Classificação automática da doença de Alzheimer por meio de processamento de imagens e algoritmos de inteligência artificial

Abstract

Alzheimer's disease, considered the most common cause of dementia, affects more than 44 million people worldwide. Diagnosis consists of a comprehensive clinical assessment with long processes and high costs. However, with advances in medical imaging technology, professionals are able to perform a non-invasive assessment of brain anatomy. In this context, this study aimed to develop an automated algorithm using digital image processing and artificial intelligence on T1-weighted magnetic resonance images to aid the early diagnosis of Alzheimer's disease. This analysis was performed by classifying patients with Alzheimer's disease (AD) and healthy controls (CN) in a binary way. In digital image processing, the FSL library was used to adjust and segment the regions of interest. Furthermore, the extracted features used the analysis strategy of volume reduction and asymmetry of the subcortical regions of the brain, which occur before the first symptoms of dementia. To perform the classification, a supervised machine learning model with Support Vector Machine (SVM) was used, which obtained 85.71% accuracy in the polynomial kernel.