Endoscopic Image Processing for Polyp Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current endoscopic procedures face challenges in detecting small, flat, and concealed polyps within the colon due to their similarity in texture and color to regular colon tissue, leading to missed polyps during colonoscopy, with regular tests missing up to 28% of all polyps and 24% of adenoma polyps, which are critical for cancer detection.
Innovation Solution
An image processing system connected to the endoscopy device that employs real-time video frame processing, utilizing conventional detectors and machine learning algorithms, including Support Vector Machines, Convolutional Neural Networks, and Long-Short Term Memory Recurrent Neural Networks, to identify and classify suspicious regions by enhancing image quality, removing irrelevant frames, and applying adaptive segmentation techniques to highlight polyps and other suspicious tissues.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional visual inspection methods are used during colonoscopy, then the procedure remains simple and quick to perform, but the detection precision for small and flat polyps is low, missing up to 28% of all polyps and 24% of adenoma polyps
Solution Approach 1:
The patent introduces an image processing system as an intermediary between the endoscope camera and the physician's visual inspection. This system includes a processor that receives video frames from the endoscope, applies machine learning algorithms (such as convolutional neural networks and recurrent neural networks) to detect suspicious regions, and overlays detection results on the video feed. The intermediary system enhances detection precision without requiring the physician to perform complex manual analysis, thus resolving the contradiction between improved detection and system complexity.
Solution Approach 2:
The patent replaces the mechanical/visual inspection method with an automated image processing and machine learning-based detection system. Instead of relying solely on the physician's visual capabilities, the system uses computer vision algorithms to automatically identify polyps in real-time video frames from the endoscope. This substitution significantly improves detection precision while the automated nature of the system keeps the operational complexity manageable.
2Productivity
If the endoscope is moved quickly through the colon to complete the procedure efficiently, then the productivity of the colonoscopy increases, but the detection precision of polyps decreases due to limited observation time
Solution Approach 1:
The patent implements continuous real-time image processing and analysis throughout the entire colonoscopy procedure. The machine learning system continuously processes video frames as they are captured by the endoscope, providing ongoing detection and feedback without interruption. This continuous analysis ensures that polyps are detected regardless of the endoscope's movement speed, maintaining high detection precision while allowing the procedure to proceed efficiently without repeated stops for manual inspection.
Solution Approach 2:
The system provides real-time feedback to the physician by overlaying detected suspicious regions directly on the video feed displayed during the procedure. This immediate feedback allows the physician to adjust the endoscope's position or focus on identified areas without leaving the procedure flow, thus maintaining both high productivity and detection precision. The feedback mechanism ensures that no potential polyps are missed even when the endoscope moves quickly through the colon.
3Measurement precision
If multiple detection methods and machine learning algorithms are applied to improve polyp detection accuracy, then the detection precision increases, but the processing time and computational complexity increase
Solution Approach 1:
The patent employs preliminary processing steps that prepare video frames before applying complex machine learning algorithms. This includes pre-processing operations such as noise reduction, contrast enhancement, and region of interest identification that simplify subsequent detection tasks. By performing these preliminary actions, the system reduces the computational burden on the main detection algorithms, thereby maintaining high detection accuracy while minimizing processing time and preventing delays in the colonoscopy procedure.
Data Source
Figure 1A~1B
Figure 2
Figure 3
AI summary
An image processing system connected to an endoscope and processing in real-time endoscopic images to identify suspicious tissues such as polyps or cancer. The system applies preprocessing tools to clean the received images and then applies in parallel a plurality of detectors both conventional detectors and models of supervised machine learning-based detectors. A post processing is also applied in order select the regions which are most probable to be suspicious among the detected regions. Frames identified as showing suspicious tissues can be marked on an output video display. Optionally, the size, type and boundaries of the suspected tissue can also be identified and marked.