Differentiation Of Breast Lesions Research Articles

Discriminant analysis of neural style representations for breast lesion classification in ultrasound

Ultrasound imaging is widely used for breast lesion differentiation. In this paper we propose a neural transfer learning method for breast lesion classification in ultrasound. As reported in several papers, the content and the style of a particular image can be separated with a convolutional neural network. The style, coded by the Gram matrix, can be used to perform neural transfer of artistic style. In this paper we extract the neural style representations of malignant and benign breast lesions using the VGG19 neural network. Next, the Fisher discriminant analysis is used to separate those neural style representations and perform classification. The proposed approach achieves good classification performance (AUC of 0.847). Our method is compared with another transfer learning technique based on extracting pooling layer features (AUC of 0.826). Moreover, we apply the Fisher discriminant analysis to differentiate breast lesions using ultrasound images (AUC of 0.758). Additionally, we extract the eigenimages related to malignant and benign breast lesions and show that these eigenimages present features commonly associated with lesion type, such as contour attributes or shadowing. The proposed techniques may be useful for the researchers interested in ultrasound breast lesion characterization.

A Normalized Shear Deformation Indicator for Ultrasound Strain Elastography in Breast Tissues: An In Vivo Feasibility Study.

The shear deformation under loads contains useful information for distinguishing benign breast lesions from malignant ones. In this study, we proposed a normalized shear deformation indicator (NSDI) that was derived from the concept of principal strains. Since the NSDI requires both high-quality axial and lateral (parallel and perpendicular to the beam, resp.) displacement estimates, a strategy combining high-quality speckle tracking with signal “denoising” was employed. Both techniques were previously published by our group. Finite element (FE) models were used to identify possible causes for elevated NSDI values in and around breast lesions, followed by an analysis of ultrasound data acquired from 26 biopsy-confirmed in vivo breast lesions. We found that, theoretically, the elevated NSDI values could be attributed to two factors: significantly hardened tissue stiffness and increasing heterogeneity. The analysis of in vivo data showed that the proposed NSDI values were higher (p < 0.05) among malignant cancers as compared to those measured from benign ones. In conclusion, our preliminary results demonstrated that the calculation of NSDI value is feasible and NSDI could add value to breast lesion differentiation with current clinical equipment as a postprocessing tool.

A multicenter study to confirm further diagnostic value of strain ratio method giving to 5-point scoring by elastography for differentiation of breast lesions

Objective To confirm whether strain ratio should be added after evaluation of lesions with 5-point elasticity scoring for differentiating breast lesions on ultrasonic elastography (UE). Methods Eight hundred and forty-six consecutive female patients with 1 071 breast lesions were recruited into a multicenter retrospective study, which involved 8 centers across China. All the patients underwent the UE procedure and the strain ratios were calculated. The sensitivity, specificity, accuracy, PPV and NPV were calculated for each of the two evaluation systems and the areas under the ROC curve were compared. Results The strain ratios of 559 benign lesions (2.7±1.4) and 512 malignant lesions (8.1±5.9) were significantly different (P<0.001). When the cutoff point was 3.03, strain ratio method had 81.6% sensitivity, 89.7% specificity, 81.7% accuracy, 80.2% PPV and 84.2% NPV. The areas under the ROC curve with the strain ratio method and 5-point scoring method were 0.865 and 0.862, respectively (P=0.622). For the lesions with elasticity score 3 and 4, the areas under the ROC curve of the two methods were 0.711 and 0.629, respectively(P=0.020). Conclusions Although the two UE methods have similar diagnostic performance, separate calculation of the strain ratios seems compulsory, especially for the lesions with elasticity score 3 and 4. Key words: Elasticity imaging techniques; Breast diseases; Strain ratio; Elasticity score

Evaluation of diffusion kurtosis imaging and its combination with diffusion weighted imaging and proton MR spectroscopy in differentiation of breast lesions

