Francesco Cardinale | Searcharxiv

arXiv2024

Desynchronization Index: a New Connectivity Approach for Exploring Epileptogenic Networks

In drug-resistant epilepsy, Stereo-Electroencephalography (SEEG) monitoring is one of the most effective techniques to identify the Epileptogenic Zone (EZ), the fundamental prerequisite for epilepsy surgery. Despite recent technological advances, SEEG recordings remain difficult to interpret, and SEEG-guided surgery still achieves success rates below 70%. In this work, we develop a novel computational framework for SEEG analysis, with the ultimate aim of improving the accuracy in EZ definition. Specifically, we investigate the hypothesis that epileptogenic regions exhibit a tendency to behave independently and thus desynchronize from neighboring brain structures before seizure onset. To this end, we design the Desynchronization Index (DI), an algorithm that identifies the Epileptogenic Zone (EZ) as the subset of channels that disconnect from the SEEG network during the ictal transition. We evaluate the DI algorithm against Epileptogenicity Index (EI), one of the most common tools for EZ definition, on a clinical dataset of 20 patients, considering the channels that were thermocoagulated at the end of SEEG monitoring as the detection target. Our results show that DI overcomes EI in terms of area under the ROC curve (AUC=0.86 vs. AUC=0.83), while combining the two algorithms into a single framework leads to the best performance (AUC=0.88). Overall, the DI algorithm underscores anomalous connectivity patterns that are difficult to detect through visual inspection, improving the accuracy in the EZ definition and providing new insights into the dynamics of seizure generation.

eess.SP↗

arXiv2021

A self-supervised learning strategy for postoperative brain cavity segmentation simulating resections

Accurate segmentation of brain resection cavities (RCs) aids in postoperative analysis and determining follow-up treatment. Convolutional neural networks (CNNs) are the state-of-the-art image segmentation technique, but require large annotated datasets for training. Annotation of 3D medical images is time-consuming, requires highly-trained raters, and may suffer from high inter-rater variability. Self-supervised learning strategies can leverage unlabeled data for training. We developed an algorithm to simulate resections from preoperative magnetic resonance images (MRIs). We performed self-supervised training of a 3D CNN for RC segmentation using our simulation method. We curated EPISURG, a dataset comprising 430 postoperative and 268 preoperative MRIs from 430 refractory epilepsy patients who underwent resective neurosurgery. We fine-tuned our model on three small annotated datasets from different institutions and on the annotated images in EPISURG, comprising 20, 33, 19 and 133 subjects. The model trained on data with simulated resections obtained median (interquartile range) Dice score coefficients (DSCs) of 81.7 (16.4), 82.4 (36.4), 74.9 (24.2) and 80.5 (18.7) for each of the four datasets. After fine-tuning, DSCs were 89.2 (13.3), 84.1 (19.8), 80.2 (20.1) and 85.2 (10.8). For comparison, inter-rater agreement between human annotators from our previous study was 84.0 (9.9). We present a self-supervised learning strategy for 3D CNNs using simulated RCs to accurately segment real RCs on postoperative MRI. Our method generalizes well to data from different institutions, pathologies and modalities. Source code, segmentation models and the EPISURG dataset are available at https://github.com/fepegar/ressegijcars .

eess.IV↗