2025
WaveMind: Towards a Generalist EEG Chat Model Aligned to Textual and Visual Modalities
Ziyi Zeng, Zhenyang Cai, Yixi Cai, Xidong Wang, Junying Chen, Rongsheng Wang, Siqi Cai, Haizhou Li, Benyou Wang
Preprint 2025
This study proposes unifying EEG signals and their paired modalities into a shared semantic space to resolve modality mismatch. It introduces WaveMind-Instruct-338k, the first cross-task instruction-tuning EEG dataset, enabling conversational EEG interpretation. The resulting MLLM achieves strong classification performance and supports open-ended dialogue across tasks.
WaveMind: Towards a Generalist EEG Chat Model Aligned to Textual and Visual Modalities
Ziyi Zeng, Zhenyang Cai, Yixi Cai, Xidong Wang, Junying Chen, Rongsheng Wang, Siqi Cai, Haizhou Li, Benyou Wang
Preprint 2025
This study proposes unifying EEG signals and their paired modalities into a shared semantic space to resolve modality mismatch. It introduces WaveMind-Instruct-338k, the first cross-task instruction-tuning EEG dataset, enabling conversational EEG interpretation. The resulting MLLM achieves strong classification performance and supports open-ended dialogue across tasks.
NeuroCLIP: A Multimodal Contrastive Learning Method for rTMS-treated Methamphetamine Addiction Analysis
Chengkai Wang, Di Wu, Yunsheng Liao, Wenyao Zheng, Cheng PengChai, Ziyi Zeng, Xurong Gao, Hemmings Wu, Zhoule Zhu, Jie Yang, Lihua Zhong, Weiwei Cheng, Yun-Hsuan Chen, Mohamad Sawan
Preprint 2025
This study proposes NeuroCLIP, a deep multimodal framework fusing EEG and fNIRS via progressive learning. It objectively assesses methamphetamine dependence and rTMS treatment effects, outperforming unimodal models. The resulting biomarker is robust, clinically interpretable, and correlates strongly with craving scores.
NeuroCLIP: A Multimodal Contrastive Learning Method for rTMS-treated Methamphetamine Addiction Analysis
Chengkai Wang, Di Wu, Yunsheng Liao, Wenyao Zheng, Cheng PengChai, Ziyi Zeng, Xurong Gao, Hemmings Wu, Zhoule Zhu, Jie Yang, Lihua Zhong, Weiwei Cheng, Yun-Hsuan Chen, Mohamad Sawan
Preprint 2025
This study proposes NeuroCLIP, a deep multimodal framework fusing EEG and fNIRS via progressive learning. It objectively assesses methamphetamine dependence and rTMS treatment effects, outperforming unimodal models. The resulting biomarker is robust, clinically interpretable, and correlates strongly with craving scores.
Specific Endophenotypes in EEG Microstates for Methamphetamine Use Disorder
Xurong Gao, Yun-Hsuan Chen, Ziyi Zeng, Wenyao Zheng, Cheng PengChai, Hemmings Wu, Zhoule Zhu, Jie Yang, Lihua Zhong, Hua Shen, Mohamad Sawan
Frontiers in Psychiatry 2025 Accepted
This study applied EEG microstate features classification within EEG across frequency bands and tasks to classify methamphetamine use disorder (MUD) versus controls. The main conclusion involves MUD endophenotyping, pinpointing alpha-band microstate A as a key biomarker.
Specific Endophenotypes in EEG Microstates for Methamphetamine Use Disorder
Xurong Gao, Yun-Hsuan Chen, Ziyi Zeng, Wenyao Zheng, Cheng PengChai, Hemmings Wu, Zhoule Zhu, Jie Yang, Lihua Zhong, Hua Shen, Mohamad Sawan
Frontiers in Psychiatry 2025 Accepted
This study applied EEG microstate features classification within EEG across frequency bands and tasks to classify methamphetamine use disorder (MUD) versus controls. The main conclusion involves MUD endophenotyping, pinpointing alpha-band microstate A as a key biomarker.
Repetitive TMS-based Identification of Methamphetamine-Dependent Individuals Using EEG Spectra
Ziyi Zeng, Yun-Hsuan Chen, Xurong Gao, Wenyao Zheng, Hemmings Wu, Zhoule Zhu, Jie Yang, Lihua Zhong, Weiwei Cheng, Mohamad Sawan
IEEE Sensors 2025 Accepted
This study uses EEG-based relative band power (RBP) analysis and random forest classification to assess treatment effects in Methamphetamine use disorders (MUD) patients. Gamma RBP was found to be a potential biomarker for rTMS efficacy and a candidate signal for closed-loop neuromodulation in MUD.
Repetitive TMS-based Identification of Methamphetamine-Dependent Individuals Using EEG Spectra
Ziyi Zeng, Yun-Hsuan Chen, Xurong Gao, Wenyao Zheng, Hemmings Wu, Zhoule Zhu, Jie Yang, Lihua Zhong, Weiwei Cheng, Mohamad Sawan
IEEE Sensors 2025 Accepted
This study uses EEG-based relative band power (RBP) analysis and random forest classification to assess treatment effects in Methamphetamine use disorders (MUD) patients. Gamma RBP was found to be a potential biomarker for rTMS efficacy and a candidate signal for closed-loop neuromodulation in MUD.
DentalGPT: Incentivizing Multimodal Complex Reasoning in Dentistry
Zhenyang Cai, Jiaming Zhang, Junjie Zhao, Ziyi Zeng, Yanchao Li, Jingyi Liang, Junying Chen, Yunjin Yang, Jiajun You, Shuzhi Deng, Tongfei Wang, Wanting Chen, Chunxiu Hao, Ruiqi Xie, Zhenwei Wen, Xiangyi Feng, Zou Ting, Jin Zou Lin, Jianquan Li, Guangjun Yu, Liangyi Chen, Junwen Wang, Shan Jiang, Benyou Wang
arXiv Preprint 2025
DentalGPT is a specialized dental multimodal large language model that addresses limitations in capturing fine-grained dental visual details and performing complex reasoning. Using the largest annotated dental dataset (120k+ images) and reinforcement learning, it achieves superior performance in disease classification and dental VQA tasks, outperforming larger MLLMs with only 7B parameters.
DentalGPT: Incentivizing Multimodal Complex Reasoning in Dentistry
Zhenyang Cai, Jiaming Zhang, Junjie Zhao, Ziyi Zeng, Yanchao Li, Jingyi Liang, Junying Chen, Yunjin Yang, Jiajun You, Shuzhi Deng, Tongfei Wang, Wanting Chen, Chunxiu Hao, Ruiqi Xie, Zhenwei Wen, Xiangyi Feng, Zou Ting, Jin Zou Lin, Jianquan Li, Guangjun Yu, Liangyi Chen, Junwen Wang, Shan Jiang, Benyou Wang
arXiv Preprint 2025
DentalGPT is a specialized dental multimodal large language model that addresses limitations in capturing fine-grained dental visual details and performing complex reasoning. Using the largest annotated dental dataset (120k+ images) and reinforcement learning, it achieves superior performance in disease classification and dental VQA tasks, outperforming larger MLLMs with only 7B parameters.