Objective To evaluate the diagnostic performance of diffusion kurtosis imaging (DKI) and its combination with DWI and proton MR spectroscopy (1H-MRS) in differentiating malignancy from benign breast lesions. Methods Fifty-three patients with 38 histopathologically confirmed malignant and 15 benign breast lesions were retrospectively studied. The patients were examined by breast MRI at 3.0 T prior to operation, including conventional T1WI, fat-suppression imaging, DWI, DKI and 1H-MRS. The shape and margin of breast lesions, and their corresponding mean values for ADC, mean kurtosis (MK) and mean diffusivity (MD) were determined by two blinded radiologists in consensus. The presence or absence of choline (Cho) peak was identified using LCModel software. Independent-samples t test or χ2 test was performed for the comparison of clinical characteristics, shape and margin of lesions, and imaging parameters between malignancy and benign lesions. ROC analysis was performed to evaluate the diagnostic accuracy of DKI, DWI and 1H-MRS alone or in combination, in comparison with the histopathologic findings. Results The onset age of breast malignancy was higher than that of benign ones, and the difference has statistical significant (P 0.05). The mean ADC,MD and MK of benign lesions were (1.464±0.348) ×10-3mm2/s, (1.726±0.268) ×10-3mm2/s and (0.692±0.227) , the mean ADC,MD and MK of malignancy were (0.963±0.170) ×10-3mm2/s, (1.158±0.262) ×10-3mm2/s and (1.311±0.218) , respectively. Significant differences were obtained between benign and malignant lesions for all parameters (P<0.05) .The area under the ROC curve (AUC) of ADC, MD and MK for differentiating malignancy from benign lesions was 0.913, 0.933 and 0.968, respectively. Taken the maximum Youden's index of MK (1.110) as the ROC optimal cut-off point, MK exhibited better diagnostic sensitivity, specificity and accuracy for distinguishing malignancy from benign lesions [89.5% (34/38) ,93.3% (14/15) and 90.6% (48/53) , respectively], compared with MD and ADC. Multiparametric imaging with combination of DKI, DWI and 1H-MRS improves the diagnostic specificity (with the highest as 100.0%) but decreases the sensitivity (with the highest as 81.6% and lowest as 71.1%), compare with the single parametric imaging. Conclusions MK generated from DKI enables differentiation of breast lesions with a higher diagnostic sensitivity and specificity than DWI and 1H-MRS. DKI combined with DWI and 1H-MRS increase specificity but decrease sensitivity for breast cancer characterization. Key words: Breast neoplasms; Magnetic resonance imaging; Comparative study

Accuracy of diffusion kurtosis imaging in characterization of breast lesions.

The aim of this study was to evaluate the accuracy of diffusion kurtosis in the characterization and differentiation of breast lesions. 49 females with 53 breast lesions underwent breast MRI. The MRI magnetic field is 1.5 T, and the protocol is standard MRI sequences, dynamic sequences pre- and post-contrast agent administration and diffusion images. Diffusion kurtosis imaging (DKI) was applied as part of our standard breast MRΙ protocol. Two experienced radiologists on breast MRI, blinded to the final diagnosis, reviewed the parametric maps and placed a volume of interest on all slices including each lesion. Kurtosis [K apparent (Kapp)] and corrected apparent diffusion coefficient [D apparent (Dapp)] median values were then calculated from the whole-lesion histogram analysis. Receiver-operating characteristic analysis was used to determine the most effective cut-off values for the differentiation between benign and malignant pathologies. Histological analysis of the breast lesions was performed, and further comparative analysis of the results was performed to investigate the accuracy of the method. Benign (n = 19) and malignant lesions (n = 34) had mean diameters of 20.8 mm (10.1-31.5 mm) and 26.4 mm (10.5-42.3 mm), respectively. The lowest and the highest kurtosis values (Kapp) of malignant lesions were significantly higher than those of benign lesions. A cut-off of 0.71 provided specificity of 93.7% and sensitivity 97.1%, and the area under the curve (AUC) was 0.976 (p < 0.0001). The lowest and the highest Dapp values of malignant lesions were lower than those of benign lesions. A cut-off value of 1.57 × 10-3 mm2 s-1 provided specificity of 93.7% and sensitivity of 91.2% with AUC of 0.949 (p < 0.0001). DKI is an accurate additional tool for the characterization and differentiation of breast lesions with high Kapp and Dapp sensitivity and specificity rates. Advances in knowledge: DKI is able to distinguish benign from malignant breast pathologies. DKI increases the specificity of breast MRI.

