Medical image dectection system and method
Inventors
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-9947090-B2
Publication Date
2018-04-17
Expiration Date
Abstract
The present invention involves a foreign object detection system and method which involves combining several algorithms specifically enhanced for foreign object detection to provide an improved detection system.
Core Innovation
The problem being addressed is detecting specific foreign objects in an image scan, such as retained surgical foreign objects including sponges and needles, within portable X-ray and OR X-ray images. The document frames the goal as detecting possible foreign object locations and reporting any detected foreign objects from the image scan.
The core workflow classifies an image, detects possible foreign object locations based upon the classifying step, and enhances at least one of the possible foreign object locations by creating at least one additional enhanced image. The document then classifies and clusters detected possible foreign object locations, analyzes clusters for presence of foreign objects, and reports any detected foreign object.
To support learning with real RSI incidence, the document describes using real RSI images plus training data augmentation by superimposing segmented foreign-object X-rays onto patient images, with intensity normalization and random warping and blending. It further includes image enhancement and candidate detection steps, producing candidate boxes and feature extraction, spatial clustering into candidate-location clusters, and supervised classification with classifiers selected based on image type such as anatomical region, exposure level, and clutter level.
Claims Coverage
The document includes one independent claim that covers a full method pipeline with two inventive features: image classification and candidate foreign-object localization with enhancement, and classifying and clustering detected candidate locations followed by cluster analysis and reporting.
Classifying an image to detect possible foreign object locations
The method classifies an image, detects possible foreign object locations based upon the classifying step, and enhances at least one of the possible foreign object locations by creating at least one additional enhanced image.
Classifying and clustering detected possible foreign object locations
The method classifies and clusters detected possible foreign object locations, analyzes clusters for presence of foreign objects, and reports any detected foreign object.
Overall, the claim coverage focuses on a pipeline that combines image classification with candidate foreign-object localization, candidate enhancement, and then classification with clustering followed by cluster-level analysis and reporting.
Stated Advantages
Not explicitly described in patent.
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?