Breast cancer in dense breasts: comparative diagnostic merits of contrast-enhanced mammography and diffusion-weighted breast MRI

The study was done to compare the value of contrast-enhanced mammography and diffusion-weighted breast MRI in dense breast screening and accurate detection of the breast cancer with correlation of the findings to the histopathological results. The study included 32 female patients having suspicious breast lesions and underwent digital mammography then scheduled for CESM and MRI DW imaging technique. The imaging findings were correlated to the histopathological findings. The study was conducted on 40 breast lesions in 32 female patients having dense breasts; they were classified by the digital mammography into ACR C (59.4%) and ACR D (40.6%). By CESM, there were twenty three lesions (57.5%) as mass lesions and thirteen lesions (32.5%) as non-mass lesions. Four lesions (10%) showed no contrast enhancement. According to the lesion characteristics in diffusion-weighted imaging, the breast lesions were classified into thirty three lesions (82.5%) with restricted diffusion and seven lesions (17.5%) with non-restricted diffusion. The study showed a cutoff ADC value to detect the malignant lesions in the dense breasts ≤ 1.1 × 10-3 s/mm2 at b value of 1000 s/mm2 with a sensitivity of 96.77%, specificity of 66.67%, PPV of 96.77%, NPV of 55.55%, and an overall total accuracy of 92.5%. On comparing the diagnostic accuracy of the CESM to that of the DW MRI, the sensitivity of DW MRI (96.77%) was higher than that of CESM (90.32%). The specificity of DW MRI (66.67%) was higher than that of CESM (33.33%). Total accuracy of DW MRI was higher than that of CESM; they were 90% and 77.5%, respectively. Also, PPV and NPV of DW MRI were 90.91 and 85.71% as compared with 82.35 and 50.00% in CESM, respectively. When comparing the sensitivity of CESM to DW MRI in the detection of multiple breast lesions, they were 88.8 and 100%, respectively. CESM is a useful technique in identification of hidden lesions in mammographically dense breasts. DW MRI is a fast, unenhanced modality that can be used as a breast cancer screening modality. CESM and DWI demonstrated good overall diagnostic accuracy in dense breast patients; however, DW MRI has a higher diagnostic accuracy than CESM for the detection of malignant breast lesions and their multiplicity.


Background
Breast cancer constitutes a major cause of cancer deaths in females. Mammographic screening has been shown to be useful in the reduction of breast cancer mortality; however, the limitations of mammographic screening, particularly in women with dense or non-involuted breasts, are well established [1].
Contrast-enhanced spectral mammography (CESM) is a currently established technique in which contrast enhancement is used with digital mammography to depict tumor vascularity. CESM has been confirmed to be more sensitive than mammography for the diagnosis of breast cancer in dense breasts [2,3].
CESM has a better resolution of a mammography over the MRI as it displays a better assessment of microcalcifications with its details. Furthermore, the images are acquired from one breast compression in a couple of seconds; therefore, there are no motion artifacts [3].
DWI is a fast functional non-contrast technique of the breast MR imaging without the costs and toxicity which occur with DCE-MRI. Breast cancers that are hidden mammographically or clinically and identified by DCE MRI are also detected on DWI and can discriminate them from benign breast lesions by accompanied apparent diffusion coefficient (ADC) mapping [4].
Diffusion-weighted (DW) MRI provides a promise in the detection of breast cancers that are mammographically occult and needs more studies to be used as an alternative supplemental breast cancer screening technique [5].

Methods
This is a prospective analytical study which included 40 breast lesions in 32 female patients having dense breasts; the age range was 29-72 years with mean age 46 years + 9.93 SD. The study was conducted in Baheya Charity Women's Cancer Hospital and Generalized Air Forces Hospital.

Subjects
Patients included in this study were adult females who underwent digital mammography which revealed dense breasts (American College of Radiology (ACR) C or D) and showed indeterminate or suspicious mammographic findings (BIRADS 3, 4, and 5) as well as patients had no detected mammographic abnormality (BIRADS 0) warranting further CESM/DW MRI assessment (Fig. 1).

Contrast-enhanced spectral mammography technique
Patients were examined by dual-energy CESM which was performed using Senographe Essential full-field digital mammography machine (GE Healthcare, Chalfont St-Giles, UK). A one-shot intravenous injection of nonionic contrast agent was administered with a dose of 1.5 ml/kg body weight at a rate of 3 ml/s. Then two subtracted images with contrast agent uptake information were obtained in mediolateral oblique (MLO) and cranio-caudal (CC) views.
The high-energy images were interpreted for the enhancement. Regarding patterns of enhancement seen by CESM, the lesions seen in subtracted images were classified and described into mass and non-mass enhancement lesions.

