Multi-Class Classification of Alzheimer’s Disease Stages Using SqueezeNet based Approach for Automated Diagnosis

Abstract

A neurodegenerative disorder is known as Alzheimer's disease (AD). AD causes cognitive impairment and memory loss as a result of brain cell death. Although research on AD has improved dramatically over the years, however, the early detection of this disease is difficult because of the complexity of the brain structure and its functions. This research mainly focuses on multiple stages of AD classification. In this paper, proposed is a SqueezeNet with Harris Hawks Optimization technique (HHO-SqueezeNet) for classifying the stages including MCI, LMCI, EMCI, AD, and CN. Principal component analysis (PCA) is used to minimize the feature dimension after MRI preprocessing, and also the feature set is selected by introducing the CML-ELM approach for each task to consider the intrinsic relevance of several related tasks. The accuracy, sensitivity, specificity, precision, and recall of the models are used to assess their performance. We discovered that our networks were capable of accurately classifying the subjects. Using the proposed model, we enhanced accuracy for all AD stages using the proposed model, with CN, EMCI, MCI, LMCI, and AD achieving 98.5 %, 98.54 %, 98.25 %, 99.02 %, and 99.2 %, respectively. We achieved better classification results with an average accuracy of 98.82% based on overall performance. The MTL algorithm based on SqueezeNet established in this research is an efficient, improved, and practical technique for diagnosing AD.