Three-dimensional tomosynthesis versus two-dimensional mammography in detection and characterization of different breast lesions

Breast cancer is considered the most serious lesion among different breast lesions. Mammography is the corner stone for screening for detection of breast cancer. It has been modified to digital mammography (DM) and then to digital breast tomosynthesis (DBT). Tomosynthesis is an emerging technique for diagnosis and screening of breast lesions. The aim of this study is to interrogate whether the addition of DBT to DM helps in better detection and characterization of different breast lesions. This is a prospective study carried on 38 female patients according to our inclusion criteria. All patients were evaluated by using DM alone and thereafter with the addition of DBT to DM. Recall rate was calculated, and the imaging findings of each case were correlated with the final diagnosis and follow-up. DM identified 32 lesions while DBT with DM identified 37 lesions. On DM alone, 17 lesions were characterized as masses, 5 as focal asymmetry, 2 as architectural distortion, 7 as microcalcification and 1 as macrocalcification. With the addition of DBT, 27 lesions were characterized as masses, 1 as focal asymmetry, 1 as architectural distortion, 7 as microcalcification and 1 as macrocalcification. So, there were better detection and characterization of lesions with the addition of DBT than DM alone. The sensitivity, specificity, AUC, positive and negative predictive values were significantly higher with the addition of DBT to DM (100%, 90.5%, 0.952, 90% and 100%, respectively) than with DM (77.8%, 80.9%, 0.794, 77.8% and 80.9%, respectively) for all breast lesions. The addition of DBT to DM helps in better detection and characterization of different breast lesions. This leads to early detection of breast cancer, improvement of the performance of radiologists and saving time by reduction of recall rate.

Breast cancer is the most common cancer and the primary cause of mortality from cancer among females around the world. Its survival rate in developing countries is generally poorer than in developed countries, primarily as a result of delayed diagnosis of cases. Enhancing breast cancer outcome and survival by early detection remains the foundation of breast cancer regulations [1] (Fig. 1).
In the 1970s, mammography gained widespread acceptance as a breast screening tool for cancer detection. It was shown to reduce mortality rate. From that time, technological advancements have driven the evolution from analog film mammography to full-field digital mammography (FFDM) and DBT [27] (Fig. 2).
DBT is a modified mammographic technique, which involves acquired multiple and low-dose projection images of the breast through a limited range of angles less than 60°. For the projections acquisition, the X-ray tube rotates around the static and compressed breast. The images are reconstructed into a stack of focal planes according to the height above the detector, typically at 1-mm intervals, to create a three-dimensional (3D) volume of thin sectional data. This algorithm uses the different locations in the projections of the same tissue to compute their vertical position that viewed sequentially as a stack in orientation (craniocaudal (CC) and mediolateral oblique (MLO)), so estimating the 3D distribution of the tissue [11] (Fig. 3).
3D tomosynthesis offers a potential advantage for detection of masses, architectural distortion and asymmetries compared with conventional two-dimensional digital mammography (2DDM) images [14].
The aim of this study is to interrogate whether the addition of DBT to DM helps in better detection and characterization of different breast lesions.

Patients
This is a prospective study conducted at our institute during the period from December 2018 to December 2020 as follow-up was needed in some cases. A total of 38 female patients were included in this study. Their ages ranged from 31 to 65 years with a median of 41.5 years. They were provided with a signed informed consent, and the study was conducted after institutional review board approval by Radiology Department Scientific Board as well as fulfilling the ethical guidelines of the institute (Fig. 4). All patients were evaluated by DM alone and then after adding DBT. Recall rate was calculated and the imaging findings of each case were correlated with the final diagnosis, which was made based upon histopathological assessment either by biopsy samples or by fine needle aspiration cytology, and 3 follow-up studies every 6 months for some cases.

Inclusion criteria
Women included in this study who were referred from breast clinic for either screening or follow-up of already present lesion or for diagnoses of complains such as palpable lump, nipple discharge or breast pain (Fig. 5).

