Predicting Antiseizure Medication Treatment in Children with Rare Tuberous Sclerosis Complex–Related Epilepsy Using Deep Learning

• DOI: 10.3174/ajnr.a8053
• Publication Date: 2023-12-11T19:57:41Z
• Authors: Haifeng Wang; Zhanqi Hu; Dian Jiang; Rongbo Lin; Cailei Zhao; Xia Zhao; Yihang Zhou; Yanjie Zhu; Hongwu Zeng; Dong Liang; Jianxiang Liao; Zhicheng Li
• Keywords: Fluid-attenuated inversion recovery

<h3>BACKGROUND AND PURPOSE:</h3> Tuberous sclerosis complex disease is a rare, multisystem genetic disease, but appropriate drug treatment allows many pediatric patients to have positive outcomes. The purpose of this study was predict the effectiveness antiseizure medication in children with tuberous complex–related epilepsy. <h3>MATERIALS METHODS:</h3> We conducted retrospective involving 300 included analysis clinical data and T2WI FLAIR images. consisted sex, age onset, at imaging, infantile spasms, numbers. To forecast treatment, we developed multitechnique deep learning method called WAE-Net. This used multicontrast MR imaging data. images were combined as FLAIR3 enhance contrast between lesions normal brain tissues. trained data-based model using fully connected network above-mentioned variables. After that, weighted-average ensemble built from ResNet3D architecture created final model. <h3>RESULTS:</h3> experiments had shown that numbers significantly different 2 drug-treatment outcomes (<i>P</i> &lt; .05). hybrid technique could accurately localize lesions, proposed achieved best performance (area under curve = 0.908 accuracy 0.847) testing cohort among compared methods. <h3>CONCLUSIONS:</h3> rare epilepsy be strong baseline for future studies.