Skip to main content

What’s beyond breast asymmetry? Comparative study between artificial intelligence and contrast-enhanced spectral mammography in the assessment of various types of breast asymmetries



Breast asymmetries are prevalent findings in mammograms and are commonly caused by variations in normal breast tissue. However, they may imply significant underlying causes in some cases. Such cases necessitate further assessment by adding further mammography views, targeted ultrasound, and investigations to exclude underlying pathology.


To investigate the role of artificial intelligence (AI) compared to contrast-enhanced spectral mammography (CESM) in the assessment of breast asymmetries and their performance as diagnostic modality among different types of breast asymmetries as well as the additive value of AI software to mammography in these cases.


Sixty-four female patients were diagnosed with breast asymmetries by standard mammography (MMG) on both craniocaudal (CC) and mediolateral oblique (MLO). Digital breast tomosynthesis (DBT) may have been added. After evaluating the breast asymmetry by MMG and complementary breast ultrasound (US), both CESM and AI were performed for all cases and analyzed, then the interpreted results were compared accordingly either by histopathology from suspected lesions scored as BI-RADS 4 or 5 or through further close follow-up by single-view mammography in benign cases scored as BI-RADS 2 or 3.


The sensitivity and specificity of CESM in the assessment of breast asymmetry in correlation with pathological data/follow-up results were 100% and 60% (p < 0.001). The corresponding values for AI were 70.83% and 75%; however, the estimated overall accuracy for both CESM and AI was close to each other measuring 75% and 73.44%, respectively (p < 0.001). The diagnostic accuracy of CESM to detect malignant causes of breast asymmetry was 100%; however, the detection of benign causes of breast asymmetry was 40%. The corresponding values for AI were 70.83% and 25%, respectively, with significant p-value (p < 0.001).


The CESM was more sensitive; however, the AI was more specific in the assessment of different breast asymmetries. Although the diagnostic accuracy of both is close to each other. Therefore, AI-aided reading can replace CESM in most cases, especially for those contraindicated to do CESM. AI also can reduce the radiation exposure hazards of a second dose of radiation for CESM and its financial cost as well. AI-aided reading in breast screening programs can reduce the recall of patients, unnecessary biopsies, and short-interval follow-up exams.


Breast asymmetries are diagnosed by MMG on both CC and MLO. Furthermore, investigations may be used for dedicated assessment and detection of what is beyond the asymmetry and to improve radiologists’ sensitivity and specificity [1].

Mammography has a high recall rate because it frequently produces false positive results. Hence, the primary purposes of AI in mammography are to improve the diagnosis of serious cancer and lower the recall rate [2].

Moreover, AI has been applied to the interpretation of mammography pictures to determine the likelihood of developing breast cancer, which could be a crucial step in the implementation of individualized screening [3].

In a similar manner to dynamic contrast-enhancing magnetic resonance imaging (DCE-MRI), without the added time or expense of conventional breast MRI protocols. CESM employs iodinated contrast materials to visualize breast neo-vascularity [4].

The combination of morphologic and functional information provided by CESM has been shown to provide superior sensitivity and specificity in diagnosing breast cancer when compared with standard mammography alone [5].

In most studies, AI has been used to analyze mammography images mainly for detecting and classifying breast mass and microcalcifications, breast density assessment, breast cancer risk assessment breast mass segmentation, and image quality improvement [6].

In the current study, we aimed to assess breast asymmetry using artificial intelligence compared to CESM and evaluate their performance as a diagnostic modality among different types of breast asymmetries. We also aimed to investigate the additive value of AI-aided reading to mammography as an advanced new technique in the assessment of breast asymmetry.


Between January 2022 and January 2023, this case series prospective study was carried out. The work was accepted by the institutional review board's ethical committee. Patients that were included gave their informed consent.

