Examining ‘neural soma’ compartments in paediatric epilepsy to support focal cortical dysplasia lesion detection

Motivation: Focal cortical dysplasia (FCD) lesions are small in size and subtle in feature, making radiological detection a challenge. Moreover, while automated machine learning tools are promising, they often predict false positives. Goal(s): Here, we examine whether 'neural soma' parameter estimates can differentiate between true lesions and false positives in paediatric FCD cases, thus improving specificity. Approach: We fitted a constrained three-compartment 'neural soma' model to paediatric patient data acquired using tensor-valued diffusion encoding on a clinical scanner. Results: 'Neural soma' parameter maps are sensitive to signal changes due to FCD lesions and distributions in parameter estimates can be used to differentiate between regions-of-interest. Impact: The 'neural soma' model could support FCD lesion detection in a clinical setting by improving specificity between true lesions and false positives. This can inform selection of surgical resection targets and ultimately improve post-surgical outcomes and rates of seizure freedom.