MRI diffusion-weighted imaging technique
MR imaging was performed on 1.5-T GE optima MR scanner release 450W GE medical systems and 1.5-T scanner (Siemens machine Magnetom Aera, Siemens Medical Systems, Erlangen, Germany) with the patient in prone position, using a double breast coil. All acquisitions were performed in the axial plane. T2-weighted fast spin echo sequences were obtained. DWI was performed using a diffusion-weighted echo-planar imaging sequence with parallel imaging. Diffusion gradients were applied with b = 0 and 1000 s/mm 2 . ADC maps were automatically generated from DW images by the MR software.
T2 images were first assessed then diffusion-weighted images were reviewed to detect the lesion signal intensity, the lesions were considered either nonrestricted diffusion (low-to intermediate-intensity lesions) or restricted diffusion (high-intensity lesions).
Signal intensity was interpreted in a qualitative manner using b value of 1000 s/mm 2 . For apparent diffusion coefficient (ADC) calculation, region of interest (ROI) was manually drawn on the ADC maps. Mean ADC for each lesion was calculated by the lowest ADC value from all examined ROIs within the lesion.

CESM and DW MRI image evaluation
Analysis and interpretation of lesions were detected in both CESM and DW images with determination of the BIRADS category of each lesion and the imaging findings were correlated with the final pathological diagnosis.

Qualitative data was presented by number and
percentage; quantitative data was presented by mean and standard deviation. 2. ROC curve was used to detect validation of the diagnostic value of radiological tools. 3. Inter-method agreement for binary outcomes was examined by calculation of Bennett's prevalenceadjusted bias-adjusted kappa (PABAK). 4. A P value of less than 0.05 was considered statistically significant.

Results
The study included 32 female patients; their age range, main complaints and mammographical ACR classification into ACR C (heterogeneously dense breasts) and ACR D (extremely dense breasts) are demonstrated in Table 1.
The study showed 40 different lesions, as we considered cases which had more than one lesion in the same or both breasts as separate lesions. Their detailed description for the number and percentage of each pathological breast entity are illustrated in Table 2.
From the thirty two cases, four cases (13.8%) had multifocal breast lesions, and five cases (17.2%) had multicentric malignant lesions. There were two lesions that were missed by CESM; however, all of them were diagnosed by DW MRI (Figs. 2 and 3). When comparing the sensitivity of CESM to DW MRI in the detection of multiple breast lesions, they were 88.8 and 100%, respectively. The digital mammographic main findings were  correlated to the histopathological results and demonstrated in Table 3.

CESM findings
CESM classified the lesions into twenty three mass lesions and thirteen non-mass lesions, and four lesions showed no contrast enhancement. CESM findings were correlated to the histopathological results as illustrated in Table 3. The lesions were then described in mass lesions regarding margin of the mass, pattern of enhancement, and intensity of enhancement, while non-mass lesions were described regarding distribution, pattern of enhancement, and intensity of enhancement then correlated versus the histopathological diagnosis as illustrated in Table 4. There were four lesions that showed no contrast enhancement; by digital mammography, they showed asymmetrical density apart from one lesion which showed pleomorphic microcalcifications (Fig. 4). Then the detected breast lesions were categorized according to BIR-ADS classification as illustrated in Table 5.
CESM showed six false-positive lesions; three fibroadenomas, one fibroadenosis, one hamartoma and one post-operative fibrosis (Fig. 5). Also, it showed three false-negative lesions-one IDC (Fig. 2) and two lesions in the same case were proved pathologically to be invasive ductal and tubular carcinomas (Fig. 3).
There were four lesions which showed no contrast enhancement; one lesion proved to be benign (Fig. 7), and three lesions were pathologically proved to be malignant. They were one DCIS (Fig. 4), one mixed invasive ductal and lobular carcinoma, and one invasive ductal and tubular carcinoma (Fig. 3).
ROC curve of CESM for the probability of malignancy using BIRADS showed area under the curve to be 0.579 where the sensitivity and specificity were 90.32%, and 33.33%, respectively, total accuracy 77.5%, PPV 82.3%, and NPV of 50% (Table 6).

