Extending Supervoxel-based Abnormal Brain Asymmetry Detection to the Native Image Space

Area of Science:

• Neuroimaging
• Medical image analysis
• Neurology

Background:

• Neurological diseases often manifest as abnormal brain asymmetries.
• Current unsupervised methods for detecting these anomalies are limited by their reliance on standard coordinate spaces.
• Existing outlier detection techniques treat abnormalities as outliers, potentially missing subtle differences.

Purpose of the Study:

• To develop and evaluate a novel unsupervised method, N-SAAD, for detecting abnormal brain asymmetries directly in the native image space of MR brain images.
• To improve the accuracy and reduce false positives in the detection of neurological disease-associated brain asymmetries.

Main Methods:

• Extension of a fully unsupervised supervoxel-based approach (SAAD).
• Application of the N-SAAD method to detect asymmetries in the native image space of MR brain images.
• Comparison with an unsupervised deep learning method for outlier detection.

Main Results:

• The N-SAAD method achieved higher accuracy in detecting abnormal brain asymmetries.
• N-SAAD demonstrated considerably fewer false positives compared to the unsupervised deep learning method.
• The approach was validated on a large dataset of MR-T1 images.

Conclusions:

• N-SAAD offers a more effective and reliable approach for detecting abnormal brain asymmetries in neurological disease research.
• The method's ability to operate in native image space enhances its clinical usability.
• This unsupervised supervoxel-based technique shows promise for advancing the diagnosis and understanding of neurological disorders.