Method and system to detect the microcalcifications in X-ray images using nonlinear energy operator

摘要

A method and system to detect the microcalcifications (MC) in different type of images viz. X-ray images/mammograms/computer tomography with varied densities using nonlinear energy operator (NEO) is disclosed to favor precise detection of early breast cancer. Such Microcalcifications are associated with both high intensity and high frequency content. The same NEO output is useful to detect and remove the irrelevant curvilinear structures (CLS) thereby helps in reducing the false alarms in micro calcification detection technique. This is effective on different dataset (scanned film, mammograms with large spatial resolution such as CR and DR) of varied breast composition (viz. dense, fatty glandular, fatty), demonstrated quantitatively by Free-response receiver operating characteristic (FROC). Importantly, the method and apparatus of the invention can be used in conjunction with machine learning techniques viz. SVM to favor detection of incipient or small microcalcifications, thus benefiting radiologists in confirming detection of micro-calcifications in X-rays images/mammograms and reducing death rates.

主权项

1. A method to detect microcalcifications in images or mammograms comprising: a pre-processing step comprising:
(i) eliminating all extraneous and non-breast artifacts including human introduced labels, radiopaque artifacts; and (ii) down sampling the image or mammograms to reduce computational complexity; a microcalcification segmentation step involving Non Linear Energy Operator (NEO) output based microcalcification segmentation comprising: (i) generating a smoothed NEO (SNEO) for enhancing microcalcification spikes; and (ii) selecting an optimal threshold for detecting suspicious regions containing microcalcification; and a false positive reduction step comprising removing curvilinear structure (CLS) texture appearances of the images or mammograms by reusing the NEO output to enhance computational speed and the quality of microcalcification detection, wherein the false positive reduction step comprises introducing a parameter φ to detect CLS elements, wherein:∅=Area⁢⁢of⁢⁢individual⁢⁢element⁢⁢in⁢⁢reduced⁢⁢thresholdArea⁢⁢of⁢⁢a⁢⁢individual⁢⁢element⁢⁢in⁢⁢full⁢⁢thresholdinvolving the same SNEO in the microcalcification segmentation step by reducing average thereshold Tavg which is average of local optimal threshold corresponding to each of the SNEO/NEO operated direction output to p% of the average threshold (where p<100), calculating the φ of each element, and if the φ is lower than certain threshold (say Tφ), object is likely to be calcification as its compactness is high and elements having higher φ are treated as the CLS elements and thus discarded.