Related Experiment Video
Updated: Jun 28, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.7K
Rapid and robust endoscopic content area estimation: A lean GPU-based pipeline and curated benchmark dataset
Charlie Budd1, Luis C Garcia-Peraza-Herrera1, Martin Huber1
1King's College London, UK.
Summary
Accurately estimating the endoscopic content area is crucial for medical imaging analysis. This study introduces a fast GPU-based pipeline and a new dataset, significantly improving content area detection accuracy and speed.
Area of Science:
- Medical image analysis
- Computer vision
- Surgical technology
Background:
- Estimating the endoscopic content area is vital for image processing pipelines.
- Current methods face challenges in real-time accuracy due to lack of investigation and benchmark datasets.
Purpose of the Study:
- To develop and evaluate a novel computational pipeline for accurate endoscopic content area estimation.
- To introduce a comprehensive dataset for benchmarking content area detection algorithms.
Main Methods:
- Proposed two lean GPU-based pipeline variants combining edge detection and circle fitting.
- Utilized handcrafted and learned features for edge point candidate extraction.
- Created a novel dataset with manual and pseudo-annotations of endoscopic content areas.
Main Results:
- Achieved significant improvements in accuracy (Hausdorff distance: 6.3 px vs. 118.1 px) compared to a U-Net approach.
- Demonstrated substantial reduction in computational time (0.13 ms/frame vs. 11.2 ms/frame).
- Publicly released the dataset and algorithm implementation for community use.
Conclusions:
- The proposed GPU-based pipeline offers superior accuracy and speed for endoscopic content area estimation.
- The new benchmark dataset facilitates further research and development in this critical area of medical imaging.
