Lesion Tracking via Siamese Network for Endoscopic Video
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current AI-assisted endoscopic detection methods for target lesions, such as colorectal polyps, face challenges in speed and robustness due to high frame rates and lack of timing information consideration, leading to inaccurate positioning and excessive alerts.
Innovation Solution
A target positioning method that detects a target lesion in a video frame and triggers a tracking process, using a siamese network to track the lesion across consecutive frames, incorporating timing information to enhance accuracy and reduce false alerts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If target detection method is used to detect each video frame image, then lesion detection accuracy is improved, but processing speed deteriorates and frame rate is missed
Solution Approach 1:
The system performs preliminary detection on selected frames to identify potential lesions, then uses tracking algorithms to follow the lesion across subsequent frames. This preliminary action on key frames rather than all frames resolves the contradiction by maintaining detection accuracy while improving processing speed.
Solution Approach 2:
The video stream processing is segmented into detection phases and tracking phases. Detection is performed on sampled frames rather than every frame, while tracking continuously monitors the lesion across all frames. This segmentation allows accurate lesion detection without processing every frame at full detection cost.
2Device complexity
If frame-by-frame detection is performed without timing information, then detection simplicity is maintained, but robustness deteriorates due to false positive alerts
Solution Approach 1:
The system uses tracking information as feedback to verify detection results. When a lesion is detected, the tracking module continuously monitors its position across frames, providing feedback that confirms whether the detected lesion is real or a false positive. This feedback mechanism improves robustness while maintaining relatively simple detection processes.
Solution Approach 2:
Instead of independent frame-by-frame detection, the system establishes continuous tracking of detected lesions across the video stream. This continuity allows the system to distinguish real lesions (which persist across frames) from false positives (which disappear), thereby improving reliability without significantly increasing detection complexity.
Data Source
AI summary
A target positioning method is provided to include: determining, in response to determining that a video frame image including a target lesion is detected from a video stream, location information of the target lesion on the video frame image; tracking the target lesion according to the location information of the target lesion on the video frame image, and determining location information of the target lesion on a video frame image in the video stream.