Gap Detection in Bone Images Using Bounding Box Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image segmentation methods are inadequate for accurately separating adjacent objects in images, especially when gaps between objects are small, due to noise interference and the need for combinations of methods that still fall short in precision.

Innovation Solution

A computer-implemented method and system that employs thresholding, edge detection, and region growing to identify gaps between objects by analyzing binarized images, using techniques such as histogram equalization, noise filtering, and erode kernels to enhance segmentation accuracy, particularly in identifying fractures or lesions in bone images.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If existing segmentation methods (thresholding, clustering, edge detection) are used to separate objects in an image, then segmentation can be performed, but segmentation accuracy deteriorates when gaps between objects are very small or when noise is present

Engineering Contradiction:
Improvesegmentation accuracyVSAvoiddifficulty of separating adjacent objects
Core Design Contradiction:
Measurement precisionVSDifficulty of detecting and measuring

Solution Approach 1:

The patent applies segmentation by dividing the image processing task into distinct stages: initial object detection, bounding box generation, gap region identification, and final separation. This multi-stage segmentation approach allows each stage to focus on specific characteristics, improving overall accuracy for separating adjacent objects with small gaps

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary bounding box structure that encompasses both adjacent objects and the gap between them. This bounding box serves as a mediator that isolates the region of interest, allowing the algorithm to focus computational resources on identifying and measuring the gap region specifically, rather than attempting to separate objects in the entire image

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If multiple segmentation methods are combined to improve separation accuracy, then segmentation can be performed on more complex images, but the complexity of the system increases and accuracy still becomes insufficient for very small gaps

Engineering Contradiction:
Improvesegmentation accuracyVSAvoidcomplexity of segmentation system
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the segmentation problem into distinct functional modules: object detection module, bounding box generation module, gap region identification module, and measurement module. This modular segmentation reduces system complexity by allowing each module to perform a specific function with optimized algorithms, rather than using a single complex method

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary actions by first detecting objects and generating bounding boxes before attempting to identify and measure gaps. This preliminary structuring of the image data creates a standardized format that simplifies subsequent gap detection, reducing the complexity of the overall system

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3345159B1Method and apparatus for identifying a gap between objects in an image
Publication Date: 2022.08.03 STRAXCORP PTY LTD
  • EP3345159B1 patent drawingFigure 1~2
  • EP3345159B1 patent drawingFigure 3~4
  • EP3345159B1 patent drawingFigure 5A

AI summary

A computer-implemented method for identifying a gap between objects in an image, the method comprising: detecting contours of the objects in the image or in a binarized version thereof; locating at least one path or gap between the objects, in the image or in an intersection region being a portion of the image containing at least respective portions of the objects by: determining representative attenuations or densities of respective one or more rows and columns of the image or of the intersection region, identifying first pixels, being pixels in a column that has a representative attenuation or density of 0, and second pixels, being pixels in a row that has a representative attenuation or density of 0, determining whether there exists a path or gap between two or more boundaries by identifying at least one first or second pixel on each of two or more of the boundaries, and detecting one or more paths between the at least two boundaries; and outputting a result indicative of one or more detected paths.