A comparison of classification accuracy between MRI and PET datasets in computer aided diagnosis of Alzheimer's disease

Sammanfattning: The number of people suffering from Alzheimer’s disease (AD) is expected to increase rapidly in the coming years. Diagnosing the disease early is key to giving those affected a chance to maintain a higher quality of life. One of the most common ways to detect AD is to visually inspect scans of the patients’ brains. Computer aided diagnosis (CAD)can assist a physician’s judgement when searching for the disease in the brain scans, making the assessment more accurate. Progress has been made in this field throughout the years. This paper compares the machine learning classification accuracy of AD on images from two different brain scanning procedures - Magnetic resonance imaging (MRI) and Positron emission tomography (PET). Both the MRI and PET datasets contained 60 images. 30 of the images were AD cases and 30 were normal cases in each of the datasets. The images were processed into 1-dimensional signals using Discrete wavelet transform (DWT). The classification accuracy of Support vector machine (SVM), Random forest (RF) and Naive Bayes (NB) was obtained and then evaluated using 6-fold cross validation (CV). This study showed that PET images are more suitable than MR images for diagnosing AD using machine learning classifiers. The highest accuracy for PET and MRI from 6-fold CV was 100% and 90% respectively. The lowest accuracy was 60% for PET and 40% for MRI.