Endoscope Lumen Detection Using Image and Time-Series Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems struggle to accurately and reliably identify the existence of lumens in medical images without omission, particularly in cases where the lumen is partially obscured or of poor image quality.
Innovation Solution
A medical support device utilizing a trained model to generate lumen specification information, which includes first and second lumen specification information based on input from a trained model and a time-series model, respectively, to accurately specify the lumen's existence position in sequential medical images, with conditions to handle erroneous specifications and varying image qualities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If a trained model is used to generate lumen specification information from medical images, then the automation and speed of lumen identification are improved, but the reliability deteriorates when the lumen is partially obscured or image quality is poor
Solution Approach 1:
The patent combines a trained model (for automated identification) with a time-series model (for temporal consistency verification) to process medical images. The trained model generates initial lumen specification information, while the time-series model verifies reliability by analyzing temporal patterns across multiple images. This merging allows the system to maintain high automation while improving reliability through cross-validation.
Solution Approach 2:
The system implements feedback by using the time-series model to evaluate the reliability of lumen specification information generated by the trained model. When reliability is low (e.g., obscured lumen or poor image quality), the system adjusts its behavior by relying more on temporal patterns from previous images or by flagging uncertain results for manual review. This feedback loop ensures that automation does not compromise reliability.
2Productivity
If a trained model processes each medical image independently, then the processing speed is improved, but the accuracy deteriorates due to inability to utilize temporal information from sequential images
Solution Approach 1:
The system performs preliminary processing by the trained model on each individual medical image to generate initial lumen specification information quickly. This preliminary action maintains high processing speed. Subsequently, the time-series model performs temporal analysis across multiple images to refine accuracy by leveraging motion patterns and temporal consistency. This two-stage approach preserves speed while improving precision.
Solution Approach 2:
The patent transitions from processing images in isolation (single dimension) to processing images in temporal sequence (adding time dimension). The time-series model analyzes lumen specification information across multiple time points, utilizing temporal patterns and motion trajectories to improve accuracy. This dimensional expansion allows the system to maintain fast per-image processing while achieving higher overall accuracy through temporal context.
3Loss of information
If the system outputs lumen specification information for all medical images, then the completeness of information is improved, but the reliability deteriorates due to inclusion of erroneous specifications from low-quality images
Solution Approach 1:
The system applies different quality standards and processing strategies to different medical images based on their individual characteristics. For high-quality images, the trained model's output is accepted with high confidence. For low-quality images (obscured lumen, poor contrast), the system either relies more heavily on temporal patterns from the time-series model or flags these specific cases for manual review. This localized quality-based approach ensures completeness while maintaining reliability.
Solution Approach 2:
The system dynamically adjusts confidence thresholds and weighting parameters based on image quality metrics. When image quality is poor, the system increases the weight of temporal pattern matching from the time-series model and decreases reliance on the trained model's direct output. This parameter adjustment allows the system to maintain complete information coverage while filtering out erroneous specifications through adaptive reliability assessment.
Data Source
AI summary
A processor of a medical support device is configured to: acquire a medical image; and selectively output a plurality of pieces of lumen specification information. The medical image is classified into a first medical image and a second medical image. The plurality of pieces of lumen specification information include first lumen specification information and second lumen specification information. The first lumen specification information is generated based on information obtained from a trained model in a case in which the first medical image is input to the trained model, and is information capable of specifying a first existence position. The second lumen specification information is generated based on information obtained from a time-series model in a case in which time-series information is input to the time-series model, and is information capable of specifying a second existence position.