Symmetry-Aware Machine Learning for the Diagnosis of Alzheimer’s Disease and Detection of Mild Cognitive Impairment Using Biomarkers

Abstract

Alzheimer’s disease (AD) is one of the most common causes of dementia in older adults, and there is currently no cure for this disease. Early detection can be very beneficial for patients, as it allows them to slow the progression of symptoms and improve their quality of life. This is where technology comes into play, especially artificial intelligence (AI), which can help doctors and nurses work faster and make better diagnoses. The goal is to test the performance of six machine learning (ML) algorithms—k-nearest neighbors (KNN), decision tree (DT), logistic regression (LR), random forest (RF), support vector machines (SVM), and Naive Bayes (NB)—to examine biomarkers and help diagnose AD and mild cognitive impairment (MCI). The dataset included 212 patients, 91 of whom had AD, 86 had MCI, and 35 showed no signs of the disease. The stages of the process were preprocessing, exploratory analysis, training, testing, and validation. DT and RF models achieved the best performance, with accuracy of 0.75 and 0.73, sensitivity of 0.75 and 0.72, and F1-scores of 0.75, respectively. LR obtained the highest MCC with 0.54. This demonstrates that ML models can be very useful for making better diagnoses of AD and MCI, especially when medical resources are limited. Finally, the DT and RF models demonstrate that applying symmetry in model training and performance metrics results in tools that can accelerate clinical translation.