辜嘉
下載導師簡歷CV
Section 1: Personal Particulars
辜嘉
Gu Jia
Professor
63193009
jiagu@cityu.edu.mo
澳門城市大學
澳門城市大學-數據科學學院
Research Project provided for Macao Youth Scholars Program:
消化內鏡定位與地圖構建系統關鍵技術研發
Research on Key Technologies for Digestive Endoscopy Localization and Map Reconstruction System
0831 生物医学工程
0812 计算机科学与技术
At least three first-author, peer-reviewed research paper accepted or published in Q1 Journals
Must have a PhD degree in QS500 university
Section 2: Research Interests and Grants
《Research on Key Technologies for Digestive Endoscopy Localization and Map Reconstruction System》，funded by the Science and Technology Development Fund of Macao under Grant0002/2024/RIA1, 1,720,000MOP, duration: 3years
Endoscopic Submucosal Dissection (ESD) offers advantages such as minimal invasiveness and rapid recovery, serving as an effective method for radical treatment of early-stage gastrointestinal tumors. The critical scientific challenge lies in achieving fast and high-precision real-time lesion localization and mapping, which is essential for investigating the correlation mechanism between lesion resection rates and postoperative outcomes. Addressing research difficulties including submillimeter-level monocular depth estimation, real-time pose computation, and 3D dense mapping—along with exploring the impact of complex in vivo environments and low resolution on SLAM accuracy—this project designs a low-latency, high-performance reconstruction method incorporating a graph neur
10
4
1
《Research on Key Technologies for Digestive Endoscopy Localization and Map Reconstruction System》，funded by the Science and Technology Development Fund of Macao under Grant0002/2024/RIA1, 1,720,000MOP, duration: 3years
Endoscopic Submucosal Dissection (ESD) offers advantages such as minimal invasiveness and rapid recovery, serving as an effective method for radical treatment of early-stage gastrointestinal tumors. The critical scientific challenge lies in achieving fast and high-precision real-time lesion localization and mapping, which is essential for investigating the correlation mechanism between lesion resection rates and postoperative outcomes. Addressing research difficulties including submillimeter-level monocular depth estimation, real-time pose computation, and 3D dense mapping—along with exploring the impact of complex in vivo environments and low resolution on SLAM accuracy—this project designs a low-latency, high-performance reconstruction method incorporating a graph neural network-refined generative adversarial network, multimodal input, and an attention mechanism-based endoscopic pose estimation model. It validates the feasibility of 3D reconstruction of submucosal lesions using monocular endoscopy without CT assistance and assists physician diagnosis and surgery through real-time rendering. This project provides theoretical models and methodological support for understanding the correlation mechanism between early cancer resection rates and prognostic quality, integrates algorithms via hardware-software integration, and promotes industrialization through commercial applications.
Endoscopic Submucosal Dissection (ESD) offers advantages such as minimal invasiveness and rapid recovery, serving as an effective method for radical treatment of early-stage gastrointestinal tumors. The critical scientific challenge lies in achieving fast and high-precision real-time lesion localization and mapping, which is essential for investigating the correlation mechanism between lesion resection rates and postoperative outcomes. Addressing research difficulties including submillimeter-level monocular depth estimation, real-time pose computation, and 3D dense mapping—along with exploring the impact of complex in vivo environments and low resolution on SLAM accuracy—this project designs a low-latency, high-performance reconstruction method incorporating a graph neural network-refined generative adversarial network, multimodal input, and an attention mechanism-based endoscopic pose estimation model. It validates the feasibility of 3D reconstruction of submucosal lesions using monocular endoscopy without CT assistance and assists physician diagnosis and surgery through real-time rendering. This project provides theoretical models and methodological support for understanding the correlation mechanism between early cancer resection rates and prognostic quality, integrates algorithms via hardware-software integration, and promotes industrialization through commercial applications.