Machine learning-based prediction of mild cognitive impairment among individuals with normal cognitive function

Xia Wei Zhu; Si Bo Liu; Chen Hua Ji; Jin Jie Liu; Chao Huang

doi:10.3389/fneur.2024.1352423

Machine learning-based prediction of mild cognitive impairment among individuals with normal cognitive function

Front Neurol. 2024 Feb 2:15:1352423. doi: 10.3389/fneur.2024.1352423. eCollection 2024.

Authors

Xia Wei Zhu¹, Si Bo Liu², Chen Hua Ji³, Jin Jie Liu³, Chao Huang¹

Affiliations

¹ School of Computer and Communication Engineering, University of Science and Technology Beijing, Beijing, China.
² Intensive Care Unit, Dalian Municipal Central Hospital Affiliated Dalian University of Technology, Dalian, China.
³ Department of General Medicine, Dalian Municipal Central Hospital Affiliated to Dalian University of Technology, Dalian, China.

Abstract

Background: Previous studies mainly focused on risk factors in patients with mild cognitive impairment (MCI) or dementia. The aim of the study was to provide basis for preventing MCI in cognitive normal populations.

Methods: The data came from a longitudinal retrospective study involving individuals with brain magnetic resonance imaging scans, clinical visits, and cognitive assessment with interval of more than 3 years. Multiple machine-learning technologies, including random forest, support vector machine, logistic regression, eXtreme Gradient Boosting, and naïve Bayes, were used to establish a prediction model of a future risk of MCI through a combination of clinical and image variables.

Results: Among these machine learning models; eXtreme Gradient Boosting (XGB) was the best classification model. The classification accuracy of clinical variables was 65.90%, of image variables was 79.54%, of a combination of clinical and image variables was 94.32%. The best result of the combination was an accuracy of 94.32%, a precision of 96.21%, and a recall of 93.08%. XGB with a combination of clinical and image variables had a potential prospect for the risk prediction of MCI. From clinical perspective, the degree of white matter hyperintensity (WMH), especially in the frontal lobe, and the control of systolic blood pressure (SBP) were the most important risk factor for the development of MCI.

Conclusion: The best MCI classification results came from the XGB model with a combination of both clinical and imaging variables. The degree of WMH in the frontal lobe and SBP control were the most important variables in predicting MCI.

Keywords: dementia; eXtreme Gradient Boosting; machine learning; mild cognitive impairment; random forest.

Grants and funding

The author(s) declare financial support was received for the research, authorship, and/or publication of this article. This work was supported by the Dalian High-Level and Talent Innovation Support Plan (2021RQ029, 2023RY019).