Scribd
Upload
32 views

Machine Learning With 3D Spatio-Temporal SSM For Alzheimer's Disease Patient Classification

1) The document presents a machine learning method using 3D spatio-temporal statistical shape models (stSSM) to classify Alzheimer's disease patients and healthy subjects based on volumetric MRI scans. 2) The method constructs stSSMs from MRI scans using weighted principal component analysis with a temporal weighting function. It extracts features from the stSSM to generate feature vectors for each subject. 3) These feature vectors are then classified using support vector machines (SVM) and K-nearest neighbors (KNN) classifiers. The method achieved 78.6% accuracy using SVM and 81.0% accuracy using KNN when classifying 84 subjects from a publicly available database.

Uploaded by

saadia
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
32 views

Machine Learning With 3D Spatio-Temporal SSM For Alzheimer's Disease Patient Classification

1) The document presents a machine learning method using 3D spatio-temporal statistical shape models (stSSM) to classify Alzheimer's disease patients and healthy subjects based on volumetric MRI scans. 2) The method constructs stSSMs from MRI scans using weighted principal component analysis with a temporal weighting function. It extracts features from the stSSM to generate feature vectors for each subject. 3) These feature vectors are then classified using support vector machines (SVM) and K-nearest neighbors (KNN) classifiers. The method achieved 78.6% accuracy using SVM and 81.0% accuracy using KNN when classifying 84 subjects from a publicly available database.

Uploaded by

saadia
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Machine learning with 3D spatio-temporal SSM for Alzheimer's disease

patient classification
Saadia Binte Alam*1,Manabu Nii*1, Akinobu Shimizu*3 Syoji Kobashi*1,*2,
*1 Graduate School of Engineering, University of Hyogo
*2 WPI Immunology research center, Osaka University, Japan
*3Tokyo Agriculture University, Japan
Email : [email protected] ; [email protected] , [email protected]
[email protected]

Abstract

Purpose: Alzheimer's disease (AD) is one of the most common form of dementia affecting the
elderly person. Advanced medical imaging such as the volumetric MRI can detect changes in the
shape of the human brain due to the loss of the brain tissues. Measuring degree of change in brain
shape during the progress of Alzheimer's disease can help neurologists in detecting and stage the
disease. In this paper, our purpose is to individually classify AD patients and healthy subjects by
using a spatio-temporal statistical shape model (stSSM). We present and evaluate a new automated
method of feature extraction form statistical shape model and discriminate between patients with
Alzheimer"s disease (AD) and healthy subjects using different classifiers.

Method: The proposed model constructs 3D-stSSM from volumetric MRI scan using expectation-
maximization (EM) based weighted principal component analysis (WPCA) with a temporal weight
function, by generating a set of feature vectors that characterizes the mean shape, Eigen vectors,
estimates principal component (PC) score, and finally classifies the subjects based on the extracted
features. In brief, our method consists of three stages: Step 1. Feature extraction, Step 2. Training
the classifier, Step 3. Submit new MRI brains to the trained classifier, and output the prediction.
For feature extraction, we construct feature vector matrix form PC score estimated using spatio-
temporal Statistical Shape Model (st-SSM). The obtained eigenvectors along with mean shape
constructed for one specific time point are used to estimate PC score which represents
dimensionally reduced feature vector. Two different classification methods namely Support Vector
Machine (SVM) and K-Nearest Neighbor (KNN) have been applied followed by the performance
evaluation using 2-fold cross validation.

Results: To train and test the proposed model, we used 84 adult subjects (age: 60 to 90 years)
within which 42 are normal subjects (Mean ± SD = 74.93±9.58) and 42 are very-mild to moderate
dementia (Mean ± SD = 75.57±7.34) patients with Clinical Dementia Rating (CDR ranging from
0 to 2) from publicly available imaging database called OASIS. For brain MRI images, features
extracted from st-SSM are classified with SVM using fine Gaussian kernel function gives 78.6%
accuracy whereas KNN gives the accuracy of 81.0%. Figure 1 shows the summary of the
experiment.
Figure 1: Performance evaluation (a) using SVM (b) K-Nearest Neighbor

Conclusion: In this paper, st-SSM based feature extraction and classification technique were
presented to detect Alzheimer's disease. The proposed method can be helpful to reduce the burden
of medical practitioners when a large amount of data is available as the model is efficient for
classification of the human brain normal or AD with good sensitivity, specificity and accuracy
rates.

Keywords: spatio-temporal SSM; Brain; MRI; Alzheimer's disease (AD); Machine learning;
Prediction

You might also like