Sixty-four patients had breast asymmetry detected radiologically by mammography either screening or diagnostic on both the CC and MLO. A breast ultrasound was done, and then each breast was categorized using standard reporting BI-RADS ACR 5th edition. Digital breast tomosynthesis (DBT) may have been added for dedicated evaluation and localization of the lesion. Then all patients were subjected to CESM and AI software to evaluate the breast asymmetry underlying cause.

Image interpretation was carried out by two radiologists with at least 10–15 years of experience in breast imaging and about 3 years of experience in AI-aided reading. The final diagnostic and clinical data were concealed from the radiologist.

Patients were eligible to be included if they fulfilled the following criteria: Female gender, range of age (35–70 years), and diagnosed breast asymmetry by diagnostic or screening MMG. Exclusion criteria included: Patients with known contraindications to performing CESM, e.g., pregnancy, contrast allergy, renal failure, missed pathological data or follow-up exams as well as withdrawal of consent at any time.

Breast mammography technique was performed by digital mammography (manufacture: Amulet Innovality, Fujifilm Gobal company, Japan). Mammography machines were supported with a “Bellus” workstation of resolution five megapixels. Standard two views were taken for each breast in the CC and MLO views. In ten cases (15.63%) of dense breast parenchyma and focal asymmetry, DBT had been added.

Breast ultrasound (US) was performed (HS60 Samsung ultrasound, Korea, 2019 device), and equipped with a linear probe of 9–13 MHz was used. All the real-time scanning was performed by a radiologist with at least 10 years of experience in breast ultrasound. The asymmetry was assessed by the US using the BI-RADS atlas of 2013.

Artificial intelligence images were generated from mammographic images by (Lunit INSIGHT MMG, Korea, version 2019) for Fujifilm digital mammography system. Before interpretation, the computer-aided system (CAD) examined each mammography qualitatively through a heat map or quantitatively through a probability of malignancy (POM) score (0–100%).

The computer-assisted system consists of two independent units: the processing unit, which digitizes and evaluates the film images; and the display unit, which is a dedicated mammography auto-viewer monitors that shows low-spatial-resolution digital images of the examination that are hung in the panels above. Each digital image may contain zero or more marks, indicating areas that warrant evaluation by the radiologist. According to the POM, the AI category was determined for each breast as follows: Normal—Low (< 10%), Benign—10–39%, Probably benign—40–59%, Probably malignant—60–79%, and Malignant—80–100%.

CESM (GE Healthcare, Chalfont St. Giles, UK) was used. It was based on a cesium iodide scintillator. Prior to CESM imaging, the patient’s clinical and medical history of any drug sensitivity was taken as well as kidney function tests were also checked. The intravenous iodinated contrast medium (Omnipaque) was injected (300 mg/ml, 1.5 ml/kg of body weight) manually. A set of images were acquired in CC and MLO views at 2 min following the injection, starting with the suspicious breast and moving onto the other breast. The patient was then positioned for standard mammography. Two automatic X-ray exposures were used to capture each view: one at a low energy that serves as the 2-D mammogram and the other at a high energy level which provides means for the contrast-enhanced portion of the exam. Images were then subtracted to suppress the non-enhancing tissue and accentuate enhancing lesions [7]. Positive cases in CESM were characterized in this study as having an enhancing impact stronger than the surrounding normal breast tissue. The presence or absence of contrast enhancement was mainly assessed on the recombined images and based on contrast enhancement and morphology like those described in the BI-RADS MRI lexicon developed by the American College of Radiology (ACR).

The diagnostic performance of AI and CESM was analyzed and correlated with pathology or follow-up studies according to the calculated BI-RADS. Follow-up by single-view mammography was used as a reference standard in benign cases scored as BIRADS 2 or 3. While patients with BIRADS 4 or 5 went for histopathological examination by US-guided biopsy from the most suspicious lesion.

