This framework integrates MRI and PET imaging for early detection of Alzheimer’s disease. Pre-processing methods reduce variability, and a modified 3D U-Net with skip connections segments the hippocampus accurately. Feature extraction combines morphology and intensity traits, while a hybrid Skill Optimization and Deep Sleep algorithm enhances convergence. The detection network with 3D CNNs, ZFNet, RNNs, and attention mechanisms identifies spatial-temporal patterns, improving diagnostic accuracy and supporting therapeutic applications.