DW MRI findings
The detected breast lesions were classified by T2 signal intensity into low T2, isointense T2, and high T2 signal intensities. According to the lesion characteristics in diffusion-weighted image, the breast lesions were classified into thirty-three restricted diffusion lesions and seven non-restricted diffusion lesions, then they correlated to the histopathological diagnosis as illustrated in Table 7.
ADC value of the lesions were correlated to the histopathological diagnosis, where the mean ADC value of the benign lesions was 1.5 + 0.42 × 10 -3 s/mm 2 and that of the malignant lesions was 0.91 + 0.18 × 10 -3 s/mm 2 .
According to the DW MRI characteristics of the breast lesions, they were categorized according to BIRADS classification as illustrated in Table 5.
This study showed one false-negative lesion by DW MRI, which pathologically proved to be ILC (Fig. 6) and three false-positive lesions, one fibroadenoma ( Fig. 7) and two fibroadenosis (Fig. 8).
ROC curve of DW MRI for the probability of malignancy using BIRADS showed that area under the curve is about 0.762 where the sensitivity and specificity were 96.77 and 66.67%, respectively, total accuracy of 90%, PPV of 90.91%, and NPV of 85.71% (Table 6).
ROC curve for probability of malignancy using ADC value showed that area under the curve was about 0.815. A cutoff value of ≤ 1.1 × 10 -3 s/mm 2 had a sensitivity of 96.77%, specificity of 66.67%, total accuracy of 92.5%, PPV of 96.77%, and NPV of 55.55% (Table 6).
Comparing the detected lesions by CESM to that of DW MRI are demonstrated in Table 8.