Exclusion criteria
Pregnant and lactating females. Females with open breast wounds (Fig. 6).

The technique of DM and DBT
Both examinations were done on Senographe Pristina, GE healthcare FFDM machine with 3D digital tomosynthesis option which enables the machine to generate both 2D and 3D images. For the projections acquisition, the X-ray tube rotates around the static and compressed breast between 15° (narrow range) and 60° (wide range) in a plane aligned with the chest wall allowing for 11 to 15 low-dose projection images (2D) acquired for the tomosynthesis images. The images were reconstructed into a stack of focal planes according to the height above the detector, typically at 1-mm intervals, to create a 3D volume of thin sectional data from the low-dose projection 2D images used to reconstruct 1 mm thick sections separated by 1 mm space. Images were obtained in the same standard projections (CC and MLO) in the form of a series of images through the entire breast. Images were assessed on the workstation (Fig. 7).

Image analysis and interpretation
In each case, bilateral scans were jointly reviewed by two experienced radiologists in breast imaging. All cases were categorized by age and breast density according to American College of Radiology ACR guidelines edition 2013.
Each breast was evaluated about the presence of lesions or not, location, morphological type (mass, architectural distortion, focal asymmetry, macrocalcifications and microcalcifications) and for mass lesion; number, shape, density and margins were recorded using DM alone and then after adding DBT. Any founded lesion was classified according to the BIRADS lexicon 2013 classification. The obtained data were correlated with the final diagnosis obtained by histopathological examination and close follow-up (3 follow-up studies every 6 months) (Fig. 8).

Statistical analysis
Results were performed using MedCalc statistical software for windows (MedCalc software, Mariakerke and Belgium). Data for continuous variables were expressed as either median, interquartile range, range of mean (± standard deviation) or both number and percentage for categorical data. Receiver operator characteristic (ROC) curve analysis was performed to determine the diagnostic accuracy of the various variables in distinguishing the different groups. The diagnostic accuracy of all variables was evaluated in terms of sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and area under the ROC curve (AUC). CHI-squared test was used for comparison of categorical data. For all tests, all P values were two-tailed and a P-value < 0.05 was considered significant (Fig. 9).

Results
The current study included 38 females with 39 imaging findings. Their ages ranged from 31 to 65 years with a median of 41.5 years, an IQR of 39-49 years and a mean (± standard deviation) of 44 years (± 8.91 years). Of the 38 women, predominantly fatty (ACR a), scattered fibro-glandular (ACR b), heterogeneously dense (ACR c) and extremely dense (ACR d) breasts were found in 3 (7.9%), 19 (50%), 14 (36.8%) and 2 (5.3%) women, respectively. The final diagnosis by histopathological evaluation and follow-up revealed 2 (5.1%) normal cases, 19 (48.7%) benign lesions and 18 (46.2%) malignant lesions (Fig. 10). The distribution of the lesions according to the BIRADS score on DM and after adding DBT can be seen in Table 1. In correlation with the final diagnosis according to the BIRADS score, the sensitivity, specificity, PPV, NPV, accuracy and AUC of DBT plus DM were 100%, 90.5%, 90%, 100%, 94.9% and 0.952, respectively, and of DM were 77.8%, 80.9%, 77.8%, 80.9%, 79.5% and 0.794, respectively. As regards TP, TN, FP and FN with the addition of DBT were 18, 19, 2 and 0, respectively, and with DM were 14, 17, 4 and 4, respectively. Moreover, adding DBT shows better overall efficacy reaching 94.9% as compared to 79.5% for DM. Pairwise comparison of AUC of both modalities revealed that the AUC with the addition of DBT was significantly larger than the AUC with DM (P = 0.039) indicating that addition of DBT was significantly more accurate than DM in Out of 39 imaging findings, seven lesions were missed on DM but appeared as masses on the addition of DBT. One lesion appeared as focal asymmetry on DM but as a mass on DBT. Two lesions appeared as architectural distortion on DM but as masses on DBT. Two lesions appeared as focal asymmetry on DM but were undetectable on the addition of DBT. One lesion appeared as focal asymmetry on DM but as architectural distortion on DBT. Otherwise the presentation of the remaining lesions was the same. This is shown in Table 2.
Regarding the performance of DBT and DM in the characterization of breast masses according to shape,  (Tables 3 and 4, respectively), DBT was more accurate than DM in differentiation of malignant and benign breast masses according to shape, margins and density.