Diagnostic features of quantitative comb-push shear elastography for breast lesion differentiation.

BackgroundLesion stiffness measured by shear wave elastography has shown to effectively separate benign from malignant breast masses. The aim of this study was to evaluate different aspects of Comb-push Ultrasound Shear Elastography (CUSE) performance in differentiating breast masses.MethodsWith written signed informed consent, this HIPAA- compliant, IRB approved prospective study included patients from April 2014 through August 2016 with breast masses identified on conventional imaging. Data from 223 patients (19–85 years, mean 59.93±14.96 years) with 227 suspicious breast masses identifiable by ultrasound (mean size 1.83±2.45cm) were analyzed. CUSE was performed on all patients. Three regions of interest (ROI), 3 mm in diameter each, were selected inside the lesion on the B-mode ultrasound which also appeared in the corresponding shear wave map. Lesion elasticity values were measured in terms of the Young’s modulus. In correlation to pathology results, statistical analyses were performed.ResultsPathology revealed 108 lesions as malignant and 115 lesions as benign. Additionally, 4 lesions (BI-RADS 2 and 3) were considered benign and were not biopsied. Average lesion stiffness measured by CUSE resulted in 84.26% sensitivity (91 of 108), 89.92% specificity (107 of 119), 85.6% positive predictive value, 89% negative predictive value and 0.91 area under the curve (P<0.0001). Stiffness maps showed spatial continuity such that maximum and average elasticity did not have significantly different results (P > 0.21).ConclusionCUSE was able to distinguish between benign and malignant breast masses with high sensitivity and specificity. Continuity of stiffness maps allowed for choosing multiple quantification ROIs which covered large areas of lesions and resulted in similar diagnostic performance based on average and maximum elasticity. The overall results of this study, highlights the clinical value of CUSE in differentiation of breast masses based on their stiffness.

Ultrasound-Guided Diffuse Optical Tomography for Differentiation of Benign and Malignant Breast Lesions: A Meta-analysis.

The purpose of this study was to assess the diagnostic performance of ultrasound-guided diffuse optical tomography for differentiation of benign and malignant breast lesions. The Cochrane Library, PubMed, and Embase databases were searched from inception to February 14, 2016. Sensitivity, specificity, and other information were extracted from the included studies. Sensitivity and specificity were pooled by a bivariate mixed-effects binary regression model. A summary receiver operating characteristic curve was constructed. Heterogeneity and publication bias were explored by Higgins and Deeks tests, respectively. Seven studies including 768 women with 886 lesions were analyzed. The summary sensitivity, specificity, and diagnostic odds ratio were 95% (95% confidence interval [CI], 85%-98%), 77% (95% CI, 66%-85%), and 57 (95% CI, 12-267), respectively. The area under the summary receiver operating characteristic curve was 91% (95% CI, 89%-94%). No significant heterogeneity or publication bias existed. Ultrasound-guided diffuse optical tomography is useful for differentiating breast lesions. Especially, its sensitivity is excellent.

Qualitative and quantitative analysis with a novel shear wave speed imaging for differential diagnosis of breast lesions

To evaluate the diagnostic performance of a new two-dimensional shear wave speed (SWS) imaging (i.e. Toshiba shear wave elastography, T-SWE) in differential diagnosis of breast lesions. 225 pathologically confirmed breast lesions in 218 patients were subject to conventional ultrasound and T-SWE examinations. The mean, standard deviation and ratio of SWS values (m/s) and elastic modulus (KPa) on T-SWE were computed. Besides, the 2D elastic images were classified into four color patterns. The area under the receiver operating characteristic (AUROC) curve analysis was performed to evaluate the diagnostic performance of T-SWE in differentiation of breast lesions. Compared with other quantitative T-SWE parameters, mean value expressed in KPa had the highest AUROC value (AUROC = 0.943), with corresponding cut-off value of 36.1 KPa, sensitivity of 85.1%, specificity of 96.6%, accuracy of 94.2%, PPV of 87.0%, and NPV of 96.1%. The AUROC of qualitative color patterns in this study obtained the best performance (AUROC = 0.957), while the differences were not significant except for that of Eratio expressed in m/s (AUROC = 0.863) (P = 0.03). In summary, qualitative color patterns of T-SWE obtained the best performance in all parameters, while mean stiffness (36.05 KPa) provided the best diagnostic performance in the quantitative parameters.