The statistical program SPSS (Statistical Program for the Social Sciences) version 28 was used to code and enter the data (IBM Corp., Armonk, NY, USA)., the mean, standard deviation, minimum, and maximum were used for quantitative data; however, for categorical data, frequency (count) and relative frequency (%) were used to describe the data. Using a Chi-square test, categorical data were compared. When the anticipated frequency is less than 5 [8], an exact test was utilized in its place. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and diagnostic efficacy of common diagnostic indices were calculated. Statistical significance was defined as a p-value of 0.05 [8].


The patient's ages ranged from 35 to 70 years. The average was 45.8 years (standard deviation of 7.4). The most important and frequent clinical data were obesity, hormonal contraception usage for more than five years and a history of benign breast lesions as well as positive family history (7.8%).

The main presenting symptoms were palpable breast lumps affecting 67.2%, followed by non-palpable breast lesions such as those who came for regular screening were about 23.4%, followed by mastalgia at 6.3%, nipple discharge in one patient, and another nipple retraction.

ACR B and C were the most commonly reported among the included patients accounting for 43.8% and 53.1%, respectively. Microcalcifications were detected in ten patients (15.60%).

All included patients showed unilateral breast affection. About 81.3% showed focal asymmetry, followed by global asymmetry among 14.1% of the included patients (Fig. 1).

Fig. 1
figure 1

Bar chart showing the type of asymmetry among the included patients

The right and left breast affection was almost similar accounting for 53.1% and 46.9%, respectively. The most common site of affection was the UOQ asymmetry accounting for 45.3%, followed by retro-areolar region (21.9%), UIQ (12.5%), LIQ (9.4%), and diffuse (9.4%) of the included patients and only one patient with LOQ affection.

Regarding the artificial intelligence (AI) system, the low probability cases were about 57.8%, followed by moderate probability cases were about 25%, yet the high probability cases were in 17.2% of the included patients (Fig. 2).

Fig. 2
figure 2

Pie chart showing AI findings

All patients were assessed by contrast-enhanced spectral mammography (CESM) (Fig. 3). The examinations elicited 21 (32.8%) patients were negative studies; however, the positive studies were among 1.6% showed benign-looking enhancing lesions, 4.7% revealed benign-looking non-mass enhancement, while 21.9% had malignant-looking enhancing mass, followed by suspicious non-mass enhancement (39.1%).

Fig. 3
figure 3

Bar chart showing CESM findings

The most common BI-RADS classification among patients was BIRAD 4 (53.1%), followed by BIRAD 3 (26.6%) then BIRAD 5 (12.5%) (Table 1).

Table 1 Sono-mammographic BI-RADS among the included patients

In this study, a follow-up assessment by single-view mammography was only conducted for 15 (23.4%) patients, who were categorized BI-RADS 2 or 3, the assessment showed that some diagnosed asymmetries were stable in the course and others regressed with no underlying suspicious lesions in other additional exams.

The histopathology revealed that 24 (37.5%) lesions were malignant, while 22 (34.4%) lesions were benign (Fig. 4).

Fig. 4
figure 4

Pie chart showing pathological findings

The subdivisions of different pathologies were as follows: DCIS 6.3%, IDC 28.1%, interlobular carcinoma in one patient, and mixed IDC and interlobular carcinoma in another one. While the remaining patients reported benign lesions such as sclerosing adenosis, periductal mastitis, granulomatous mastitis, fibrocystic, fibroadenoma, and fibro-adenosis.

The correlated results of the AI showed that the true positive results were 26.6% (17 patients), the false positive results were 15.6% (10 patients), the true negative rate was 46.9% (30 patients), and 10.9% (7 patients) of the examined patients were false negative cases (Table 2).

Table 2 Comparison of AI findings to histopathology

However, the correlated results of CESM were true positive results 37.5% (24 patients), false positive results 25.0% (16 patients), true negative 37.5% (24 patients), and none of the examined patients were false negative cases (Table 3).