Discussion
Despite of using FFDM in screening, it has a limitation in detection of different breast lesions as fibro-glandular tissue overlapping which is part of the nature of the imaging method makes it very difficult to distinguish abnormalities from normal anatomical structures [20]. So DBT is a modified 3D mammographic technique that overcome this limitation [14].
In our study, we compare the performance of 2DDM alone and with the addition of DBT in detection and characterization of different breast lesions in different breast densities at different ages in females. We find out that the performance of addition of DBT is better than 2DDM alone in correlation with the final diagnosis. Out of the included 39 lesions, the sensitivity, specificity, PPV, NPV and AUC are significantly higher with the addition of DBT (100%, 90.5%, 90%, 100% and 0.952, respectively) than that with DM (77.8%, 80.9%, 77.8%, 80.9% and 0.794, respectively). TP and TN are significantly higher with the addition of DBT (18 and 19, respectively) than with DM (14 and 17, respectively). FP and FN are lower with the addition of DBT (2 and 0, respectively) than with DM (4 and 4, respectively). Moreover, adding DBT shows better overall efficacy reaching 94.9% as compared to 79.5% for DM.
The findings of our study lie in concordance with previous studies. Mall et al. [15] evaluated 144 women aged Another mass is detected in retroareolar area which is dense rounded with partially well-defined margins by 2DDM and which is dense irregular with partially speculated margins by DBT. BIRADS 4, proved by histopathology to be invasive ductal carcinoma so the diagnosis changed to be multifocal invasive ductal carcinoma and the decision of treatment will be changed more than 40 years in Australia and found that the sensitivity, specificity, PPV, NPV and AUC were significantly higher with the addition of DBT (93%, 75%, 64%, 96% and 0.927, respectively) than that with DM (90%, 56%, 0.49%, 92% and 0.872, respectively), TP and TN were significantly higher with the addition of DBT (226 and 375, respectively) than with DM (218 and 283, respectively) and FP and FN were lower with the addition of DBT (126 and 16, respectively) than with DM (222 and 24, respectively). Singla et al. [24] evaluated 100 women and found that the sensitivity and specificity were significantly higher with the addition of DBT (100% and 76.4%, respectively) than with DM (83.6% and 38.78%, respectively). Tucker et al. [28] evaluated 7060 women and found that the sensitivity and specificity were significantly higher with the addition of DBT (91% and 68%, respectively) than with DM (86% and 56%, respectively). Alakhras et al. [2] evaluated 50 women and found that the sensitivity, specificity and AUC were significantly higher with the addition of DBT (70.4%, 78.3% and 0.788, respectively) than with DM (63%, 65.2% and 0.681, respectively). Gillbert et al. [10] evaluated 7060 women and found that the sensitivity, specificity and AUC were significantly higher with the addition of DBT (89%, 69% and 0.89, respectively) than with DM (87%, 58% and 0.84, respectively). Michell et al. [16] evaluated 738 women and found that the sensitivity and specificity were significantly higher with the addition of DBT (100% and 76.4%, respectively) than with DM (83.6% and 38.78%, respectively). In contrast, Ohashi et al. [19] evaluated 628 women and found that there is no significant difference for AUC with the addition of DBT (0.9376) and DM (0.9160), also a statistically significant difference for specificity with the addition of DBT (98.9%) over DM (99.1%) but the sensitivity was significantly higher with the addition of DBT (83%) than with DM (61%). Yi et al. [31] evaluated 265 women in Korea and found that the sensitivity, specificity, PPV and NPV were non-statistically significant with the addition of DBT