Artificial Neural Network Analysis to Identify Risk Factors for Drug-resistant Epilepsy
Abstract number :
2.078
Submission category :
16. Epidemiology
Year :
2024
Submission ID :
440
Source :
www.aesnet.org
Presentation date :
12/8/2024 12:00:00 AM
Published date :
Authors :
Presenting Author: Roberta Coa, MD, PhD – AOU Cagliari
Davide Fonti, MD – PO Sirai, ASL Carbonia
Federico Arippa, PhD – University of Cagliari
Rosamaria Lecca, MD, PhD – ASL Cagliari
Lorenzo Polizzi, MD – AOU Cagliari
Antonella Muroni, MD, PhD – AOU Cagliari
Marta Melis, MD, PhD – AOU Cagliari
Michela Figorilli, MD, PhD – AOU Cagliari
Enzo Grossi, PhD – Villa Santa Maria Foundation
Monica Puligheddu, MD, PhD – University of Cagliari
Rationale: Thirty percent of individuals with epilepsy experience drug resistance, meaning they continue to have seizures after undergoing at least two drug trials with appropriate, tolerated, full-dose medications, leading to an increased risk of Sudden Unexpected Death in Epilepsy (SUDEP) and psychiatric comorbidity, and a reduced quality of life. We aimed to identify the risk factors of DRE through the analysis of clinical data of subjects referred to the AOU Cagliari Epilepsy Center.
Methods: Data were extracted from medical records of the patients treated at our Epilepsy Centre. In order to identify predictive risk factors for DRE we used a combination of univariate analysis and logistic regression to evaluate the combined effect of factors. We analyzed different conditions linked to DRE: gender, age of onset, etiology (structural, genetic…) history of status epilepticus, type of seizures (focal, generalized, combined), learning disabilities, comorbidities, illness duration, and family history of epilepsy.
Furthermore, data mining approach with fourth-generation artificial neural networks has been used in order to discover subtle trends and associations for DRE among the different variables.
Results: Results: 183 out of a total of 724 patients were identified as DRE (25,2%). Univariate analysis and subsequent multiple regression showed a significant association for DRE for a few factors: age of onset, structural etiology, psychiatric and neurological comorbidities, and learning disability. ANN methodology and the auto-contractive map exploited the links among the full spectrum of variables, revealing the simultaneous connections among them, facing the complexity of DRE (fig1).
Conclusions: Our data show that the presence of a structural etiology, psychiatric and neurological comorbidities, learning disabilities, and early age of onset are significant risk factors for DRE. Early identification of these factors holds promise for improving management of affected individuals and refining clinical strategies.
Funding: IFAIR regional council funds
Epidemiology