The diagnosis of AD poses a major challenge to modern healthcare. A diagnosis is necessary for effective management and intervention. Through the integration of ML and DL, the early detection and classification of AD may be aided. In this paper, we propose a deep learning framework leveraging DTCWT for multi-class classification of AD across six distinct stages. The DTCWT has several advantages such as multiresolution analysis and directional sensitivity. Using DTCWT for the preprocessing step, we build a CNN and ANN architecture to extract discriminative features from structural MRI data. DTCWT offers an accurate and precise early detection of Alzheimer’s disease across different stages. Our model is tested and trained on a large dataset of people at different phases of AD, which is divided into six different classes according to the severity and the illness. We utilize thorough cross-validation methods to ensure the model is reliable and applicable to a wide range of patients and situations. Results show that this method works well for obtaining extraordinary results in multi-class AD classification problems. In addition, we use SVM and RF methods for better training and testing of disease diagnosis. Hybridizing CNN with RF and SVM and hybridizing ANN with RF and SVM give better and more accurate results.