top of page

Machine learning-based Alzheimer’s disease Designation (MAD)

 

This SPM extension evaluates whether the given brain FDG PET images “look like” an Alzheimer’s disease (AD) patient’s or not (non-AD).

 

The brain FDG PET images must be already pre-processed (normalized and smoothed with [8 8 8]mm Gaussian filter) using SPM12 (with “old normalize” routine on to a PET template). Please note that the original study was conducted without structural MRI co-registration.

 

The program computes subject scores and labels (AD vs. non-AD) estimated by 5 different algorithms described in [1].

1.  General Linear Model

2.  Scaled Subprofile Modeling - single principal component

3.  Scaled Subprofile Modeling - linearly regressed principal components

4.  Support Vector Machine – Iterative Single Data Algorithm

5.  Support Vector Machine – Sequential Minimal Optimization

 

The package contains MAT file (~1GB). Please contact us for further instructions on how to download the software package.

Requirement: SPM12 must be installed running on MATLAB (with statistics toolbox).

Disclaimer:

The Software and code samples available on this website are provided "as is" without warranty of any kind, either express or implied. Use at your own risk.

 

Citation:[1] Katako A, Shelton P, Goertzen A, Levin D, Bybel B, Aljuaid M, Yoon HJ, Kang DY, Kim SM, Lee CS, Ko JH, “Machine learning identified an Alzheimer’s disease-related FDG-PET pattern which is also expressed in Lewy body dementia and Parkinson’s disease dementia,” Scientific Reports, vol. 8(1), pp.13236, September 2018.

MAchine learning based prediction of COgnitive decline in Parkinson’s Disease patients with mild cognitive impairment (MACOPD)

 

This MATLAB script uses a Support Vector Machine (SVM) classifier to evaluate whether Parkinson’s Disease patients will develop Dementia (PDD)

or remain as stable mild cognitive impairment (MCI). The algorithm was trained in a set of 43 Parkinson’s Disease patients with clinically diagnosed

MCI. Within a 5-year follow-up period, 23 progressed to PDD diagnosis whereas 20 remained stable MCI. Also included is a voxel map in MNI

space of the linear predictor coefficients used to determine the diagnosis prediction. The FDG-PET images must be spatially normalized to MNI

space and smoothed with an 8mm Gaussian kernel. 

The package contains MAT file (~1GB). Please contact us for further instructions on how to download the software package.

Requirement: SPM12 must be installed running on MATLAB (with statistics toolbox). 

Disclaimer:

The Software and code samples available on this website are provided "as is" without warranty of any kind, either express or implied. Use at your own risk. 

Citation: Booth S, Park KW, Lee CS, Ko JH. Predicting cognitive decline in Parkinson's disease using FDG-PET-based supervised learning.

J Clin Invest. 2022 Oct 17;132(20):e157074. doi: 10.1172/JCI157074. PMID: 36040832; PMCID: PMC9566889. 

 

bottom of page