Automated Polyp Detection via Color Filtering and Edge Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current polyp detection methods during optical colonoscopy face challenges due to reliance on shape and texture, which are unreliable, leading to missed detections, especially for flat and pedunculated polyps, as they are susceptible to partial segmentation and vary in morphology.
Innovation Solution
A system and method for automated polyp detection using color filtering, edge pixel localization, and a vote accumulation scheme that generates probabilistic outputs, enhancing low-level features and distinguishing polyp boundaries from other colonic objects through a two-stage classification framework, providing rotation and illumination invariance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If shape-based polyp detection is used, then detection can be performed, but detection accuracy deteriorates due to partial segmentation and morphological variability
Solution Approach 1:
The patent applies color filtering to enhance the contrast between polyps and surrounding tissue. By transforming the original color image into multiple color-filtered images (e.g., using red, green, blue channels or other color spaces), the system emphasizes color differences that are characteristic of polyps, making them more distinguishable from normal colon tissue regardless of shape or segmentation quality.
Solution Approach 2:
The patent segments the image processing into multiple stages: color filtering, edge detection, oriented patch extraction, and classification. This multi-stage segmentation allows each component to focus on specific features, with the classification system learning to recognize polyp patterns from oriented patches that capture local structural information invariant to rotation and illumination changes.
2Ease of manufacture
If texture-based polyp detection is used, then detection can be performed, but reliability deteriorates due to camera-polyp distance dependency
Solution Approach 1:
The patent transitions from analyzing texture in the original image space to analyzing oriented image patches in a transformed space. By extracting patches oriented along edge directions and representing them with rotation-invariant descriptors, the system creates a new dimensional representation that is insensitive to camera distance and orientation changes, thereby improving reliability.
3Productivity
If automated detection system is implemented, then productivity increases, but device complexity increases
Solution Approach 1:
The patent performs preliminary color filtering and edge detection before the main classification stage. By pre-processing the images to enhance relevant features and reduce noise, the subsequent classification system operates on more informative data, which improves accuracy without requiring an excessively complex classifier. The oriented patch extraction also prepares data in advance in a rotation-invariant manner.
Solution Approach 2:
The patent introduces oriented image patches as an intermediary representation between the original image and the classification system. These patches, extracted and oriented along edge directions, serve as a bridge that captures local structural information in a standardized format, making the classification task more manageable and accurate while keeping the overall system architecture modular and organized.
4Measurement precision
If edge detection is applied to locate polyp boundaries, then boundary localization is achieved, but false detections increase due to irrelevant objects in complex endoluminal scene
Solution Approach 1:
The patent applies different processing and analysis to different parts of the image. By extracting oriented patches centered on edge pixels and analyzing their local intensity patterns, the system evaluates each edge segment in its local context rather than treating all edges uniformly. The classification system learns to distinguish polyp boundaries from other structures based on local patch characteristics, reducing false detections from irrelevant objects.
Data Source
AI summary
A system and method for automated polyp detection in optical colonoscopy images is provided. In one embodiment, the system and method for polyp detection is based on an observation that image appearance around polyp boundaries differs from that of other boundaries in colonoscopy images. To reduce vulnerability against misleading objects, the image processing method localizes polyps by detecting polyp boundaries, while filtering out irrelevant boundaries, with a generative-discriminative model. To filter out irrelevant boundaries, a boundary removal mechanism is provided that captures changes in image appearance across polyp boundaries. Thus, in this embodiment the boundary removal mechanism is minimally affected by texture visibility limitations. In addition, a vote accumulation scheme is applied that enables polyp localization from fragmented edge segmentation maps without identification of whole polyp boundaries.