Related Experiment Video
Updated: Nov 28, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.2K
A deep learning framework for quality assessment and restoration in video endoscopy
Sharib Ali1, Felix Zhou2, Adam Bailey3
1Institute of Biomedical Engineering and Big Data Institute, Oxford, UK; University of Oxford, Old Road Campus, Oxford, UK; Oxford NIHR Biomedical Research Centre, Oxford, UK.
Medical Image Analysis
|November 27, 2020
Summary
This study introduces an automated framework to detect, classify, and restore multiple artifacts in endoscopy videos, significantly improving image quality and enabling more reliable automated analysis for clinical applications.
Area of Science:
- Medical Imaging
- Computer Vision
- Artificial Intelligence
Background:
- Endoscopy is crucial for diagnosis and surgery, but artifacts like blur and reflections hinder interpretation.
- Current methods address only specific artifacts, leaving a need for comprehensive solutions.
Purpose of the Study:
- To develop a fully automatic framework for detecting, classifying, segmenting, and restoring multiple artifacts in endoscopy videos.
- To enhance the reliability of automated analysis and visual interpretation of endoscopic imagery.
Main Methods:
- Utilized a fast, multi-scale, single-stage convolutional neural network detector for artifact classification.
- Employed an encoder-decoder model for pixel-wise segmentation of irregular artifact shapes.
- Integrated generative adversarial networks (GANs) for frame restoration, including deblurring, saturation correction, and inpainting.
Main Results:
- The framework achieved high mean average precision (mAP) for artifact detection with near real-time processing (88 ms).
- Restoration models showed significant improvements over existing methods for deblurring, saturation correction, and inpainting.
- An average of 68.7% of video frames met quality standards (≥0.9) post-restoration, retaining 25% more frames.
Conclusions:
- The proposed framework offers a comprehensive solution for endoscopy video artifact management.
- Artifact detection and restoration enhance the robustness of medical image analysis methods.
- This work advances automated analysis and interpretation in clinical endoscopy.
