Deep learning applied in epilepsy: Bibliometric and visual analysis.

BACKGROUND

Epilepsy is a prevalent chronic neurological disease with significant impacts. Traditional diagnostic methods and machine learning face challenges due to the complexity of electroencephalography (EEG). Deep learning (DL) shows potential in epilepsy management through EEG processing and brain imaging analysis. This study aims to fill the gap of missing bibliometric analysis in this field by conducting a visual bibliometric analysis of DL applications in epilepsy, revealing research trends, hotspots, and cutting-edge developments to guide future research directions.

METHODS

In this study, a comprehensive search of original research articles and reviews, limited to English-language, published from 2006 to 2025, was conducted by the Web of Science Core Collection database. Bibliometric analyses and visualizations were conducted by CiteSpace, VOSviewer, and Bibliometrix.

RESULTS

This search retrieved 1266 papers related to DL applied in epilepsy from the Web of Science Core Collection database, showing a consistent upward trend. These papers were from 1957 organizations across 290 countries/regions, primarily from China and the United States. "Biomedical Signal Processing and Control" ranked as the journal with the most published papers. Acharya, U. Rajendra from Ngee Ann Polytechnic, was the most authoritative author. DL-based seizure detection, prediction, and epilepsy management are key research hotspots. Moreover, multimodal data integration approaches are gaining more attention.

CONCLUSIONS

This study innovatively employs bibliometric methods to visually analyze research on DL applications in epilepsy, which reveals a rising trend in research within the field and identifies key hotspots, such as DL in seizure detection and prediction. The potential of DL to enhance epilepsy management is highlighted, particularly in improving the accuracy of seizure detection and prediction, thereby enhancing patients' quality of life. Furthermore, the findings highlight the importance of increasing seizure prediction accuracy, exploring multimodal data integration, and developing more interpretable DL models for future research.