Identification of Amyotrophic Lateral Sclerosis Based on Diffusion Tensor Imaging and Support Vector Machine

Area of Science:

• Neuroimaging
• Neurology
• Machine Learning

Background:

• White matter (WM) impairments are recognized in amyotrophic lateral sclerosis (ALS), affecting both motor and non-motor functions.
• Diffusion tensor imaging (DTI) measures microstructural changes in WM.
• Support vector machine (SVM) is a machine learning algorithm capable of classification.

Purpose of the Study:

• To evaluate the potential of DTI-derived WM measurements for identifying ALS using SVM.
• To correlate DTI findings with ALS severity as assessed by the ALS Functional Rating Scale-Revised (ALSFRS-R).

Main Methods:

• Voxel-wise fractional anisotropy (FA) values from DTI were extracted from 22 ALS patients and 26 healthy controls.
• Feature selection was performed using Fisher scores, and a linear kernel SVM was trained.
• Leave-one-out cross-validation (LOOCV) was employed for model evaluation.

Main Results:

• An optimal feature set of 2,400-3,400 ranked features yielded a classification accuracy of 83.33% (sensitivity 77.27%, specificity 88.46%).
• The area under the ROC curve was 0.862, indicating good diagnostic performance.
• Predicted function values positively correlated with ALSFRS-R scores (r=0.493, P=0.020).
• Key contributing regions included the corticospinal tract, postcentral gyrus, and frontal/parietal areas.

Conclusions:

• SVM analysis of DTI-derived WM measurements is a feasible approach for ALS diagnosis.
• The findings suggest DTI-based SVM can reflect disease severity.
• Further validation with larger cohorts is recommended.