Visual Filter Neural Network for Anatomical Image Pre-processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Manual visual assessment of medical anatomical images, such as x-ray images, is time-consuming and prone to missing critical findings due to the large amount of information that needs to be processed, leading to potential delays in urgent treatments.
Innovation Solution
A system utilizing a visual filter neural network to prioritize anatomical images by classifying them into relevant and irrelevant categories, adjusting outlier pixel intensity values, and rotating images to a baseline orientation, thereby enhancing the accuracy of a classification neural network in detecting acute medical conditions for early and rapid treatment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual visual assessment is used to review anatomical images, then comprehensive analysis of all images is performed, but the time required increases significantly and critical findings may be missed
Solution Approach 1:
The system segments the large set of anatomical images by using a visual filter neural network to classify images into relevant and irrelevant categories based on target body region and sensor orientation. This segmentation allows the classification neural network to focus only on relevant images, significantly reducing processing time while maintaining detection accuracy for critical findings.
Solution Approach 2:
The system performs preliminary filtering of anatomical images before they reach the classification neural network. The visual filter neural network pre-processes images by rejecting irrelevant ones (e.g., non-target body regions or incorrect sensor orientations), so that only potentially relevant images require detailed analysis, thereby reducing overall processing time without compromising detection accuracy.
2Measurement precision
If all anatomical images are processed by the classification neural network, then detection accuracy is maximized, but processing time and computational resources increase
Solution Approach 1:
The system divides the image processing workload into two stages: first, the visual filter neural network segments images into relevant and irrelevant groups; second, only relevant images are processed by the classification neural network. This segmentation maintains high detection accuracy for acute medical conditions while significantly improving processing throughput by excluding irrelevant images from resource-intensive classification.
Solution Approach 2:
The system applies partial processing by using the visual filter neural network to quickly assess and reject irrelevant images before applying the more computationally intensive classification neural network. This partial action approach processes only the necessary subset of images (those potentially containing target body regions with correct sensor orientation), thereby maintaining detection accuracy while improving overall processing throughput.
3Measurement precision
If images with outlier pixel intensity values are not adjusted, then processing is faster, but detection accuracy of fine features deteriorates
Solution Approach 1:
The system applies local quality adjustment by specifically targeting and adjusting only the outlier pixel intensity values that represent injected content, while leaving the rest of the image data unchanged. This localized adjustment improves detection accuracy of fine features in relevant images without requiring comprehensive reprocessing of entire images, thereby managing preprocessing complexity efficiently.
4Measurement precision
If images are not rotated to baseline orientation, then processing is simpler and faster, but classification accuracy decreases
Solution Approach 1:
The system performs preliminary rotation of relevant images to baseline orientation before they are classified by the visual filter neural network. This preliminary action ensures that images are in the correct orientation for accurate classification of target body regions and sensor orientations, thereby improving classification accuracy while managing preprocessing complexity through automated orientation correction.
Data Source
AI summary
A system for prioritizing patients for treatment, comprising: at least one hardware processor executing a code for: feeding anatomical images into a visual filter neural network for outputting a category indicative of a target body region depicted at a target sensor orientation and a rotation relative to a baseline, rejecting a sub-set of anatomical images classified into another category, rotating to the baseline images classified as rotated, identifying pixels for each image having outlier pixel intensity values denoting an injection of content, adjusting the outlier pixel intensity values to values computed as a function of non-outlier pixel intensity values, feeding each the remaining sub-set of images with adjusted outlier pixel intensity values into a classification neural network for detecting the visual finding type, generating instructions for creating a triage list for which the classification neural network detected the indication, wherein patients are selected for treatment based on the triage list.