in 55 women with extremely dense breast (63.6%, 84.8%,79.2% and 90.3%, respectively) than with DM (59.1%, 75.8%, 61.9% and 73.5%, respectively), but in 210 women with other breast density they found that specificity and PPV were significantly higher with the addition of DBT (98.4% and 97.6%, respectively) than DM (90.5% and 76.8%, respectively). In our study, the mass detection rate is higher with the addition of DBT (69.2%) than with DM (43.6%) and there is accurate detection of mass margins with the addition of DBT in comparison with DM (92.6% vs. 76.5%). This agrees with Mohindra et al. [17], Yang et al. [30], Mun et al. [18], Hakim et al. [12], Andersson et al. [4] and Poplack et al. [21] studies. Mohindra et al. [17] evaluated 164 women and found that there was statistically significant with the addition of DBT in detection of masses comparing to DM (97.6% vs. 87.6%). Also, there was statistically significant with the addition of DBT in detection of speculated margins in comparison with DM (56.5% vs. 34.7%).
Regarding focal asymmetry in our study, one lesion appeared as focal asymmetry on DM but was detected as a mass on DBT. Two lesions appeared as focal asymmetry on DM but were undetectable on DBT (normal cases on follow-up). Our study goes with Skaane et al. [25] study, where 7 cases were categorized as normal as their lesions were obscured and 2 another normal case were categorized as focal asymmetry on DM. However, upon interrogating the DBT slices, the lesions were clearly seen, and focal asymmetry faded away.
In this study, detection and characterization of calcifications were similar with using DM or DBT. This agrees with Li et al. [13] and Chu et al. [7] in which they found that calcifications can be diagnosed using DM and DBT with similar sensitivity.
Regarding the architectural distortion in our study, two lesions appeared as architectural distortion on DM but appeared as masses on DBT. one lesion appeared as focal asymmetry on DM but presented as architectural distortion on DBT. Thus, our study agrees with Dibble et al. [9], in their study mentioned that the sensitivity   of detection of architectural distortion with DBT was found to outperform DM, but specificity was found to be similar between DM and DBT.
Regarding BIRADS score in our study, the addition of DBT allowed more confident up or down grading of the BIRADS score of a lesion. For example, three lesions were upgraded from BIRADS 1 to 2, three lesions were upgraded from BIRADS 1 to 3, one lesion was upgraded from BIRADS 1 to 4, two lesions were upgraded from BIRADS 3 to 4, two lesions were upgraded from BIRADS 3 to 5 and one lesion was upgraded from BIRADS 4 to 5. On the other hand, two lesions were downgraded from BIRADS 3 to 2, one lesion was downgraded from BIRADS 4 to 3 and two lesions were downgraded from BIRADS 4 to 1. This goes with Bahrs et al. [5] whom evaluated 87 patients and found that 4.6% lesions were upgraded to BIRADS 4 and 57.1% lesions were downgraded from BIRADS 3 to 1 or 2 by the addition of DBT.
Regarding recall rate in our study, there is a decrease in recall rate with the addition of DBT 5% versus 13% with DM alone. Many studies were done to compare recall rate with DM alone versus with the addition of DBT, Cohen et al. [8]    found significantly decrease in the recall rate with the addition of DBT 14% versus 16% with DM. In contrast, Pattacini et al. [32] evaluated 19,560 women and found that no significant decrease in the recall rate with the addition of DBT, it was the same 3.5%.
There is a limitation in our study which is a relatively small number of cases.

Conclusions
Addition of DBT to DM helps in better detection and characterization of different breast lesions. This leads to early detection of breast cancer, improvement of the performance of radiologists and saving time by reduction of recall rate.