Automated Segmentation of Post-Surgical Resection Cavities on MRI in Focal Epilepsy: a MELD Study
Authors
Affiliations (1)
Affiliations (1)
- UCL Great Ormond Street Institute of Child Health, London, UK
Abstract
ObjectiveQuantitative assessment of extent of tissue resection following epilepsy surgery requires accurate delineation of the resection cavity on postoperative MRI. Current methods for resection cavity masking are time-consuming and labour-intensive, while existing automated approaches exhibit variable segmentation accuracy, particularly on extra-temporal resections. We developed MELD-PostOp, a deep learning tool trained and evaluated on a large, international, heterogeneous cohort to automatically segment resection cavities. MethodsThe study included 1.5 and 3T postoperative 3D T1-weighted MRI images from the Multicentre Epilepsy Lesion Detection (MELD) project (nsubjects=969, 27 centres) and from the EPISURG dataset (n=133). The cohort included both children and adults, alongside a range of resection locations, pathologies, and MRI characteristics. Resection cavities were individually segmented in 285 subjects and used to train an nnU-Net prototype model. The prototype model was used to generate an additional 680 resection masks, which were subsequently quality-controlled, edited and then combined with the original 285 to train the final MELD-PostOp model (n=965). A Stratified (STC, n=50) and Independent Test Cohort (ITC, n=87) were masked and withheld for model evaluation. Performance was evaluated using Dice Similarity Coefficient (DSC), 95th percentile Hausdorff distance (HD95), number of predicted clusters and inference runtime, and compared against established tools (Epic-CHOP and ResectVol). ResultsMELD-PostOp achieved a median DSC of 0.85 and HD95 of 3.61 on the combined test cohort, outperforming Epic-CHOP (DSC 0.68, HD95 9.54) and ResectVol (DSC 0.66, HD95 12.07), with significant improvements seen in both temporal and especially extra-temporal resections. The model detected 99% (135/137) of resection cavities. MELD-PostOp runtime was 17s per MRI, compared to 612s (ResectVol) and 3205s (Epic-CHOP). MELD-PostOp performance remained high across clinical and imaging subgroups (median DSC > 0.8). SignificanceMELD-PostOp provides an accurate, efficient and generalisable solution for postoperative resection cavity segmentation using only postoperative MRI scans. This open-source tool facilitates large-scale quantitative analysis to define what tissue is essential to resect for optimal epilepsy surgical outcomes. Key points- MELD-PostOp automatically segments resection cavities from a postoperative MRI scan in 17 seconds, a 100-200x faster runtime than existing tools - The model was trained on 965 annotated T1-weighted postoperative scans from sites worldwide, supporting its generalisation across diverse clinical and imaging settings. - The model achieves high segmentation accuracy (DSC > 0.8), with consistently high-quality segmentations across age groups, sex, image resolutions, and resection locations.