Table 3 Comparison of CESM findings to histopathology

The diagnostic performance indices of AI versus CESM revealed that the CESM was more sensitive; however, the AI was more specific in diagnosing different breast asymmetries. Although the diagnostic accuracy of both is close to each other (Fig. 5).

Fig. 5
figure 5

Bar chart showing the diagnostic performance indices CESM and AI among the included patients


Using additional modalities such as digital breast tomosynthesis, AI-CAD, and/or CESM in the assessment of breast asymmetries will improve the efficiency of breast cancer screening and diagnosis which subsequently leads to lower the frequency of unnecessary biopsies as well as short-interval follow-up exams (Fig. 6).

Fig. 6
figure 6figure 6

A 42-year-old female patient complained of a left breast lump. Pathology revealed: invasive duct carcinoma grade III. a both MLO and CC mammography views revealed heterogeneously dense breasts (ACR c) along with left LOQ global asymmetry (white arrows). b spot image shows overlying groups of amorphous micro-calcifications. c) complementary US shows a large area of altered parenchyma, with multiple confluent masses with internal echogenic calcific foci from 3 to 6 o’clock, the largest measuring about 5.2 × 3.8 cm. d AI shows a left high abnormity score (96%). e CESM images revealed a left LOQ large irregular heterogeneously enhancing mass and non-mass lesions measuring collectively about 9 × 9 cm. Both AI and CESM were true positives in this case

We aimed in this study to evaluate the diagnostic performance of AI compared to CESM in the assessment of what is beyond different types of breast asymmetry.

In the present study, breast asymmetry was detected mainly in UOQ accounting for 45.3%, 81.3% of the included patients showed focal asymmetry, and 15.6% showed associated microcalcifications (Fig. 7).

Fig. 7
figure 7figure 7

A 52-year-old female patient came complaining of a left lump. Pathology revealed invasive mixed ductal and lobular carcinoma. a both MLO and CC mammography views: revealed heterogeneously dense breast parenchyma (ACR c) along with the left retro-areolar focal area of asymmetry as well as right multiple scattered oval-shaped circumscribed and partially obscured lesions. b Spot image shows internal malignant-looking pleomorphic clusters of microcalcifications and surrounding parenchymal distortion. c Complementary US of left breast shows an area of altered parenchyma. d US of the right breast shows adenosis with tiny cystic changes. e AI revealed a left high abnormality score of 96% and a right low probability score. f CESM (CC views) revealed left retro areolar intensely enhancing, spiculated mass along with outer nodular non-mass enhancement. Both AI and CESM were true positives in this case

These findings were matching with the evidence in the literature that UOQ is the most reported site for breast masses and asymmetry detection, as well as focal asymmetries are more commonly reported compared to global or non-mass asymmetries. The prevalence of microcalcifications is different in each study and usually corresponds to a number of screened females, and it ranges from 18.5 to 40% of breast cancer patients [9, 10].

In a retrospective study conducted by Wessam et al. [11], results showed that after screening 125 patients with breast asymmetries, 88 (70.4%) females had focal asymmetry while 26 (20.8%) had global asymmetry, these findings were consistent with our results (Fig. 8).

Fig. 8
figure 8figure 8

A 41-year-old female patient complaining of a left palpable lesion. Pathology revealed PASH with no underlying malignancy cells. a) Mammography CC and MLO views showing heterogeneously dense breast parenchyma (ACR c) along with an area of left upper asymmetry. b Complementary US shows an area of condensed glandular tissue (palpable clinically) with no underlying suspicious lesions c enlarged left axillary lymph node, showing rather uniform cortical thickening (5 mm), yet with preserved fatty hilum. d AI result shows left low abnormality score. e CESM shows delayed faint non-mass enhancement, more appreciated at MLO view, measuring about 4.6 × 3.4 cm along its maximum dimensions.AI was true negative however the CESM was false positive in this case

Our data showed that CESM elicited 21 patients (32.8%) were negative studies; however, the positive studies revealed that the most common pattern of enhancement was the suspicious non-mass enhancement (39.1%) followed by malignant looking mass (21.9%). This was similar to Dawoud et al. [12] who found that the pattern of contrast mammography findings was no-enhancement in 71/540 asymmetry (13.1), enhancing mass in 248/540 asymmetries (45.9), enhancing foci in 20/540 asymmetries (3.7) and non-mass enhancement in 201/540 asymmetries (37.2).

