Automated method and system for computerized detection of masses and parenchymal distortions in medical images
AssigneeArch Development Corporation
InventorsMaryellen L. Giger, Ming Zhang, Ping Lu
Life of this patent
Estimating the term…
Abstract
A method and system for the automated detection of lesions such as masses and/or tissue (parenchymal) distortions in medical images such as mammograms. Dense regions and subcutaneous fat regions within a mammogram are segmented. A background correction may be performed within the dense regions. Hough spectrum within ROIs placed in the breast region of a mammogram are calculated and thresholded using the intensity value .eta. in order to increase sensitivity and reduce the number of false-positive detections. Lesions are detected based on the thresholded Hough spectra. The thresholded Hough spectra are also used to differentiate between benign and malignant masses.
Claims44
A method for enhancing visualization of a mammographic image, comprising:
- identifying in said image an anatomically dense portion of said image;
- and
- processing said dense portion to produce a processed image having a more uniform density.
A method of detecting a lesion in a mammographic image, comprising:
- determining a fat value of said image;
- subtracting said fat value from said image to produce a subtracted image;
- extracting features from said subtracted image;
- and
- inputting said features to a detection scheme.
A method of detecting a lesion in a mammographic image, comprising:
- obtaining right and left mammograms;
- identifying in at least one of said right and left mammograms an anatomically dense portion;
- processing said dense portion identified in said identifying step so that the mammogram in which the dense portion is identified has a more uniform density;
- processing said right and left mammograms after said step of processing said dense portion to obtain a combined image;
- performing feature analysis on said combined image to extract features from said combined image;
- and
- detecting a lesion based on the extracted features.
A method of analyzing a mammographic image, comprising:
- selecting a plurality of regions of interest in said image;
- calculating a Hough spectrum for each of said regions of interest;
- determining a threshold using said Hough spectra;
- thresholding said Hough spectra using said threshold;
- identifying a region suspected of being a lesion in said image;
- and
- analyzing said region.
A method of analyzing an image, comprising:
- locating a region suspected of containing a lesion;
- selecting a plurality of regions of interest including said region;
- calculating a Hough spectrum for each of said regions of interest;
- determining a threshold using said Hough spectra;
- thresholding said Hough spectra using said threshold to produce a thresholded Hough spectra;
- determining an accumulation for each of said thresholded Hough spectra;
- extracting at least one feature of said region using a feature analysis technique;
- and
- analyzing said region using said accumulation and said at least one feature.
A system for visualization in an image, comprising:
- an image acquisition device configured to acquire said image;
- and
- a dense portion processing circuit connected to said image acquisition device and configured to identify an anatomically dense portion of said image and process said dense portion to produce a processed image having more uniform density.
A system for detecting lesions in an image;
- an image acquisition device;
- a lesion location circuit connected to said image acquisition device;
- a dense portion processing circuit connected to said lesion location circuit;
- a Hough spectrum analysis circuit connected to said lesion location circuit and said dense portion processing circuit;
- a lesion extraction circuit connected to said Hough spectrum analysis circuit;
- a feature extraction circuit connected to said lesion extraction circuit;
- and
- a lesion detection circuit connected to said feature analysis circuit.
Dependencies are inferred from the claim text.