Computer-automated focus lateralization of temporal lobe epilepsy using fMRI

Area of Science:

• Neuroimaging
• Epilepsy Research
• Machine Learning in Medicine

Background:

• Temporal lobe epilepsy (TLE) diagnosis relies on accurate lateralization of the seizure focus.
• Standard clinical practice often uses MRI, but its accuracy in lateralization can be limited.
• Interictal functional magnetic resonance imaging (fMRI) offers insights into brain network alterations.

Purpose of the Study:

• To compare computer-automated diagnosis using fMRI graph theory (CADFIG) with standard MRI analysis for lateralizing the epileptogenic hemisphere in TLE.
• To evaluate the diagnostic performance of CADFIG against expert manual analysis (MA) of MRI.
• To determine if CADFIG can improve lateralization accuracy, especially in challenging cases.

Main Methods:

• Interictal resting-state fMRI and high-resolution MRI were acquired from 24 TLE patients.
• Graph theory analysis of fMRI data was used to compute functional topology measures.
• Quadratic discriminant analysis was applied for hemisphere lateralization, with leave-one-out cross-validation.

Main Results:

• CADFIG achieved a high lateralization accuracy of 95.8% (23/24) compared to 66.7% (16/24) for expert MA of MRI.
• In cases where MRI failed to lateralize (8/8), CADFIG correctly identified the affected hemisphere.
• Combining MA with CADFIG resulted in 100% accurate lateralization for all patients.

Conclusions:

• CADFIG, utilizing fMRI and graph theory, demonstrates superior performance in lateralizing the affected hemisphere in TLE compared to expert MRI analysis.
• Functional brain patterns detected by fMRI, when analyzed with machine learning, offer a powerful tool for TLE diagnosis.
• Integrating fMRI-based assessments into presurgical protocols can enhance diagnostic accuracy and potentially improve surgical outcomes for TLE patients.