Machine learning for filtering out false positive grey matter atrophies in single subject voxel based morphometry: A...

Area of Science:

• Neuroimaging
• Machine Learning
• Medical Image Analysis

Background:

• Single subject VBM (SS-VBM) is an alternative to standard VBM for single case studies.
• SS-VBM often yields a high number of false positive detections, complicating analysis.

Purpose of the Study:

• To propose and evaluate a machine learning technique, Support Vector Machine (SVM), to refine SS-VBM findings.
• To reduce false positive detections in SS-VBM while preserving true positive results.

Main Methods:

• Utilized Support Vector Machine (SVM) for automated data classification.
• Conducted experiments using 3D T1 MRI scans from the IXI dataset.
• Artificially induced atrophy in MRI scans to simulate neurological disorders.

Main Results:

• The proposed SVM approach significantly reduced false positive clusters (p < 0.05).
• No significant statistical differences were observed in true positive findings (p > 0.05).
• Results were consistent across various atrophy locations and sizes.

Conclusions:

• SVM effectively refines SS-VBM analysis by minimizing false positives.
• This method enhances the usability of SS-VBM for clinical image reading.
• Reduces the need for extensive manual filtering by radiologists and clinicians.