Search

PubMed ID: 17854795 A novel fully automated system is introduced to facilitate lesion detection in dynamic contrast-enhanced, magnetic resonance mammography (DCE-MRM). The system extracts breast regions from pre-contrast images using a cellular neural network, generates normalized maximum intensity-time ratio (nMITR) maps and performs 3D template matching with three layers of 12 × 12 cells to detect lesions. A breast is considered to be properly segmented when relative overlap > 0.85 and misclassification rate < 0.10. Sensitivity, false-positive rate per slice and per lesion are used to assess detection performance. The system was tested with a dataset of 2064 breast MR images (344 slices × 6 acquisitions over time) from 19 women containing 39 marked lesions. Ninety-seven percent of the breasts were segmented properly and all the lesions were detected correctly (detection sensitivity = 100 %), however, there were some false-positive detections (31%/lesion, 10%/slice). © 2007 Elsevier Ltd. All rights reserved.

Cellular neural networks(CNNs) are massively parallel cellular structures with learning abilities. They can be used to realize complex image processing applications efficiently and in almost real time. In this preliminary study, we propose a novel, robust, and fully automated system based on CNNs to facilitate lesion localization in contrast-enhanced MRmammography, a difficult task requiring the processing of a large number of images with attention paid to minute details. The data set consists of 1170 slices containing one precontrast and five postcontrast bilateral axial MRmammograms from 39 patients with 37 malignant and 39 benign mass lesions acquired using a 1.5 Tesla MRscanner with the following parameters: 3D FLASH sequence, TR/TE 9.80/4.76 ms, flip angle 25 ° , slice thickness 2.5 mm, and 0.625 × 0.625 mm 2 in-plane resolution. Six hundred slices with 21 benign and 25 malignant lesions of this set are used for training the CNNs; the remaining data are used for test purposes. The breast region of interest is first segmented from precontrast images using four 2D CNNs connected in cascade, specially designed to minimize false detections due to muscles, heart, lungs, and thoracic cavity. To identify deceptively enhancing regions, a 3D nMITR map of the segmented breast is computed and converted into binary form. During this process tissues that have low degrees of enhancements are discarded. To boost lesions, this binary image is processed by a 3D CNN with a control template consisting of three layers of 11 × 11 cells and a fuzzy c -partitioning output function. A set of decision rules extracted empirically from the training data set based on volume and 3D eccentricity features is used to make final decisions and localize lesions. The segmentation algorithm performs well with high average precision, high true positive volume fraction, and low false positive volume fraction with an overall performance of 0.93 ± 0.05 , 0.96 ± 0.04 , and 0.03 ± 0.05 , respectively (training: 0.93 ± 0.04 , 0.94 ± 0.04 , and 0.02 ± 0.03 ; test: 0.93 ± 0.05 , 0.97 ± 0.03 , and 0.05 ± 0.06 ). The lesion detection performance of the system is quite satisfactory; for the training data set the maximum detection sensitivity is 100 % with false-positive detections of 0.28/lesion, 0.09/slice, and 0.65/case; for the test data set the maximum detection sensitivity is 97 % with false-positive detections of 0.43/lesion, 0.11/slice, and 0.68/case. On the average, for a detection sensitivity of 99 % , the overall performance of the system is 0.34/lesion, 0.10/slice, and 0.67/case. The system introduced does not require prior information concerning breast anatomy; it is robust and exceptionally effective for detecting breast lesions. The use of CNNs, fuzzy c -partitioning, volume, and 3D eccentricity criteria reduces false-positive detections due to artifacts caused by highly enhanced blood vessels, nipples, and normal parenchyma and artifacts from vascularized tissues in the chest wall due to oversegmentation. We hope that this system will facilitate breast examinations, improve the localization of lesions, and reduce unnecessary mastectomies, especially due to missed multicentric lesions and that almost real-time processing speeds achievable by direct hardware implementations will open up new clinical applications, such as making feasible quasi-automated MR-guided biopsies and acquisition of additional postcontrast lesion images to improve morphological characterizations.

