CAD Algorithm Detecting Outlying Tissue Abnormalities
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing CAD systems for medical imaging often ignore outlying tissue near the image border, missing potentially valuable information about anatomical abnormalities that could indicate the need for further medical evaluation or intervention.
Innovation Solution
A computer-aided detection (CAD) algorithm that processes medical images to detect likely abnormalities in outlying tissue near the image border by computing neighborhood-based features in an extended region beyond the image border, and provides annotated maps to highlight these suspected abnormalities for radiologists.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If CAD algorithms analyze only the imaged tissue within the image border, then the processing is simple and fast, but anatomical abnormalities in outlying tissue are missed
Solution Approach 1:
The patent extends the analysis from the traditional 2D image border into a virtual 3D extended region. By computing neighborhood-based features that spill over the image border into outlying tissue, the system adds a spatial dimension beyond the conventional imaging boundaries, enabling detection of abnormalities in tissue that was previously inaccessible to CAD analysis.
Solution Approach 2:
The system performs preliminary computation of neighborhood-based features in the extended region before final abnormality detection. By pre-calculating features in outlying tissue and preparing annotation maps in advance, the system enables comprehensive screening without significantly increasing real-time processing complexity during the actual diagnosis workflow.
2Area of stationary object
If the CAD system extends analysis beyond the image border, then anatomical coverage is improved, but computational resources and processing time increase
Solution Approach 1:
Rather than uniformly analyzing the entire extended region, the system applies local quality by computing neighborhood-based features only in specific regions where abnormalities are most likely to occur near the image border. This targeted approach extends anatomical coverage while minimizing unnecessary computational resources spent on analyzing large portions of outlying tissue.
Solution Approach 2:
The patent implements partial action by analyzing only the portion of outlying tissue that is within a predetermined distance from the image border, rather than the entire outlying region. This partial analysis of the most critical areas provides sufficient anatomical coverage for detecting abnormalities while keeping computational resource consumption manageable.
3Measurement precision
If neighborhood-based features are computed in the extended region, then detection precision for outlying abnormalities improves, but false positives may increase
Solution Approach 1:
The system incorporates feedback mechanisms where the classified and labeled candidate locations in the extended region are fed back into the annotation map generation process. This feedback loop allows the system to refine its detections by comparing neighborhood-based feature results with the overall image context, thereby maintaining high detection precision while filtering out false positives through iterative validation.
Data Source
AI summary
A method, system, and related computer program products are provided for processing a medical image of a body part according to a computer-aided detection (CAD) algorithm, the medical image having an image border, the body part comprising imaged tissue appearing inside the image border and outlying tissue not appearing in the medical image, wherein likely anatomical abnormalities in the outlying tissue near the imaged tissue border are detected by the CAD algorithm. In one example, the detected likely abnormalities in the outlying tissue are located within a first distance from the imaged tissue border, wherein the first distance corresponds to a spatial ambit of a neighborhood-based feature computed by the CAD algorithm.