In research by Wessam et al. [11] in a cohort similar to ours, they found a prevalence of malignant lesions of 72.8%. They also found that 47.3% of patients with malignant focal asymmetry had mass enhancement on CESM, whereas 23.5% of patients with benign focal asymmetry had no enhancement on CESM (Fig. 9).

Fig. 9
figure 9figure 9

A 49-year-old female patient came with left uniorificial bloody nipple discharge. Pathology revealed mixed IDC and ILC with intermediate DCIS. a Both CC and MLO views of mammography revealed dense breast parenchyma (ACR d) along with left UOQ asymmetry. b Spot image shows the overlying microcalcifications. c The complementary US shows an underlying rather well-defined heterogeneous hypoechoic lesion, measured about 1.1 × 0.8 cm. d AI shows a high abnormality score (96%). e CESM CC and MLO views revealed left UOQ intense heterogenous non-mass enhancement. Both AI and CESM were true positives in this case

In the current study, 34.4% of the lesions were benign, while 37.5% of the lesions were malignant, according to histopathology. The correlation between CESM findings and histopathology was statistically significant. The diagnostic accuracy was 75%, the specificity was 60%, and the sensitivity was 100%.

These findings were consistent with a cross-sectional study conducted by Lobbes et al. [13] including 116 females during the screening program, the results showed CESM sensitivity to be 100%, and specificity to 87.7% (Fig. 10).

Fig. 10
figure 10

A 52-year-older old female patient complaining of right mastalgia, breast redness, and tenderness. Pathology revealed peri-ductal mastitis. a Mammography CC and MLO revealed scattered fibro glandular breast parenchyma (ACR b) along with retro-areolar and para-areolar trabecular thickening and focal asymmetry with overlying areolar skin thickening (blue circles). b Complementary US revealed diffuse right inflammatory changes in the form of skin thickening, echogenic fat lobules, and dilated ducts (5 mm) with internal echogenic non-vascular contents, likely inspissated secretions. c AI revealed a right moderate abnormality score (42%). d CESM revealed Right UOQ para-areolar heterogeneous linear non-mass enhancement. Both AI and CESM were false positives in this case

The earlier review conducted by Lobbes et al. [14] mentioned that the mean sensitivity of CESM was 85.2% (range 62.0–96.0%) and a mean specificity of 66.1% (range 50.0–83.3%). Jochelson et al. [15] have reported a sensitivity of CESM in the detection of malignant breast asymmetry 96% compared to conventional mammography.

A large multi-observer study conducted as part of a Dutch screening program revealed that CESM improved diagnostic accuracy in all readers. The results for all readers utilizing CESM were: sensitivity of 96.9%, specificity of 69.7% [16].

For 129 randomly chosen breast lesions that were annotated by an expert radiologist, a further fivefold cross-validation of CESM pictures was performed. Each annotation includes the BI-RADS descriptors, the biopsy-proven diagnosis of the benign or malignant tumor, and the shape of the mass visible in the image. With 100% sensitivity and 66% specificity [17].

Although several studies discussed the role of AI in the detection of cancer among screening mammograms, few studies addressed its significance in detecting different types of breast asymmetries. Our study showed that AI sensitivity in the assessment of what was beyond the breast asymmetry was 70.8%, specificity was 75% and diagnostic accuracy was 73.4%. Kim et al. [18] reported that AI sensitivity in detecting soft tissue lesions (mass, asymmetry, and distortion) was 89.8% (Fig. 11).