Effectiveness of morphological descriptors based on normalized maximum intensity-time ratio (nMITR) maps generated using a 3 x 3 pixel moving mask on dynamic contrast-enhanced magnetoresistance (MR) mammograms are studied for assessment of malignancy. After a rough indication of volume of interest on the nMITR maps, lesions are automatically segmented. Two-dimensional (2D) convexity, normalized complexity, extent, and eccentricity as well as three-dimensional (3D) versions of these descriptors and contact surface area ratio are computed. On a data set consisting of dynamic contrast-enhanced MR DCE-MR mammograms from 51 women that contain 26 benign and 32 malignant lesions, 3D convexity, complexity, and extent are found to reflect aggressiveness of malignancy better than 2D descriptors. Contact surface area ratio which is easily adaptable to different imaging resolutions is found to be the most significant and accurate descriptor (75% sensitivity, 88% specificity, 89% positive predictive values, and 74% negative predictive values).

Cellular neural networks (CNNs) are massively parallel cellular structures with learning abilities. They can be used to realize complex image processing applications efficiently and in almost real time. In this preliminary study, we propose a novel, robust, and fully automated system based on CNNs to facilitate lesion localization in contrast-enhanced MR mammography, a difficult task requiring the processing of a large number of images with attention paid to minute details. The data set consists of 1170 slices containing one precontrast and five postcontrast bilateral axial MR mammograms from 39 patients with 37 malignant and 39 benign mass lesions acquired using a 1.5 Tesla MR scanner with the following parameters: 3D FLASH sequence, TR/TE 9.80/4.76 ms, flip angle 250, slice thickness 2.5 mm, and 0.625 x 0.625 mm2 in-plane resolution. Six hundred slices with 21 benign and 25 malignant lesions of this set are used for training the CNNs; the remaining data are used for test purposes. The breast region of interest is first segmented from precontrast images using four 2D CNNs connected in cascade, specially designed to minimize false detections due to muscles, heart, lungs, and thoracic cavity. To identify deceptively enhancing regions, a 3D nMITR map of the segmented breast is computed and converted into binary form. During this process tissues that have low degrees of enhancements are discarded. To boost lesions, this binary image is processed by a 3D CNN with a control template consisting of three layers of 11 x 11 cells and a fuzzy c-partitioning output function. A set of decision rules extracted empirically from the training data set based on volume and 3D eccentricity features is used to make final decisions and localize lesions. The segmentation algorithm performs well with high average precision, high true positive volume fraction, and low false positive volume fraction with an overall performance of 0.93 +/- 0.05, 0.96 +/- 0.04, and 0.03 +/- 0.05, respectively (training: 0.93 +/- 0.04, 0.94 +/- 0.04, and 0.02 +/- 0.03; test: 0.93 +/- 0.05, 0.97 +/- 0.03, and 0.05 +/- 0.06). The lesion detection performance of the system is quite satisfactory; for the training data set the maximum detection sensitivity is 100% with false-positive detections of 0.28/lesion, 0.09/slice, and 0.65/case; for the test data set the maximum detection sensitivity is 97% with false-positive detections of 0.43/lesion, 0.11/slice, and 0.68/case. On the average, for a detection sensitivity of 99%, the overall performance of the system is 0.34/lesion, 0.10/slice, and 0.67/case. The system introduced does not require prior information concerning breast anatomy; it is robust and exceptionally effective for detecting breast lesions. The use of CNNs, fuzzy c-partitioning, volume, and 3D eccentricity criteria reduces false-positive detections due to artifacts caused by highly enhanced blood vessels, nipples, and normal parenchyma and artifacts from vascularized tissues in the chest wall due to oversegmentation. We hope that this system will facilitate breast examinations, improve the localization of lesions, and reduce unnecessary mastectomies, especially due to missed multicentric lesions and that almost real-time processing speeds achievable by direct hardware implementations will open up new clinical applications, such as making feasible quasi-automated MR-guided biopsies and acquisition of additional postcontrast lesion images to improve morphological characterizations.

A fully automated software is introduced to facilitate MR mammography (MRM) examinations and overcome subjectiveness in diagnosis using normalized maximum intensity-time ratio (nMITR) maps. These maps inherently suppress enhancements due to normal parenchyma and blood vessels that surround lesions and have natural tolerance to small field inhomogeneities and motion artifacts. The classifier embedded within the software is trained with normalized complexity and maximum nMITR of 22 lesions and tested with the features of remaining 22 lesions. Achieved diagnostic performances are 92% sensitivity, 90% specificity, 91% accuracy, 92% positive predictive value and 90% negative predictive value. DynaMammoAnalyst shortens evaluation time considerably and reduces inter and intra-observer variability by providing decision support.