Qualitative Analysis with Virtual Touch Tissue Imaging & Quantification for Differential Diagnosis of Breast Lesions Compared with Quantitative Analysis

To evaluate the diagnostic performance of Virtual Touch Tissue Imaging ＆Quantification (VTIQ) for differential diagnosis of breast lesions, using a new qualitative analysis (i.e. the elasticity score in the travel time map) compared with quantitative analysis. From June 2014 to July 2015, 266 pathologically proven breast lesions were enrolled in this study. The maximum, mean, median, minimum, and standard deviation of shear wave speed (SWS) values (m/s) were assessed. The elasticity score, a new qualitative feature, was evaluated in the travel time map. The area under the receiver operating characteristic (AUROC) curves were plotted to evaluate the diagnostic performance of both qualitative and quantitative analyses for differentiation of breast lesions. Among all quantitative parameters, SWS-max showed the highest AUROC (0.805; 95% CI: 0.752, 0.851) compared with SWS-mean (0.786; 95% CI:0.732, 0.834; P=0.094), SWS-median (0.775; 95% CI:0.720, 0.824; P=0.046), SWS-min (0.675; 95% CI:0.615, 0.731; P=0.000), SWS-SD (0.0738; 95% CI:0.712, 0.817; P=0.074). The AUROC of qualitative analysis in this study obtained the best diagnostic performance (0.871; 95% CI: 0.825, 0.909; P=0.011), compared with the best parameter of SWS-max in quantitative analysis. The new qualitative analysis showed the superior diagnostic performance in the differentiation of breast lesions in comparison with quantitative analysis.

Real-Time Elastography for Differentiation of Breast Lesions.

SummaryBackroungTo investigate the diagnostic performance of the elastography-based strain index ratio in the differential diagnosis of malignant and benign breast lesions.Material/MethodsSeventy-nine breast masses that were classified as BI-RADS category 3, 4, and 5 on B-mode ultrasonography (US) were further prospectively evaluated by real-time sonoelastography (RTE). To obtain an optimal cut-off value of the strain ratio for differentiating between malignant and benign breast lesions, RTE findings were compared with histopathology of core needle biopsy samples or with ultrasound follow-up data of the analyzed masses.ResultsSeventy-nine breast lesions [BI-RADS category 3 (n=15), BI-RADS category 4 (n=34), and BI-RADS category 5 (n=30)] were classified as malignant (n=36) or benign (n=43). The mean strain index value was 6.59±3.44 (range 0.6–14) for malignant lesions and 2.79±2.16 (range 0.6–8.7) for benign lesions, respectively (p<0.05). As regards the detection of malignant lesions, US was characterized by sensitivity and specificity of 100% (CI 95%; 88–100) and 90% (CI 95%; 76–97), respectively. When an optimal value of the strain ratio (4.25) was obtained by ROC curve analysis, the sensitivity and specificity for diagnosing malignant lesions were 86% (CI 95%; 70–95) and 76% (CI 95%; 60–87), respectively.ConclusionsRTE can play an important role in the differentiation between malignant and benign breast masses, but it should be used in conjunction with ultrasonography.

Usefulness of combined BI-RADS analysis and Nakagami statistics of ultrasound echoes in the diagnosis of breast lesions