Fig. 11
figure 11

A 52-year-older old female patient came complaining of a right breast lump. Pathology revealed fibro-adenosis with no underlying atypia or malignant cells. a Mammography CC and MLO views revealed heterogeneously dense breast parenchyma (ACR c) along with right UOQ focal asymmetry (blue circles). b Complementary US shows condensed fibro-glandular tissue (at the site of clinical concern). c AI result shows the right UOQ abnormality score of 49%. d) CESM revealed a right UOQ enhancing nodule (blue circle) along with heterogeneous nodular linear non-mass enhancement seen anterior to it, extending toward the retro areolar region to an approximate distance of 2.9 cm from the ipsilateral nipple (blue arrow). Both AI and CESM were false positives in this case

The sensitivity of AI to detect DCIS was 75% (Fig. 12), which was unlike Rafaat et al. [19] who interpreted that the AI system showed 100% sensitivity in the detection of DCIS.

Fig. 12
figure 12

A 45-year-older old female patient with strong positive family history came for screening. Pathology revealed DCIS. a Mammography CC and MLO views revealed heterogeneously dense breast parenchyma (ACR c) along with left LOQ subtle linear focal asymmetry (blue circles). b Spot image shows faint pleomorphic microcalcifications with subtle parenchymal distortion. c Complementary US shows dilated ducts with internal echogenic contents. d AI result shows a low abnormality score. e CESM revealed a left LOQ linear non-mass enhancement reaching (blue circle). AI was a false negative however CESM was a true positive in this case

Another study showed that AI can differentiate between benign and malignant breast lesions with sensitivity and specificity of 87.5% and 91.7%, respectively; they concluded that AI can be used as a tool to assist radiologists in the interpretation of CESM pictures, lowering the number of false positives and reducing biopsies and surgeries [20], which is similar to our conclusions.

In our study, most of the cases were ACR b and c; however, the diagnostic accuracy of AI to detect underlying lesions was reasonable (73.4%) which was similar to that interpreted by Freer [21], who found that the diagnostic performance of AI was less affected by breast density than the performance of radiologists. Radiologists' performance can decrease with dense breasts since dense parenchymal tissue is more likely to mask cancer lesions in mammograms.

We had some limitations in our study such as a small sample size and a short interval of follow-up. We also did not correlate risk factors of breast cancer as prolonged use of hormonal contraception and positive family history in breast cancer patients. We do recommend the conduction of large population-based dataset validation of AI systems used in governmental healthcare facilities, estimation of the reduction rate of tissue biopsies based on the AI system, and establishment of longitudinal cohort studies with a prolonged follow-up period to study the high-risk findings which will require shorter follow-up intervals.


The diagnostic performance of AI-CAD versus CESM revealed that the CESM was more sensitive; however, the AI software was more specific in assessing different breast asymmetries. Although the diagnostic accuracy of both is consistent with each other, all of these findings support the role of AI-aided reading as a safe and valuable diagnostic tool in mammography interpretation which will also improve the efficiency of breast cancer screening programs with subsequently reduced recall of patients and the frequency of unnecessary biopsies as well as short-interval follow-up exams.

We also concluded that AI-aided reading can replace CESM in most cases, especially for those who have contrast sensitivity or impaired kidney function. Also, AI can reduce the radiation exposure hazards of the second dose of radiation for CESM and its financial cost as well.

Availability of data and materials

The corresponding author is responsible for sending the user data and materials upon request.



American college of radiology


Artificial intelligence


Breast imaging-reporting and data system


Computer-aided system




Contrast-enhanced spectral mammography


Digital breast tomosynthesis


Dynamic contrast-enhancing magnetic resonance imaging


Invasive duct carcinoma


Invasive lobular carcinoma


Standard mammography


Medio-lateral oblique


Negative predictive value


Probability of malignancy


Positive predictive value




  1. Chesebro AL, Winkler NS, Birdwell RL, Giess CS (2016) Developing asymmetries at mammography: a multimodality approach to assessment and management. Radiogr Rev Publ Radiol Soci N Am 36(2):322–334.

