Colonoscopy Lesion Detection Using Condition-Specific AI Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing endoscopic examination systems struggle to accurately detect lesions in the large intestine, particularly when inflammation is present, as they do not adequately consider the varying conditions of the intestine.
Innovation Solution
An endoscopic examination support apparatus equipped with machine learning models to detect lesions in both inflammation and non-inflammation states of the large intestine, allowing for switching between different lesion detectors based on the intestine's condition, and outputting detection results accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single machine learning model is used for lesion detection, then the device complexity is reduced, but the measurement precision deteriorates when inflammation is present in the large intestine
Solution Approach 1:
The detection system is segmented into multiple specialized machine learning models: a first model trained on normal large intestine images and a second model trained on inflamed large intestine images. Each model is responsible for detecting lesions in specific intestinal conditions, thereby improving detection accuracy for each segment while managing overall system complexity through modular architecture.
Solution Approach 2:
Different detection models are applied based on the local condition of the large intestine. The system dynamically selects or switches between the first model (for normal conditions) and the second model (for inflamed conditions) according to the specific intestinal state being examined, ensuring optimal detection quality for each local condition.
2Adaptability or versatility
If multiple machine learning models are used for different intestinal conditions, then the adaptability is improved, but the device complexity increases
Solution Approach 1:
The system implements dynamic model selection and switching capability, allowing it to adaptively choose between the first and second machine learning models based on the real-time condition of the large intestine. This dynamic adaptation enables the system to handle varying intestinal conditions (normal, inflamed, or mixed states) while maintaining a manageable complexity level through automated selection logic.
Solution Approach 2:
The detection system achieves multi-functionality by incorporating multiple machine learning models that can handle different intestinal conditions within a single unified system. The system can detect lesions in normal intestines, inflamed intestines, or switch between modes based on the detected condition, providing universal applicability across various examination scenarios.
3Measurement precision
If lesion detection does not consider the intestinal state, then the ease of operation is improved, but the measurement precision deteriorates
Solution Approach 1:
The system performs self-service by automatically detecting the intestinal condition and selecting the appropriate machine learning model without requiring manual intervention from the operator. The automated condition assessment and model selection process maintains high detection accuracy while preserving ease of operation, as the system handles the complexity internally without burdening the user.
Solution Approach 2:
The system incorporates feedback mechanisms where the detection process continuously monitors intestinal conditions and adjusts the selected model accordingly. This feedback loop ensures that the most appropriate detection algorithm is applied based on the current intestinal state, maintaining high precision while keeping the operation simple through automated adjustments.
Data Source
AI summary
In the endoscopic examination support apparatus, the image acquisition means acquires an endoscopic image taken by an endoscope. The first lesion detection means detects a lesion candidate from the endoscopic image, using a machine learning model that learned a relationship between the lesion candidate and a normal state of a large intestine. The second lesion detection means detects a lesion candidate from the endoscopic image, using a machine learning model that learned a relationship between the lesion candidate and a predetermined state of the large intestine. The output means outputs at least one of a detection result of the first lesion detection means and a detection result of the second lesion detection means.