To develop a method combining the statistics of the ultrasound backscatter and the Breast Imaging-Reporting and Data System (BI-RADS) classification to enhance the differentiation of breast tumours. The Nakagami shape parameter m was used to characterise the scatter properties of breast tumours. Raw data from the radiofrequency (RF) echo-signal and B-mode images from 107 (32 malignant and 75 benign) lesions and their surrounding tissue were recorded. Three different characteristic values of the shape parameters of m (maximum [mLmax], minimum [mLmin] and average [mLavg]) and differences between m parameters (Δmmax, Δmmin, Δmavg) of the lesions and their surrounding tissues were assessed. A lesion with a BI-RADS score of 3 was considered benign, while a lesion with a score of 4 was considered malignant (a cut-off of BI-RADS 3/4 was set for all patients). The area under the receiver operating characteristic (ROC) curve (AUC) was equal to 0.966 for BI-RADS, with 100% sensitivity and 54.67% specificity. All malignant lesions were diagnosed correctly, whereas 34 benign lesions were biopsied unnecessarily. In assessing the Nakagami statistics, the sum of the sensitivity and specificity was the best for mLavg (62.5% and 93.33%, respectively). Only four of 20 lesions were found over the cut-off value in BI-RADS of 4a. When comparing the differences in m parameters, Δmavg had the highest sensitivity of 90% (only three of 32 lesions were false negative). These three lesions were classified as BI-RADS category 4c. The combined use of B-mode and mLmin parameter improve the AUC up to 0.978 (p=0.088), compared to BI-RADS alone. The combination of the parametric imaging and the BI-RADS assessment does not significantly improve the differentiation of breast lesions, but it has the potential to better identify the group of patients with mainly benign lesions that have a low level of suspicion for malignancy with a BI-RADS score of 4a.

Diagnostic Value of Contrast-Enhanced Sonography for Differentiation of Breast Lesions: A Meta-analysis.

The purpose of this study was to systematically review and evaluate the diagnostic accuracy of contrast-enhanced sonography in the differential diagnosis of benign and malignant breast lesions. The scientific literature databases PubMed and Embase were comprehensively searched for relevant studies before January 2015. Data were pooled to yield the summary sensitivity, specificity, and diagnostic odds ratio using meta-analysis software. A total of 29 studies with 2296 lesions were included in the analysis. The pooled sensitivity and specificity were 0.88 (95% confidence interval [CI], 0.86-0.90; inconsistency index [I(2)] = 77.9%) and 0.80 (95% CI, 0.78-0.83; I(2) = 84.0%), respectively. The pooled diagnostic odds ratio was 30.35 (95% CI, 15.75-58.48; I(2)= 82.1%), and the area under the summary receiver operating characteristic curve was 0.9115 (SE, 0.0243). The comprehensive results suggest that contrast-enhanced sonography could be a potentially effective method for differential diagnosis of benign and malignant breast lesions.

Classification of breast lesions using segmented quantitative ultrasound maps of homodyned K distribution parameters.

Statistical modeling of an ultrasound backscattered echo envelope is used for tissue characterization. However, in the presence of complex structures within the analyzed area, estimation of parameters is disturbed and unreliable, e.g., in the case of breast tumor classification. In order to improve the differentiation of breast lesions, the authors proposed a method based on the segmentation of homodyned K distribution parameter maps. Regions within lesions of different scattering properties were extracted and analyzed. In order to improve the classification, the best-performing features were selected from various regions and then combined. A radio-frequency data set consisting of 103 breast lesions was used in the authors' analysis. Maps of homodyned K distribution parameters were created using an algorithm based on signal-to-noise ratio, kurtosis, and skewness of fractional-order envelope moments. A Markov random field model was used to segment parametric maps. Features of different segments were extracted and evaluated based on bootstrapping and the receiver operating characteristic curve. To determine the best-performing feature subset, the authors applied the joint mutual information criterion. It was found that there were individual features which performed better than the ones commonly used for lesion characterization, like the parameter obtained through averaging of values over the whole lesion. The authors selected and discussed the best-performing features. Properties of different extracted regions were important and improved the distinction between benign and malignant tumors. The best performance was obtained by combining four features with the area under the receiver operating curve of 0.84. The study showed that the analysis of internal changes in lesion parametric maps leads to a better classification of breast tumors. The authors recommend combining multiple features for characterization, instead of using only one parameter, especially in the case of heterogeneous lesions.