    Article  Google Scholar 

  2. Geras KJ, Mann RM, Moy L (2019) Artificial intelligence for mammography and digital breast tomosynthesis: current concepts and future perspectives. Radiology 293(2):246–259.

    Article  PubMed  Google Scholar 

  3. Vourtsis A, Berg WA (2019) Breast density implications and supplemental screening. Eur Radiol 29(4):1762–1777.

    Article  PubMed  Google Scholar 

  4. Patel BK, Lobbes MBI, Lewin J (2018) Contrast enhanced spectral mammography: a review. Semin Ultrasound CT MR 39(1):70–79.

    Article  PubMed  Google Scholar 

  5. Perry H, Phillips J, Dialani V, Slanetz PJ, Fein-Zachary VJ, Karimova EJ, Mehta TS (2019) Contrast-enhanced mammography: a systematic guide to interpretation and reporting. Am J Roentgenol 212(1):222–231.

    Article  Google Scholar 

  6. Lei YM, Yin M, Yu MH, Yu J, Zeng SE, Lv WZ, Li J, Ye HR, Cui XW, Dietrich CF (2021) Artificial intelligence in medical imaging of the breast. Front Oncol 11:600557.

    Article  PubMed  PubMed Central  Google Scholar 

  7. Li L, Roth R, Germaine P, Ren S, Lee M, Hunter K, Tinney E, Liao L (2017) Contrast-enhanced spectral mammography (CESM) versus breast magnetic resonance imaging (MRI): a retrospective comparison in 66 breast lesions. Diagn Interv Imaging 98(2):113–123.

    Article  CAS  PubMed  Google Scholar 

  8. Chan YH (2003) Biostatistics 103: qualitative data—tests of independence. Singap Med J 44(10):498–503

    CAS  Google Scholar 

  9. Huang PC, Lin YC, Cheng HY, Juan YH, Lin G, Cheung YC (2020) Performance of stereotactic vacuum-assisted biopsy on breast microcalcifications: comparison of 7-gauge and 10-gauge biopsy needles. J Radiol Sci 45:25–31

    Google Scholar 

  10. Naseem M, Murray J, Hilton JF, Karamchandani J, Muradali D, Faragalla H, Polenz C, Han D, Bell DC, Brezden-Masley C (2015) Mammographic microcalcifications and breast cancer tumorigenesis: a radiologic-pathologic analysis. BMC Cancer 15:307.

    Article  PubMed  PubMed Central  Google Scholar 

  11. Wessam R, Gomaa MMM, Fouad MA, Mokhtar SM, Tohamey YM (2019) Added value of contrast-enhanced mammography in assessment of breast asymmetries. Br J Radiol 92(1098):20180245.

    Article  PubMed  PubMed Central  Google Scholar 

  12. Bassant MD, Abdelmonem ND, Mohamed MH, Rasha MK, Rasha LY (2022) Diagnostic value of contrast-enhanced mammography in the characterization of breast asymmetry. Egypt J Radiol Nucl Med 53:259.

    Article  Google Scholar 

  13. Lobbes MB, Lalji U, Houwers J, Nijssen EC, Nelemans PJ, van Roozendaal L, Smidt ML, Heuts E, Wildberger JE (2014) Contrast-enhanced spectral mammography in patients referred from the breast cancer screening programme. Eur Radiol 24(7):1668–1676.

    Article  PubMed  Google Scholar 

  14. Lobbes MB, Smidt ML, Houwers J, Tjan-Heijnen VC, Wildberger JE (2013) Contrast enhanced mammography: techniques, current results, and potential indications. Clin Radiol 68(9):935–944.

