EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Predicting Epileptogenic Tubers in Patients With Tuberous Sclerosis Complex Using a Fusion Model Integrating Lesion Network Mapping and Machine Learning

Tinghong Liu, Qi Wang, Suhui Kuang, Dezhi Cao, Ping Ding, Shaohui Zhang, Haihua Wei, Zhirong Wei, Jinshan Xu, Xinyu Huang, Bing Liu iD, Shuli Liang iD

Annals of Clinical and Translational Neurology · 2025

DOI: 10.1002/acn3.70277 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Annals of Clinical and Translational Neurology, 2025. Verfasst von Tinghong Liu und 11 weiteren Autor:innen. Beteiligte Einrichtung: Functional Neurosurgery Department, National Center for Children's Health, Beijing Children's Hospital, Capital Medical University, Beijing, China. Hauptthemen laut MEDLINE: Tuberous Sclerosis, Machine Learning, Epilepsy, Brain Mapping. Die Arbeit nennt 28 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0). Gefördert von: National Natural Science Foundation of China, Beijing Municipal Natural Science Foundation, Beijing Municipal Science and Technology Commission, Administrative Commission of Zhongguancun Science Park.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Objective

Accurate localization of epileptogenic tubers (ETs) in patients with tuberous sclerosis complex (TSC) is essential but challenging, as these tubers lack distinct pathological or genetic markers to differentiate them from other cortical tubers. Approximately 60% of patients fail to have their ETs identified through noninvasive preoperative evaluations, creating an urgent clinical need for effective, noninvasive localization strategies.

Methods

A novel fusion model was developed, integrating lesion network mapping‐based risk assessment with a machine learning prediction model that utilizes brain functional connectivity and random forest algorithms. The model was built based on magnetic resonance imaging data. Retrospective analysis was conducted on patients with TSC‐related epilepsy who had undergone resective surgery and achieved seizure freedom at the 1‐year follow‐up; tubers were classified as true epileptogenic tubers (true ETs) or true non‐epileptogenic tubers (true non‐ETs) according to the resected regions. The model calculated and ranked the probability of each tuber being an ET for every patient.

Results

A total of 47 patients were enrolled in the study. The fusion model successfully ranked the true ETs within the top three in 91% of the cases. Significant differences in the probability rankings of ETs were observed among true ETs, true non‐ETs, and random tubers ( p < 0.01). Receiver operating characteristic curves were plotted to evaluate the accuracy of true ET localization across different methods, and the fusion model exhibited an area under the curve of 0.86. This performance significantly outperformed that of scalp electroencephalography, semiology, and positron emission tomography based on structural magnetic resonance imaging in the same cohort. Cross‐validation in three independent epilepsy centers confirmed the model's high generalizability.

Interpretation

Overall, this fusion model demonstrates high accuracy and robust clinical utility as a noninvasive tool for the localization of ETs. It effectively addresses the current challenges in identifying ETs, providing valuable support for surgical planning in patients with TSC‐related epilepsy.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Tuberous SclerosisMachine LearningEpilepsyBrain MappingHumansMaleFemaleMagnetic Resonance ImagingChildRetrospective StudiesAdolescentChild, PreschoolYoung AdultAdult

Stichwörter der Autor:innen

epilepsyepileptogenic tuberfusion modelmagnetic resonance imagingtuberous sclerosis complex

Autor:innen und Institutionen

  • Functional Neurosurgery Department, National Center for Children's Health, Beijing Children's Hospital, Capital Medical University, Beijing, China.
  • Key Laboratory of Major Diseases in Children, Ministry of Education, Beijing, China.
  • Brainnetcome Center and National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, Beijing, China.
  • School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China.
  • Epilepsy Center, Neurology Department, Shenzhen Children's Hospital. Shenzhen, Shenzhen, Guangdong, China.
  • Neurosurgery Department, PLA General Hospital, Beijing, China.

ORCID-Profile: Bing Liu, Shuli Liang

Förderung

  • National Natural Science Foundation of China
  • Beijing Municipal Natural Science Foundation
  • Beijing Municipal Science and Technology Commission, Administrative Commission of Zhongguancun Science Park
  • Beijing Municipal Hospital Administration Center
  • Natural Science Foundation of Beijing Municipality
  • Beijing Municipal Science and Technology Commission, Adminitrative Commission of Zhongguancun Science Park

Zitieren

APA

Liu, T., Wang, Q., Kuang, S., Cao, D., Ding, P., Zhang, S., Wei, H., Wei, Z., Xu, J., Huang, X., Liu, B., & Liang, S. (2025). Predicting Epileptogenic Tubers in Patients With Tuberous Sclerosis Complex Using a Fusion Model Integrating Lesion Network Mapping and Machine Learning. Annals of Clinical and Translational Neurology. https://doi.org/10.1002/acn3.70277

Vancouver

Liu T, Wang Q, Kuang S, Cao D, Ding P, Zhang S, et al. Predicting Epileptogenic Tubers in Patients With Tuberous Sclerosis Complex Using a Fusion Model Integrating Lesion Network Mapping and Machine Learning. Annals of Clinical and Translational Neurology. 2025. doi:10.1002/acn3.70277

BibTeX
@article{Liu2025,
  author = {Liu, Tinghong and Wang, Qi and Kuang, Suhui and Cao, Dezhi and Ding, Ping and Zhang, Shaohui and Wei, Haihua and Wei, Zhirong and Xu, Jinshan and Huang, Xinyu and Liu, Bing and Liang, Shuli},
  title = {Predicting Epileptogenic Tubers in Patients With Tuberous Sclerosis Complex Using a Fusion Model Integrating Lesion Network Mapping and Machine Learning},
  journal = {Annals of Clinical and Translational Neurology},
  year = {2025},
  doi = {10.1002/acn3.70277}
}
Kostenlos starten →