Classification of Alzheimer’s Disease From MRI Images
Date
2019
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Tartu Ülikool
Abstract
In English: In this thesis work machine learning techniques are used to classify MRI brain scans of people
with Alzheimers Disease. This work deals with binary classification between Alzheimers Disease
(AD) and Cognitively Normal (CN). Supervised learning algorithms were used to train a
classifier using MATLAB Classification Learner App in which the accuracy is being compared.
The dataset used is from The Alzheimers Disease Neuroimaging Initiative (ADNI). Histogram
is used for all slices of all images. Based on the highest performance, specific slices were selected
for further examination. Majority voting and weighted voting is applied in which the
accuracy is calculated and the best result is 69.5% for majority voting.
Eesti keeles: Käesolevas töös kasutatakse masinõppe meetodeid, et klassifitseerida Alzheimeri tõvega inimeste
MRI aju skaneeringuid. Töös rakendatakse binaarset liigitust Alzheimeri tõve (AD)
ja kognitiivse normaalsuse (CD) vahel. Kasutati juhendatud masinõppealgoritme, et treenida
klassifikaatoreid MATLAB’i klassifikaatorite õpperakenduses (Classification Learner App), kus
võrreldi algoritmi täpsust. Kasutatav andmestik pärineb ADNI andmebaasist (The Alzheimer’s
Disease Neuroimaging Initiative). Kõikidest piltidest võetud osadele arvutati histogrammid.
Kõrgeima jõudluse põhjal valiti konkreetsed osad edasiseks uurimiseks. Võtteldi enamus ja
kaalutud valikute täpsust ja parimaks tulemuseks saadi enamusvalikuid kasutades 69.5%.
Description
Keywords
Computer Vision, Machine Learning, Alzheimer’s Disease, feature extraction, magnetic resonance imaging, masinõpe, Alzheimeri tõbi, magnetresonantstomograafia, arvutinägemine