    Article  CAS  PubMed  Google Scholar 

  15. Jochelson MS, Dershaw DD, Sung JS, Heerdt AS, Thornton C, Moskowitz CS, Ferrara J, Morris EA (2013) Bilateral contrast-enhanced dual-energy digital mammography: feasibility and comparison with conventional digital mammography and MR imaging in women with known breast carcinoma. Radiology 266(3):743–751.

    Article  PubMed  Google Scholar 

  16. Lalji UC, Houben IP, Prevos R, Gommers S, van Goethem M, Vanwetswinkel S, Pijnappel R, Steeman R, Frotscher C, Mok W, Nelemans P, Smidt ML, Beets-Tan RG, Wildberger JE, Lobbes MB (2016) Contrast-enhanced spectral mammography in recalls from the Dutch breast cancer screening program: validation of results in a large multireader, multicase study. Eur Radiol 26(12):4371–4379.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  17. Perek S, Kiryati N, Zimmerman-Moreno G, Sklair-Levy M, Konen E, Mayer A (2019) Classification of contrast-enhanced spectral mammography (CESM) images. Int J Comput Assist Radiol Surg 14(2):249–257.

    Article  PubMed  Google Scholar 

  18. Kim HE, Kim HH, Han BK, Kim KH, Han K, Nam H, Kim EK (2020) Changes in cancer detection and false-positive recall in mammography using artificial intelligence: a retrospective, multireader study. Lancet Digit Health 2(3):e138–e148.

    Article  PubMed  Google Scholar 

  19. Raafat M, Mansour S, Kamal R, Ali HW, Shibel PE, Marey A, Taha SN, Alkalaawy B (2022) Does artificial intelligence aid in the detection of different types of breast cancer? Egypt J Radiol Nucl Med 53:182.

    Article  Google Scholar 

  20. Fanizzi A, Losurdo L, Basile TMA, Bellotti R, Bottigli U, Delogu P, Diacono D, Didonna V, Fausto A, Lombardi A, Lorusso V, Massafra R, Tangaro S, La Forgia D (2019) Fully automated support system for diagnosis of breast cancer in contrast-enhanced spectral mammography images. J Clin Med 8(6):891.

    Article  PubMed  PubMed Central  Google Scholar 

  21. Freer PE (2015) Mammographic breast density: impact on breast cancer risk and implications for screening. Radiographics 35(2):302–315.

    Article  PubMed  Google Scholar 

Download references


We would like to acknowledge Prof. Dr. Rasha Kamal who always supports the research works at our unit, the radiology department, at Cairo University. And we would like to acknowledge Prof. Dr. Soha Talaat, head of the women imaging unit at the radiology department, at Cairo University for great help.


No source of funding.

Author information

Authors and Affiliations



SA is the guarantor of the integrity of the entire study. AE and SA contributed to the study concepts and design. AN, SA, and AE contributed to the literature research. SA and AN contributed to the clinical studies. All authors contributed to the experimental studies/data analysis. AN and AN contributed to the statistical analysis. SA contributed to the manuscript preparation. SA and AE contributed to the manuscript editing. All authors have read and approved the final manuscript.

Corresponding author

Correspondence to Aalaa Sobhi.

Ethics declarations

Ethics approval and consent to participate

The study was approved by the ethical committee of the Radiology Department of Kasr –Al-Ainy Hospital, Cairo University which is an academic governmental supported highly specialized multidisciplinary hospital. The included patients gave written informed consent.

Consent for publication

All patients included in this research were legible and above 16 years of age. They gave written informed consent to publish the data contained within this study.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Sobhi, A., Mohammed, N.A. & Ali, E.A. What’s beyond breast asymmetry? Comparative study between artificial intelligence and contrast-enhanced spectral mammography in the assessment of various types of breast asymmetries. Egypt J Radiol Nucl Med 54, 107 (2023).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: