Endoscopic Observation Canvas for Examined Area Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current endoscopic systems rely heavily on the skill and experience of the operator for thorough examination and accurate recording of observed areas, and the application of artificial intelligence is hindered by the scarcity and quality of training data, particularly in procedures like cystoscopy, leading to inconsistent diagnostic accuracy.
Innovation Solution
An endoscopic diagnosis support method using a simulated unfolded observation canvas to record examined and unexamined areas, combined with an annotation augmented model to enhance diagnostic accuracy without requiring additional training data, by utilizing key frames, key points, and image feature points, and employing an annotation augmented model to generate useful annotation information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If artificial intelligence is applied to endoscopic diagnosis, then diagnostic accuracy can be improved, but the scarcity and poor quality of training data (especially in cystoscopy) prevents effective AI implementation
Solution Approach 1:
The patent creates a simulated three-dimensional space that copies and reconstructs the actual endoscopic examination space. By mapping endoscopic images to a virtual 3D model of the organ cavity, the system generates synthetic training data that replicates real examination conditions without requiring additional physical examinations or patient data
Solution Approach 2:
The system performs preliminary creation of a simulated 3D space and pre-marking of examination positions before actual diagnosis. By establishing the virtual examination environment in advance with predefined anatomical structures and examination pathways, the system prepares training data structures ahead of time, eliminating the need to collect and process large volumes of raw medical images
2Reliability
If a doctor manually records examination results, then diagnostic support can be provided, but the accuracy varies depending on the doctor's skill and experience
Solution Approach 1:
The system enables automatic self-recording of examination results by the simulated 3D space. As the endoscope moves through the organ cavity, the system automatically tracks position, maps observed areas to the virtual model, and records examination status without requiring manual intervention. The examination canvas is automatically updated based on the endoscopic image sequence and position data
Solution Approach 2:
The patent replaces the mechanical manual recording process with an automated computer-based system. Instead of doctors manually noting observations, the system uses image processing, position tracking, and automatic mapping to the 3D model to record examination data objectively and consistently
3Measurement precision
If all areas of the organ cavity are thoroughly examined, then lesion detection accuracy improves, but it is difficult to determine whether all required areas have been observed
Solution Approach 1:
The patent transitions from two-dimensional endoscopic images to a three-dimensional virtual space representation. By unfolding the 3D organ cavity into a 2D examination canvas with spatial coordinates, the system provides a comprehensive overview of all examined areas. The 3D position information allows the system to track and visualize which regions have been observed and which remain unexamined
Data Source
AI summary
An endoscopic diagnosis support method whereby an examined area and an unexamined area can be clearly discriminated. After a preparatory step of an observation canvas is performed in advance, a frame marking step, a key point calculation step, a preceding and following frame displacement amount calculation step, a preceding and following frame marking step are executed to thereby perform observation recording. In an image diagnosis support step IDS, support is performed such that the existence of a lesion is diagnosed in an organ on the basis of a plurality of position data marked with respect to a plurality of frames in the observation canvas data and an endoscopic image in the plural frames.