Detecting Cortical Thickness Changes in Epileptogenic Lesions Using Machine Learning.

Area of Science:

• Neurology
• Medical Imaging
• Machine Learning

Background:

• Epilepsy is a neurological disorder causing unpredictable seizures, often leading to awareness loss.
• Drug-resistant epilepsy necessitates alternative diagnostic and treatment strategies.
• Cortical thickness measurement is used to guide surgical interventions in epilepsy patients.

Purpose of the Study:

• To apply machine learning for classifying drug-resistant epilepsy using T1-weighted MRI measurements.
• To evaluate the significance of different brain measurements in epilepsy classification.
• To compare the performance of various machine learning models for epilepsy detection.

Main Methods:

• T1-weighted MRI data from epilepsy patients and healthy controls were preprocessed using BrainSuite.
• Two trials were conducted: one evaluating combinations of four measurements, and another comparing four machine learning classifiers.
• Principal component analysis and 10-fold cross-validation were applied in the second trial.

Main Results:

• The first trial showed 80.00% accuracy, with grey matter volume and white matter volume proving more significant than cortical thickness.
• The K-nearest neighbors model in the second trial achieved 97.11% accuracy, 75.00% recall, and 75.00% precision.
• Machine learning models demonstrated high potential in classifying drug-resistant epilepsy.

Conclusions:

• Machine learning, particularly K-nearest neighbors, is a promising tool for classifying drug-resistant epilepsy.
• Brain volume measurements (grey and white matter) are crucial for differentiating epilepsy patients.
• This approach offers a non-invasive method for epilepsy diagnosis and treatment planning.