and normal control (NC) combined with the Bayesian information criterion for model selection. Several T1-weighted MRI-based radiomic features were considered explanatory variables in the prediction model. Results: The best radiomic predictor was the relative brain volume. The proposed method confirmed its quality by achieving a balanced accuracy of 95.18%, AUC of 93.25%, NPV of 97.93%, and PPV of 90.48% for classifying AD versus NC for the European DTI Study on Dementia (EDSD). The comparison of the two models: with the MMSE score only as an independent variable and corrected for the relative brain value and age, shows that the addition of the T1-weighted MRI-based biomarker improves the quality of MCI detection (AUC: 67.04% versus 71.08%) while maintaining quality for AD (AUC: 93.35% versus 93.25%). Additionally, among MCI patients predicted as AD inconsistently with the original diagnosis, 60% from ADNI and 76.47% from EDSD were re-diagnosed as AD within a 48-month follow-up. It shows that our model can detect AD patients a few years earlier than a standard medical diagnosis. Conclusion: The created method is non-invasive, inexpensive, clinically accessible, and efficiently supports AD/MCI screening.

Machine Learning Approach for Improved Longitudinal Prediction of Progression from Mild Cognitive Impairment to Alzheimer’s Disease

“…This was similar to the approach of Marcisz and Polanska, who trained a linear regression model on a small set of features including the MMSE to investigate AD but did not factor in changes in disease status over time and did not include other neuropsychological tests and important comorbidities in building their model. They similarly used the MMSE on its own as a base comparison model [ 30 ]. In our MLA, MMSE input features allowed for the comparison of the predictive performance of the MMSE alone to a model built using MMSE components along with other features of interest, rather than comparing the MMSE to a model built solely using other features.…”

“…To gauge the performance of the XGBoost-based MLA versus models built with other machine learning algorithms, we trained three additional models for comparison. We selected one model which is substantially simpler compared to XGBoost, k -nearest neighbors (KNN), one which is substantially more complex, a multi-layer perceptron (MLP) neural network, and one based on a logistic regression approach previously published in the context of the ADNI dataset [ 30 ].…”

“…The demographics for each of the training sets were substantially consistent between different timepoints, and representative training set demographic information is shown in Table 1 for the 12-, 24-, and 48-month timepoints. The demographic information for the other timepoints (18,30,36, and 42 months) is given in Supplementary Table S2. The age in the demographics tables is given for the baseline of the study, i.e., 12 months prior to the first time point (the 12-month timepoint or prediction window).…”

“…Data were derived from imaging, demographics, neuropsychological assessments, and genetic data and the model achieved an AUROC of 0.925 for predicting MCI conversion to AD within 3 years and AUROC of 1.00 for differentiation between participants with AD and those with typical cognition. Marcisz and Polanska used a logistic regression model to classify individuals as having MCI, AD, or normal cognition using magnetic resonance imaging radiomic data to evaluate brain volume and yielded the highest AUROC (0.997) for classification of AD and the lowest AUROC (0.793) for classification of MCI vs. normal cognition [ 30 ]. Peng et al predicted the progression of MCI to AD using feature inputs to analyze white matter from the brain measured by positron-emission computed tomography (PET) scans [ 31 ].…”

Artificial Intelligence in Neuroradiology: A Review of Current Topics and Competition Challenges

“…Serial neuroimaging is not routinely performed in patients with dementia unless there is a new rapid loss of cognition, focal neurological signs, or seizure. There exist ongoing datasets of longitudinal biomarkers of dementia, such as the ADNI, although many of these biomarkers are obtained primarily in research settings and not part of standard clinical practice [ 132 , 133 ].…”