Alzheimer’s disease (AD) is the common type of dementia, which is a decline in cognition with significant memory loss that cannot be reversed causing the loss of independent functionality. Early detection is thus important for proper management because the current diagnostic methods, among them being cognitive testing, behavioral assessments, brain imaging, and history, are both unreliable and insufficient for the early-stage diagnosis. The paper will propose a novel approach for early-stage AD detection based on MRI capability with enhanced image processing, using Convolutional Neural Networks in combination with Wavelet Transform, Random Forest, and Support Vector Machine techniques. Our approach applies the Discrete Wavelet Transform of the MRI images to decompose them into multiple frequency frames, and further features are extracted by processing the wavelet coefficients with kurtosis-based thresholding for denoising enhanced representations. Then, the findings are used to train on a broad data set offered by Kaggle with CNN, Random Forest, and SVM models which can classify different stages of Alzheimer’s diseases. The proposed approach improves the accuracy of detection significantly, which provides a more reliable solution for early diagnosis. Future work will be based on further optimization of the model’s performance and its extension to the application of the tool for other neurodegenerative conditions.