Object Detection Device Using Deformation Displacement Field for Polyp Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current large intestine CT examination methods rely heavily on human expertise, leading to variable detection accuracy and a risk of false detection of lesions, particularly in images taken from different postures, where water droplets or residues may be misidentified as polyps.
Innovation Solution
An object detection device and method that compares and aligns candidate regions from images taken in different postures using a deformation displacement field, transforming coordinates and determining the presence of a polyp through a combination of convolution neural networks and scoring systems to reduce false positives and omissions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If a discriminator trained through machine learning is used to extract lesion regions from CT images, then automated detection is achieved, but false detection occurs due to reliance on single-image analysis and human expertise variability
Solution Approach 1:
The patent transitions from analyzing a single CT image to analyzing multiple CT images taken in different postures (supine and prone positions). By adding the dimension of multiple imaging angles, the system can distinguish true lesions from artifacts like water droplets that appear in only one posture, thereby improving detection reliability while maintaining automation.
Solution Approach 2:
The patent combines multiple candidate region detection results from different postures by transforming coordinates to a common reference frame and merging the results. This integration of information from multiple sources allows the system to confirm true lesions that appear consistently across postures while eliminating false positives that appear in only one posture.
2Reliability
If images are taken in two different postures to reduce false detection, then detection reliability improves, but the complexity of the detection system increases due to multiple image processing requirements
Solution Approach 1:
The patent introduces a coordinate transformation unit as an intermediary that converts candidate region coordinates from different posture images into a common reference coordinate system. This mediator enables the merging of detection results without requiring complex manual alignment procedures, thus improving reliability while managing system complexity through automated transformation processes.
3Ease of operation
If candidate regions are individually detected in each image without coordination, then processing simplicity is maintained, but false detection increases due to inability to distinguish true lesions from artifacts
Solution Approach 1:
The patent implements a feedback mechanism where detection results from one posture are used to verify and refine detections in another posture. By comparing candidate regions across multiple postures through coordinate transformation and merging, the system provides feedback that confirms true lesions and eliminates false positives, thereby improving reliability while maintaining operational simplicity through automated verification.
Data Source
AI summary
An object detection device that detects a specific object included in an input image includes a first candidate region specifying unit that specifies a first candidate region in which an object candidate is included from a first input image obtained by imaging a subject in a first posture, a second candidate region specifying unit that specifies a second candidate region in which an object candidate is included from a second input image obtained by imaging the subject in a second posture different from the first posture, a deformation displacement field generation unit that generates a deformation displacement field between the first input image and the second input image, a coordinate transformation unit that transforms a coordinate of the second candidate region to a coordinate of the first posture based on the deformation displacement field, an association unit that associates the first candidate region with the transformed second candidate region that is close to the first candidate region, and a same object determination unit that determines that the object candidates included in the candidate regions associated with each other by the association unit are the same object and are the specific object.