Convolutional neural network‐aided tuber segmentation in tuberous sclerosis complex patients correlates with electroencephalogram

Abstract

Objective: One of the clinical hallmarks of tuberous sclerosis complex (TSC) is radiologically identified cortical tubers, which are present in most patients.Intractable epilepsy may require surgery, often involving invasive diagnostic procedures such as intracranial electroencephalography (EEG). Identifying the location of the dominant tuber responsible for generating epileptic activities is a critical issue. However, the link between cortical tubers and epileptogenesis is poorly understood. Given this, we hypothesized that tuber voxel intensity may be an indicator of the dominant epileptogenic tuber. Also, via tuber segmentation based on deep learning, we explored whether an automatic quantification of the tuber burden is feasible. Methods:We annotated tubers from structural magnetic resonance images across 29 TSC subjects, summarized tuber statistics in eight brain lobes, and determined suspected epileptogenic lobes from the same group using EEG monitoring data. Then, logistic regression analyses were performed to demonstrate the linkage between the statistics of cortical tuber and the epileptogenic zones. Furthermore, we tested the ability of a neural network to identify and quantify tuber burden.Results: Logistic regression analyses showed that the volume and count of tubers per lobe, not the mean or variance of tuber voxel intensity, were positively correlated with electrophysiological data. In 47.6% of subjects, the lobe with the largest tuber volume concurred with the epileptic brain activity. A neural network model on the test dataset showed a sensitivity of .83 for localizing individual tubers. The predicted masks from the model correlated highly with the neurologist labels, and thus may be a useful tool for determining tuber burden and searching for the epileptogenic zone.

Deep Learning Approaches for Imaging-Based Automated Segmentation of Tuberous Sclerosis Complex

“…The study participants were epilepsy patients seeking treatment at the Epilepsy Center, Beijing Fengtai Hospital and Beijing Tiantan Hospital between January 2015 and December 2020. Subjects included in the study met the diagnostic criteria of TSC according to the International TSC Consensus Group in 2012 (Supplemental Material Content S1) [14]. Four classic central nervous system findings are associated with TSC: cortical and subcortical tubers, cerebral white matter heterotopia, subependymal nodules (SENs), and subependymal giant cell astrocytoma (SEGAs) [15].…”

“…The epileptogenic tubers were unambiguously defined only if the patient received surgery and remained seizure-free in the last follow-up [14]. Unfortunately, surgical resection and resultant data on seizure freedom were unavailable for some of our patients.…”

“…The current state of machine learning for cybersecurity presents several challenges and limitations, including issues such as data quality, feature selection, and class imbalance, which warrant careful consideration for the development of innovative solutions [12,13]. To address these challenges, this study utilized diverse datasets and preprocessing protocols to train the model, with the aim of exploring the minimal input requirements for epileptogenic tuber segmentation [14].…”

Recognizing Pediatric Tuberous Sclerosis Complex Based on Multi-Contrast MRI and Deep Weighted Fusion Network

“…Previously, deep learning models and multi-contrast MRIs have been successfully used for automatically detecting strokes [15] and classifying brain tissues [16]. Until now, convolutional neural networks (CNNs) have been applied to assist in tuber segmentation in TSC patients [17]. Sanchez et al [18] used two types of contrast MRI, T2W and FLAIR, for the detection task of TSC tubers and achieved the receiver operating characteristic curve that can have an area under the curve (AUC) of 0.99.…”