EndoControlMag GitHub
Advanced Engineering Informatics Volume 72 · Article 104505 · 2026

EndoControlMag

Robust Endoscopic Vascular Motion Magnification with Periodic Reference Resetting and Hierarchical Tissue-aware Dual-Mask Control

An Wang* Rulin Zhou* Mengya Xu* Yiru Ye Longfei Gou Yiting Chang Hao Chen Chwee Ming Lim Jiankun Wang Hongliang Ren†

* Equal contribution    † Corresponding author

The Chinese University of Hong Kong · CUHK Shenzhen Research Institute · Clinical and engineering collaborators

Input Original endoscopic video
Ours · ×4 EndoControlMag
24 Endoscopic videos
4 Surgical procedures
4 Challenge categories
0 Task-specific training

TL;DR

EndoControlMag makes subtle vascular pulsations visible in dynamic surgical videos while keeping surrounding tissue stable—even through occlusion, deformation, instrument motion, and view change.

Abstract

Motion magnification built for real surgical scenes.

Accurate visualization of subtle vascular dynamics is a knowledge-intensive challenge in minimally invasive surgery. Conventional imaging systems struggle to reveal these imperceptible motions amidst camera movement, tissue deformation, and instrument occlusion.

EndoControlMag is a training-free Lagrangian framework that applies mask-conditioned magnification to target vessels while preserving surrounding tissue. Periodic Reference Resetting limits temporal drift, while Hierarchical Tissue-aware Magnification maintains spatial control with tracked inner masks and adaptively softened outer masks.

Method

Temporal stability meets tissue-aware control.

Two complementary mechanisms address the temporal and spatial failure modes of endoscopic motion magnification.

Graphical abstract showing the EndoControlMag framework
EndoControlMag combines Periodic Reference Resetting and Hierarchical Tissue-aware Magnification in a training-free pipeline.
01

Temporal consistency

Periodic Reference Resetting

The input video is divided into short overlapping clips. Each clip receives an updated reference frame, preventing error accumulation without introducing abrupt temporal transitions.

02

Spatial selectivity

Hierarchical Tissue-aware Magnification

A pretrained tracker follows the vessel core while motion- or distance-based softening controls amplification across nearby deformable tissue.

Qualitative results

One framework across four surgical procedures.

Switch procedures to compare the original footage with EndoControlMag. The paired videos remain synchronized.

LC

Laparoscopic Cholecystectomy

Magnification factor ×4

Input Original
Ours EndoControlMag
×2Subtle
×4Default
×8Enhanced
×16Strong
×32Extreme

Challenging scenarios

Robust when the surgical scene changes.

EndoControlMag maintains vessel localization and smooth magnification across four realistic sources of disruption.

01

Occlusion

Temporary coverage by surgical gauze or instruments.

Original
EndoControlMag
02

Tool disturbance

Strong local motion caused by instrument interaction.

Original
EndoControlMag
03

Vessel deformation

Non-rigid tissue stretching changes vessel appearance.

Original
EndoControlMag
04

View change

Large camera movement and temporary vessel disappearance.

Original
EndoControlMag

Evaluation

Measured for accuracy, quality, and clinical usefulness.

01

Image quality

SSIM, PSNR, and MUSIQ quantify fidelity and perceptual quality across magnification factors.

02

Magnification accuracy

Spatial-temporal analysis verifies periodic enhancement at the target vessel while surrounding tissue remains stable.

03

Surgeon assessment

A double-blind expert study evaluates visual quality, robustness, and perceived clinical usefulness.

Publication

Advanced Engineering Informatics

EndoControlMag: Robust endoscopic vascular motion magnification with periodic reference resetting and hierarchical tissue-aware dual-mask control

Volume
72
Article
104505
Published
May 2026
Access
Open access
doi.org/10.1016/j.aei.2026.104505
BibTeX
@article{wang2026endocontrolmag,
  title   = {EndoControlMag: Robust endoscopic vascular
             motion magnification with periodic reference
             resetting and hierarchical tissue-aware
             dual-mask control},
  author  = {Wang, An and Zhou, Rulin and Xu, Mengya and
             Ye, Yiru and Gou, Longfei and Chang, Yiting
             and Chen, Hao and Lim, Chwee Ming and
             Wang, Jiankun and Ren, Hongliang},
  journal = {Advanced Engineering Informatics},
  volume  = {72},
  pages   = {104505},
  year    = {2026},
  doi     = {10.1016/j.aei.2026.104505}
}