Scholarlink

Electroencephalography (EEG) and electrocorticography (ECoG) are complementary neuroimaging techniques, balancing non-invasiveness (EEG) and high spatial resolution (ECoG). Conventional EEG inverse solutions face spatial blurring and mislocalization due to mathematical constraints. We propose Cortical Signal Transformation Network (CSTNet), a deep learning framework that leverages Optimal Transport (OT) to directly map EEG scalp potentials to ECoG-equivalent cortical signals. Through OT-based geometric alignment of paired EEG-ECoG data, CSTNet bypasses explicit noise modeling assumptions while preserving cortical signal topology. This approach bridges EEG’s safety with ECoG-like accuracy, advancing applications in epilepsy surgery planning and brain-computer interfaces.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

CSTNet: A Generative Framework for EEG-to-ECoG Mapping via Optimal Transport

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ruslan Kalimullin,
  • Ekaterina Antipushina,
  • Alexandra Razorenova,
  • Georgiy Kormakov,
  • Nikolay Koshev

摘要

Electroencephalography (EEG) and electrocorticography (ECoG) are complementary neuroimaging techniques, balancing non-invasiveness (EEG) and high spatial resolution (ECoG). Conventional EEG inverse solutions face spatial blurring and mislocalization due to mathematical constraints. We propose Cortical Signal Transformation Network (CSTNet), a deep learning framework that leverages Optimal Transport (OT) to directly map EEG scalp potentials to ECoG-equivalent cortical signals. Through OT-based geometric alignment of paired EEG-ECoG data, CSTNet bypasses explicit noise modeling assumptions while preserving cortical signal topology. This approach bridges EEG’s safety with ECoG-like accuracy, advancing applications in epilepsy surgery planning and brain-computer interfaces.