Abstract:
Alzheimer's disease is a degenerative disease that affects the age progression and causes the brain to be unable to fulfill its expected functions. Depending on the stage, the effects of Alzheimer's disease (AD) vary from forgetting the names of the surrounding people to not being able to continue daily life without assistance. To the best of our knowledge, there are currently no generally accepted diagnostic or treatment methods. In this study, a binary version of the artificial bee colony algorithm (BABC) is proposed as a feature selector for classifying AD from volumetric and statistical data of brain magnetic resonance images (MRIs). MRIs were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI). Volumetric and statistical data from the collected MRIs were obtained from an online system called volBrain. Then, for comparison, binary particle swarm optimization (BPSO), binary grey wolf optimization (BGWO), and binary differential evolution (BDE) were employed. For a comprehensive comparison, three algorithms, K-nearest Neighborhood (KNN), Random Forest (RF), and Support Vector Machine (SVM), are used as classifiers in feature selection progress. The results of this comparison demonstrate that BGWO outperforms BABC, which is a competitive method for this purpose. The outputs of the experiments show that all methods achieve their personal best by using RF as the classifier. Additionally, traditional data mining methods such as the Info Gain (IG), Gain Ratio (GR), Chi-square (CHI), and ReliefF methods were utilized for comparison. The results also demonstrate the superiority of the BABC over traditional methods. Another research point that this study focused on was to explore which parts of the brain are more relevant for AD diagnosis. The novelty of this study lies in the output of this point. Alongside the hippocampus and amygdala, the globus pallidus can also help in AD diagnosis.