Prediksi Gangguan Kognitif Ringan Menggunakan Pencitraan Resonansi Magnetik dan Deep Learning: Sebuah Studi Meta-Analisis

Introduction: Mild Cognitive Impairment (MCI) is a condition characterized by cognitive decline that does not interfere with daily activities but increases the risk of progressing to Alzheimer's Dementia (AD). Early detection of MCI progression to AD is crucial for early intervention. Purpose: The purpose of this meta-analysis to evaluate the performance of Convolutional Neural Networks (CNNs), an artificial intelligence used to analyze complex data such as images, in predicting the conversion of MCI to AD using MRI data. Methods: A meta-analysis was conducted following PRISMA guidelines, utilizing articles from PubMed and Wiley Online Library. Inclusion criteria focused on studies that used CNN in conjunction with MRI to diagnose MCI. A total of 39 articles with 40 comparative studies were analyzed. Results: The CNN showed an average accuracy of 0.7910 (p < 0.0001), sensitivity of 0.7698, specificity of 0.8467, and an Area Under the Curve (AUC) of 0.8063. High heterogeneity (I² = 90.90%) indicated significant variation across studies. Sub meta-analysis revealed consistent performance despite the heterogeneity. Conclusion: CNN is potentially useful for predicting the progression of MCI to AD. Further research is needed to address heterogeneity, improve algorithms, expand datasets, and include more diverse populations.