Alzheimer's Disease Classification Using StyleGAN3-Based MRI Data Synthesis and Machine Learning Techniques

Abstract

Abstract:Alzheimer’s Disease (AD) is a progressive neurodegenerative disorder responsible for nearly 60% of all dementiacases. The prevalence of AD is rapidly increasing, with approximately 46.8 million people affected worldwide.Currently, there is no effective treatment to reverse or halt the progression of AD, emphasizing the critical needfor disease prevention and early diagnosis. The framework uses Medical Resonance Images (MRI) datapreprocessing, Style-based Generative Adversarial Network 3 (styleGAN3) -based synthetic image generation todevelop an Alzheimer's disease classifier. High-quality synthetic MRI images are combined with real images toincrease dataset diversity and improve the robustness and classification performance of the model. Initializesa model with Random Forest, Logistic Regression, Support vector machine, Adaptive Boosting and GaussianNaïve Bayes classifiers to find which model is best among these classifiers. The Support Vector Machine (SVM)produces the average accuracy is 0.8890 (88.90%), indicating that the SVM correctly classifies approximately89% of the samples. The average precision is 0.7903 (79.03%), indicating the proportion of predicted positivecases that are correctly classified. The average recall is 0.8890 (88.90%), showing that the model identifies mostof the actual positive cases. The average F1-score is 0.8368 (83.68%), demonstrating a good balance betweenprecision and recall.