This is a page that curates AI-related papers published worldwide. All content here is summarized using Google Gemini and operated on a non-profit basis. Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.
EndoControlMag: Robust Endoscopic Vascular Motion Magnification with Periodic Reference Resetting and Hierarchical Tissue-aware Dual-Mask Control
Created by
Haebom
Author
An Wanga, Rulin Zhou, Mengya Xu, Yiru Ye, Longfei Gou, Yiting Chang, Hao Chen, Chwee Ming Lim, Jiankun Wang, Hongliang Ren
Outline
EndoControlMag is a training-free Lagrange-based framework for visualizing microscopic vessel motion during endoscopic surgery. It is designed for precise surgery and decision-making in complex and dynamic surgical environments, using a periodic re-reference (PRR) technique to prevent error accumulation and a tissue-aware hierarchical augmentation (HTM) framework. HTM tracks the vessel center using a pre-trained visual tracking model and adjusts the augmentation effect of surrounding tissue through two adaptive softening strategies: motion-based or distance-based. Evaluation results using the EndoVMM24 dataset show that it outperforms existing methods in terms of accuracy and visual quality, while maintaining robustness across a variety of surgical conditions. The code, dataset, and video results are available at https://szupc.github.io/EndoControlMag/ .
We improve the accuracy and efficiency of visualizing microvascular motion during endoscopic surgery with a Lagrange-based framework that requires no training.
◦
Effectively reduce error accumulation and visual noise through PRR and HTM modules.
◦
It demonstrates robust performance in various surgical environments (occlusion, instrument interference, field of view changes, vascular deformation, etc.).
◦
It has demonstrated superior performance in quantitative indicators and expert evaluations compared to existing methods.
◦
We increased the reproducibility and scalability of our research through open code and datasets.
•
Limitations:
◦
The EndoVMM24 dataset may be limited in scope and further evaluation across a wider range of surgical types and settings is needed.
◦
Specific information on real-time processing speed is lacking. Further studies may be needed to determine the feasibility of real-time applications in real surgical settings.
◦
Clarification of the selection criteria for exercise-based and distance-based softening strategies may be needed. Additional research on automatic strategy selection may be needed.
