Automated Diagnosis of Alzheimer's Using MRI Images
CSEF · 2015 Mathematics & Software Fourth Award
Overview
Objectives/Goals Over 60% of people with dementia die without ever being diagnosed as dementia patients. The reason for this is the lack of precise tools, and the extent of time and effort it takes to diagnose dementia. The purpose of this project was to create an automated tool that could accurately diagnose Alzheimer#s. The tool would use neural networks to train itself to diagnose Alzheimer's. Neural networks model the human brain by finding patterns and making connections between the inputs and the output. This would provide a quick and precise result, and enable early detection of Alzheimer#s. My tool would be an improvement from what doctors currently have to do: conduct many series of tests and brain scans, which consumes a lot of time, and come to a conclusion based largely on opinion. Methods/Materials I used the OASIS database of MRI scans and MATLAB to do my project. First, I extracted image features from the scans, such as the fractal dimension and the area of atrophy of the brain. Then, I combined those features with the patient#s basic medical data. After that, I tested out combinations of different training algorithms, structures, numbers of hidden neurons, and amounts of data. I compared results from each classifier to find the best classifier. After creating the classifier, I built a GUI using the MATLAB toolkit where doctors would only have to upload an MRI scan and enter the patient#s basic information to diagnose the disease. Results The neural network classifier with 3 chained stages, Bayesian Regulation Back propagation as a training algorithm, 12 # 16 hidden neurons, and image features combined with the patient#s basic information, achieved the highest accuracy. This classifier reached a testing accuracy of 97%, and greater than 99% overall accuracy. When the neural network was not trained with image features, not trained with the MMSE feature, or trained with a different training algorithm, the accuracy reduced significantly. The two other neural network structures I tested, single-stage and hierarchical networks, could not obtain as high an accuracy as the 3 stage chained neural network. Conclusions/Discussion I was able to meet all my project goals. The classifier I created achieved a higher diagnostic accuracy than any previous method of Alzheimer#s diagnosis. The GUI I built was functional and easy to use. My tool could be used in medical practice to change the current method of diagnosis.
Summary statement
I created a tool that quickly and precisely diagnoses Alzheimer's using MRI Images and determines the early diagnosis of a patient more accurately than any previous methods.
Help received
My brother taught me Matlab. My science teacher, Mrs. Iyer, helped me organize my ideas and develop my project.
Awards (1)
Competition history
- CSEF 2015
Resources
Related projects
CSEF · 2018
Early Classification of Alzheimer's Disease Using Machine Learning
AJAS · 2017
Automated Diagnosis of Dementia Through Supervised Learning Classification Algorithm
ISEF · 2014
Train the Artificial Brain II: Computer-Aided Diagnosis and Treatment Plan of Alzheimer's Disease using Neural Networks
ISEF · 2018
Early Diagnosis of Alzheimer's Disease Using Machine Learning
ISEF · 2021
Improving Alzheimer's Disease Diagnosis Using AI-Based Artificial Neural Networks
CSEF · 2013
The Effect of Features on Dementia Detection Accuracy
CSEF · 2023
Using a Convolutional Neural Network to Classify the Progression of Alzheimer's Patients MRI's.
ISEF · 2022
A Novel Approach to Analyzing Alzheimer's Disease in MRI Scans Using a Convolutional Neural Network
Closest projects by meaning, across every fair and year in the corpus.
Browse more like this
Source: California Science & Engineering Fair public projects