Comparative study of superb microvascular imaging and contrast-enhanced ultrasonographic microvascular imaging in diagnosis of breast tumor

Objective To compare the diagnostic efficiency of superb microvascular imaging (SMI) and contrast-enhanced ultrasonographic microvascular imaging (MVI) for differentiating breast lesions. Methods One hundred and sixteen patients with 116 breast lesions were first examined by grayscale ultrasound. Then SMI and MVI were performed on all patients. Microvascular architectures of breast lesions were depicted by both methods. The lesions were evaluated based on their microvascular architectures. The diagnostic efficacy of both methods were compared. Results The diagnostic sensitivity, specificity, and accuracy of SMI and MVI were 79.24%, 90.48%, 85.35% and 88.68%, 87.30%, 87.93%, respectively. The areas under the curve of SMI and MVI were 0.888 and 0.926. The diagnostic values of SMI and MVI were not statistically different (P=0.212). Conclusions SMI can detect tiny vessels and depict microvascular architecture of breast lesions as MVI do, which is beneficial for breast tumor differentiation. The diagnostic efficacy of SMI is almost the same as MVI. Key words: Ultrasonography; Breast neoplasms; Superb microvascular imaging; Contrast-enhanced ultrasonography

Differentiation of breast lesions based on viscoelasticity response at sub-Hertz frequencies

There is an urgent need for a noninvasive but more accurate method to assess breast masses before referring the patient to biopsy. In this study, we use the slow viscoelastic creep response to evaluate breast masses at low (<1 Hz) frequencies. This method uses an automated ramp-and-hold compression device and a high-frame-rate ultrasound imaging to track tissue strain. The study cohort included 30 pre-biopsy patients with suspicious masses. Sequential high-frame-rate images were recorded during the ramp-and-hold process. The data were used to calculate the creep response at each point. Then, using a classic linear model, the retardation-time map was created within the field of view. The resulting retardation-time maps exhibited clear distinction of the mass margins. The results showed that benign breast lesions appear with an increase in creep retardation-time compared to surrounding breast glandular tissue, and the opposite trend was true for the malignant lesions. Statistical analysis of the viscoelasticity features revealed that a contrast feature based on the retardation-time is an accurate classifier of malignant and benign masses (P<0.0003). It is concluded that the retardation time is a promising biomarker for differentiation of breast masses. [Work supported by NIH Grant CA168575.]

Efficacy of single voxel 1H MR spectroscopic imaging at 3 T for the differentiation of benign and malign breast lesions

PurposeThe aim of our study was to evaluate the effect of 1H Magnetic Resonance Spectroscopy (MRS) in differentiating breast lesions. Materials and methodsSingle voxel 1H Magnetic Resonance Spectroscopy (1H-MRS) was performed with 3T magnet in 45 women. The choline cut off point was set semi-quantitavely. Sensitivity, specificity and accuracy of MRS were calculated. ResultsTwenty-four of 25 (96%) malignant and 9 of 26 (35%) benign lesions had choline peak. With the use cutoff value of 19,5 MRS provided a 96% sensitivity, 65% specificity and 80% accuracy. ConclusionMRS has a high diagnostic accuracy in differentiating breast lesions.

Ultrasound Elastography Combined With BI-RADS–US Classification System: Is It Helpful for the Diagnostic Performance of Conventional Ultrasonography?

PurposeTo evaluate the additive diagnostic performance of ultrasound elastography (UE) to ultrasound (US) with the 2003 or 2013 Breast Imaging Reporting and Data System (BI-RADS)–US classification systems for the differentiation of benign and malignant breast lesions. MethodsFrom June 2010 to December 2012, 738 women with 770 breast lesions were recruited into this retrospective study. Breast lesions were evaluated separately by US, UE, and both. US assessment was based on the 2003 or 2013 BI-RADS–US, and UE assessment was based on a previously reported 5-point scale. Diagnostic performance of US, UE, and both was compared. ResultsBefore category 4 lesions were subdivided, the area under the receiver operating characteristic curve (AUC) for US, UE, and both were, respectively, 0.735, 0.877, 0.878 (P < .01). When subcategories of 4 lesions were considered, the AUC for US, UE, and both were, respectively, 0.865, 0.877, and 0.883 (P > .05). Adding UE to analysis of 4A lesions can decrease the percentages of malignancy to 2.56%. ConclusionWhen the 2003 BI-RADS was considered, UE could give US some help in differentiating breast lesions. However, when the 2013 BI-RADS was considered, UE gave little help to US, although it reduced unnecessary biopsies of benign category 4A lesions.

