Algorithm + results

Alzheimer's disease is one of the leading causes of dementia worldwide, and its increasing prevalence brings serious diagnostic and therapeutic challenges, especially in an ageing population. Current diagnostic methods, which include patient histories, neuropsychological tests and MRI images, often fail to achieve sufficient sensitivity and specificity. In response to these challenges, this research presents an advanced convolutional neural network (CNN) model, combining ResNet50 and Inception v3 architectures, to accurately classify the stages of Alzheimer's disease based on MRI images. The model was developed and tested on data from the Alzheimer's Disease Neuroimaging Initiative (ADNI). It classifies MRI images into four clinical categories of disease severity. The model's evaluation results, based on metrics such as precision, sensitivity and the F1 measure, confirm its effectiveness. Additional augmentation techniques and differential weights for the classes helped improve the model's accuracy. Visualization of the results using the t-SNE method and confusion matrix further illustrates the model's ability to differentiate disease categories, which can support neurological diagnosis in detecting and classifying Alzheimer's disease.