A multimodal study of brain regions in pediatric epilepsy: anatomical, radiological, and machine learning approaches
CUKUROVA MEDICAL JOURNAL, vol.51, no.1, pp.1-14, 2026 (ESCI, TRDizin)
- Publication Type: Article / Article
- Volume: 51 Issue: 1
- Publication Date: 2026
- Doi Number: 10.17826/cumj.1718730
- Journal Name: CUKUROVA MEDICAL JOURNAL
- Journal Indexes: Emerging Sources Citation Index (ESCI), Directory of Open Access Journals, TR DİZİN (ULAKBİM)
- Page Numbers: pp.1-14
- Ataturk University Affiliated: Yes
Abstract
Purpose: This study aims to investigate the anatomical and radiological differences in brain regions between pediatric epilepsy patients and age-matched healthy controls without neurological sequelae, by machine learning techniques. Materials and Methods: This study comprised 78 neurologically healthy children (mean age: 7.25 +/- 4.25 years) and 101 pediatric patients diagnosed with epilepsy (mean age: 6.91 +/- 4.34 years), all of whom underwent comprehensive anatomical and radiological evaluation. Magnetic Resonance Imaging was used to measure the volumes of the amygdaloid complex, hippocampus, insula, thalamus, caudate nucleus, the largest area of the brain, and the corpus callosum. Results: These volumetric measurements were generally higher in healthy children compared to those with epilepsy. Notably, the differences in the volumes of the amygdaloid complex, insula, and corpus callosum were statistically significant between the epilepsy and control groups. Significant reductions in the volumes of key brain regions including the hippocampus, corpus callosum, thalamus, and overall cortical surface in children with epilepsy compared to healthy controls. Conclusion: This study may contribute Machine learning methods combined with carefully selected features have demonstrated exceptional accuracy and reliability in the automatic detection of epilepsy, underscoring their significant potential to enhance clinical diagnostics and facilitate personalized treatment strategies.