Multiplanar Reconstructions of 3D Automated Breast Ultrasound Improve Lesion Differentiation by Radiologists

To investigate the value of multiplanar reconstructions (MPRs) of automated three-dimensional (3D) breast ultrasound (ABUS) compared to transverse evaluation only, in differentiation of benign and malignant breast lesions. Five breast radiologists evaluated ABUS scans of 96 female patients with biopsy-proven abnormalities (36 malignant and 60 benign). They classified the most suspicious lesion based on the breast imaging reporting and data system (BI-RADS) lexicon using the transverse scans only. A likelihood-of-malignancy (LOM) score (0-100) and a BI-RADS final assessment were assigned. Thereafter, the MPR was provided and readers scored the cases again. In addition, they rated the presence of spiculation and retraction in the coronal plane on a five-point scale called Spiculation and Retraction Severity Index (SRSI). Reader performance was analyzed with receiver-operating characteristics analysis. The area under the curve increased from 0.82 to 0.87 (P = .01) after readers were shown the reconstructed planes. The SRSI scores are highly correlated (Spearman's r) with the final LOM scores (range, r = 0.808-0.872) and ΔLOM scores (range, r = 0.525-0.836). Readers downgraded 3%-18% of the biopsied benign lesions to BI-RADS 2 after MPR evaluation. Inter-reader agreement for SRSI was substantial (intraclass correlation coefficient, 0.617). Inter-reader agreement of the BI-RADS final assessment improved from 0.367 to 0.536 after MPRs were read. Full 3D evaluation of ABUS using MPR improves differentiation of breast lesions in comparison to evaluating only transverse planes. Results suggest that the added value of MPR might be related to visualization of spiculation and retraction patterns in the coronal reconstructions.

The Diagnostic Value of Micropure Imaging in Breast Suspicious Microcalcificaion

The purpose of this study was to evaluate the diagnostic value of Micropure Imaging (MI) in breast lesions differentiation by comparison with B-mode ultrasonography (B-US) and Doppler ultrasonography (DU). A total of 135 consecutive patients (all females) with 135 suspicious lesions were examined and skin marked by MI before mammotome biopsies. All patients (age range, 20-86 years; mean age, 42.5 ± 15.6 years) were the first onset, not in the pregnancy or lactation and had no history of radiation or chemotherapy. The maximum diameter of lesions ranged from 0.2 to 9.6 cm (average 1.98 ± 1.3 cm). Their final diagnoses were obtained by histologic examination. The study protocol was approved by the hospital review board; each patient gave written informed consent. One hundred thirty-five breast lesions were classified into 90 nonmalignant and 45 malignant types. To 86 breast lesions with microcalcification, MI showed more microcalcification and coincided better with pathology results than B-US did (P < .05). The specificity of MI was higher than that of B-US and DU; the sensitivity of DU was significantly higher than that of B-US and MI (P < .001). By combining B-US, DU, and MI, the detection accuracy was 86.7%. Receiver-operator characteristic curves showed the area under the curve of B-US, DU, and MI was 0.865, 0.934, and 0.923(P = .000), respectively. Moreover, the interobserver agreements of MI were the highest, 0.922 (observer 1 vs. observer 2), 0.866 (observer 1 vs. observer 3), and 0.916 (observer 2 vs. observer 3). MI as an adjunct ultrasound modality holds some promise in locating and differentiating breast lesions.

Image manifold revealing for breast lesion segmentation in DCE-MRI.

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is widely used for breast lesion differentiation. Manual segmentation in DCE-MRI is difficult and open to viewer interpretation. In this paper, an automatic segmentation method based on image manifold revealing was introduced to overcome the problems of the currently used method. First, high dimensional datasets were constructed from a dynamic image series. Next, an embedded image manifold was revealed in the feature image by nonlinear dimensionality reduction technique. In the last stage, k-means clustering was performed to obtain final segmentation results. The proposed method was applied in actual clinical cases and compared with the gold standard. Statistical analysis showed that the proposed method achieved an acceptable accuracy, sensitivity, and specificity rates.