Discussion
This study presented four multifocal cases; one of them was missed by CESM, while all of them were diagnosed by DW MRI. Furthermore, our study revealed five cases having multicentric malignant lesions; one of them was  [6,7].
CESM mass lesions in this study showed the most common features that described malignant lesions were noncircumscribed margins (speculated or irregular) (69.5%), heterogeneous contrast enhancement (47.9%), and rim enhancement (26.1%). This was matching Schnall et al.'s study which showed that the most important feature of image interpretation is the characterization of the focal mass margin. Irregular or speculated margins have a positive predictive value (PPV) of 84-91%. Rim-like enhancement highly correlates with a cancer diagnosis (PPV, 84%). Also, intense (43.5%) and moderate contrast In the study, CESM non-mass lesions showed that the most common features that described malignant lesions were focal and segmental distribution of non-mass lesions (23.07% for each), regional distribution (15.38%), clumped contrast enhancement (38.46%), heterogeneous contrast enhancement (30.76%), and intense contrast enhancement (38.46%).
In contrast to Schnall et al.'s study, a moderate to marked non-mass regional enhancement provides a PPV of 59% in malignant lesion detection. Stippled enhancement was found that it has a low incidence of malignancy (25%), while clumped, heterogeneous, and homogeneous enhancements were found to have a 60%, 53%, and 67% likelihood of cancer, respectively [8].
In the current study, CESM showed six false-positive lesions (19.3% false-positive rate), three fibroadenomas, one fibroadenosis, one hamartoma, and one postoperative fibrosis.
In contrast to Luczyn´ska et al. and Muller et al.'s studies, there was a relatively high rate (20%) of falsepositive results and no false-negative findings with CESM. Benign lesions, such as fibroadenomas, fibrosclerosis, hamartoma, intraductal papillomas, and phyllodes tumors had shown contrast enhancement [7,9].
In this study, CESM showed three false-negative lesions (33.3% false-negative rate); one IDC and two lesions invasive ductal and tubular carcinomas in the same case. Furthermore, there was one DCIS case which showed no contrast enhancement; however, the presence of pleomorphic microcalcifications raised the   [10]. CESM in this study presented two postoperative cases; one of them was pathologically proved to be recurrent IDC, while the other one showed suspicious contrast enhancement by CESM and revealed pathologically to be postoperative fibrosis.
In line with Helal et al.'s study, they had nine falsepositive cases that were found out in the comparison of the CESM diagnoses with the histopathology results. These cases were wrongly diagnosed because the operative bed showed areas of enhancement, but these enhancements were caused by benign postoperative sequel [11].
This DW MRI study showed that the mean ADC value of the benign lesions was 1.5 + 0.42 × 10 -3 s/mm 2 which is higher than that of the malignant lesions which was 0.91 + 0.18 × 10 -3 s/mm 2 . This was in concordance with Moukhtar and Abo El Maati's study, where they reported that the mean ADC value of all benign lesions was 1.41 ± 0.36 × 10 -3 s/mm 2 , which was higher than the mean ADC of all malignant lesions (1.05 ± 0.30 × 10 -3 s/mm 2 ) [12].
Our study showed that a cutoff ADC value ≤ 1.1 × 10 -3 s/mm 2 at b value of 1000 s/mm 2 had a sensitivity of 96.77%, specificity of 66.67%, PPV of 96.77%, NPV of 55.55%, and with an overall total accuracy of 92.5%. This was in concordance with Moukhtar and Abo El Maati's study which showed that the optimal cutoff value to discriminate benign from malignant lesions was 1.25 × 10 -3 s/mm 2 , with a sensitivity of 82%, a specificity of 68%, and an overall accuracy of 78% [13].
This study showed that IDCs had a mean ADC value (0.92 + 0.14 × 10 -3 s/mm 2 ) which was slightly lower than  that of ILCs which showed mean ADC value (0.93 + 0.29 × 10 -3 s/mm 2 ). Our study had only one DCIS lesion with ADC value relatively higher than that of the invasive lesions; it was 1.1 × 10 -3 s/mm 2 . Woodhams et al.'s study reported that because of their higher cellularity, most of IDCs show higher signal intensity and lower ADC values than that of benign tumors and normal breast parenchyma on diffusionweighted images [12].
In this study, DWI showed one false-negative lesion (2.5% false-negative rate); it was ILC which showed non-restricted diffusion with high ADC value of 1.5 × 10 -3 s/mm 2 .   According to Woodhams et al.'s study, invasive lobular carcinoma represents a diagnostic challenge at MR imaging as well as its size may be underestimated at diffusion-weighted imaging; this inaccuracy may be due to the spread of infiltrating cells [12].
DWI in this study showed three false-positive lesions (33.3% false-positive rate); one fibroadenoma and two fibroadenosis. Fibroadenoma showed restricted diffusion with low ADC value 0.8 × 10 -3 s/mm 2 . Two fibroadenosis lesions showed restricted diffusion as well as low ADC value 1.1 and 1.2 × 10 -3 s/mm 2 .
According to Pereira et al.'s study, fibroadenomas are supposed to have high rates of diffusion and ADC values owing to their stromal myxoid changes and consequently increased mobility of water. However, fibroadenomas with an abundant fibrous component have lower ADC values. Additionally, fibrocystic disease, which is characterized by considerable degrees of fibrosis and proliferation, may have ADC values in the range of malignant lesions [14].
This study revealed that the sensitivity and specificity of the DW MRI to detect malignant lesions were 96.77 and 66.67%, respectively, PPV of 90.91%, and NPV of 85.71% with a total accuracy of 90%.
This was in concordance with Partridge and McDonald's study which established that the diagnostic performance of quantitative breast DWI exhibited pooled sensitivity of 84% and specificity of 79% [4].
In this study, the diagnostic accuracy of DW MRI was higher than that of CESM where the sensitivity of DW MRI was 96.77% as compared with 90.32% in CESM. The specificity of DW MRI was 66.67% as compared with 33.33% in CESM. The total accuracies were 90 and 77.5%, respectively. Also PPV and NPV of DW MRI were 90.91 and 85.71% as compared with 82.35 and 50.00% in CESM, respectively. This was in concordance with Barra et al.'s study; among the 25 patients who had residual lesions, 19 were positive by CESM, and 23 were positive by MRI. Higher sensitivity was found by MRI (92%) in contrast to 76% in CESM. Also PPV and NPV were higher for MRI comparable to CESM; they were 95 and 75% as compared with 92 and 53.8%, respectively [15].
Our study had several limitations. First, our study population was small. Second, pre-and postmenopausal women were included and were examined in different phases of the menstrual cycle. Third, most cases were collected from the cancer breast center, so the negative lesions were small in number. Finally, women with no suspicious lesions at conventional mammography or CESM mostly did not undergo DW MRI and thus were not included in this study.

Conclusion
Dual-energy contrast-enhanced digital mammography is a useful technique in identification of lesions in mammographically dense breasts and capable of demonstrating lesions that are not visible by standard mammography. In comparison with MRI, CESM can detect microcalcifications easily, and there are no limitations as with MRI in terms of the ferromagnetic effect and machine design. DW MRI technique is a diagnostic technique that enables accurate detection of malignant breast lesions without need for the contrast media injection, and it avoids the irradiation exposure.
CESM and DWI demonstrated good overall diagnostic accuracy and correlation in lesion size estimation in dense breast patients. However, DW MRI has a higher diagnostic accuracy than CESM for the detection of malignant breast lesions in dense breasts with a higher sensitivity, specificity, total accuracy, negative predictive value, and positive predictive value as well as the detection of multiple lesions.   Bennet's prevalence-and bias-adjusted kappa (PABAK) 0.45