Interpretable surface-based detection of focal cortical dysplasias: a MELD study

Abstract Introduction One outstanding challenge for machine learning in diagnostic biomedical imaging is algorithm interpretability. A key application is the identification of subtle epileptogenic focal cortical dysplasias (FCDs) from structural MRI. FCDs are difficult to visualise on structural MRI but are often amenable to surgical resection. We aimed to develop an open-source, interpretable, surface-based machine-learning algorithm to automatically identify FCDs on heterogeneous structural MRI data from epilepsy surgery centres worldwide. Methods The Multi-centre Epilepsy Lesion Detection (MELD) Project collated and harmonised a retrospective MRI cohort of 1015 participants, 618 patients with focal FCD-related epilepsy and 397 controls, from 22 epilepsy centres worldwide. We created a neural network for FCD detection based on 33 surface-based features. The network was trained and cross-validated on 50% of the total cohort and tested on the remaining 50% as well as on 2 independent test sites. Multidimensional feature analysis and integrated gradient saliencies were used to interrogate network performance. Results Our pipeline outputs individual patient reports, which identify the location of predicted lesions, alongside their imaging features and relative saliency to the classifier. Overall, after including a border-zone around lesions, the developed MELD FCD surface-based algorithm had a sensitivity of 67% and a specificity of 54% on the withheld test cohort, and a sensitivity of 85% on a restricted subcohort of seizure free patients with FCD type IIB who had T1 and FLAIR MRI data. Conclusions This multicentre, multinational study with open access protocols and code has developed a robust and interpretable machine-learning algorithm for automated detection of focal cortical dysplasias, giving physicians greater confidence in the identification of subtle MRI lesions. Highlights This large, multi-centre, multi-scanner neuroimaging cohort captures the heterogeneity of histopathological subtypes and imaging features of patients with FCD. We developed a robust and interpretable surface-based algorithm which detects FCDs with a sensitivity of 67% and a specificity of 54%. The algorithm generates individual patient reports that “open the AI black-box” highlighting predicted lesion locations, alongside the imaging features and their relative saliency to the classifier.