MRI is a well-established modality in breast imaging; moreover, at present, the DCE-MRI is considered the gold standard technique in patients with breast lesions, with a reported sensitivity of 95–99% in detecting invasive cancer and of 80% in detecting ductal carcinoma in situ. It allows lesions assessment morphologically as well as semi-quantitatively by the enhancement kinetics [12,13,14,15,16,17]. So, absence of the peculiar features of malignancy in the enhanced sequence may alleviate the need of unnecessary breast biopsy [18]. However, the DCE-MRI can be time consuming, costly, or unavailable, plus in some inevitable circumstances, contrast administration can be a concern, undesired, or even contraindicated [10, 19]. Hence, with the evolution of new sequences like DWI, initiatives in setting up new protocols, alternative to DCE-MRI, has emerged. One of these protocols is the integration of DWI with T2WI and STIR images; this way, morphological information provided by the high spatial resolution of the T2WI is incorporated with the functional data offered by DWI.
Our study aimed to evaluate the role of UE-MRI of the breast in detecting and characterizing breast lesions compared to DCE-MRI of the same breast lesions. We found out that UE-MRI has close results to DCE-MRI with strong interobserver agreement in lesion characterization.
Although all lesions were depicted in both of DCE-MRI and UE-MRI in our study, we noted that the former revealed the lesions slightly more readily and confidently than the UE-MRI. Additionally, we found that the subtly visible lesions in DWI were more of benign nature than being malignant, and this could be assigned to the fact that most malignant pathologies caused diffusion restriction and so were relatively easily spotted than the isointense to hypointense benign ones. Other reasons for this could be field in-homogeneity or small and non-mass lesions. However, careful inspection for associated abnormalities like skin changes and lymphadenopathies, plus reviewing the other UE-MRI images (STIR sequence in particular) side by side the DWI, assisted in raising the overall confidence about lesion localization and overcame the possibility of the lesion being overlooked in DWI. Some studies had also reported that DWI had a relatively reduced diagnostic performance in lesions measuring 1 cm or less leading to more false-negative results [20].
For lesion characterization using the ADC values, although we found some overlapping between benign and malignant lesions, the mean average of the benign ones was significantly higher than malignant ones, and thus adding this parameter was profitable in increasing the diagnostic accuracy of UE-MRI in our study. Different studies had set different cutoff ADC values depending on their scanning parameters; nevertheless, a cutoff value of 1.30 × 10− 3 mm2/s for ADC had yielded 89.1% sensitivity and 100% specificity for the differentiation between benign and malignant lesions [8]. Some studies had questioned the value of DWI and ADC and discovered that it is a beneficial tool in lesion characterization and even in treatment monitoring and planning. ADC values showed strong correlation with lesion histology being notably lower in malignant lesions; also, the highest values were recorded in human epidermal growth factor receptor (HER2)-enriched tumors, while low ADC had suggested progesterone receptor (PR) negativity [11, 21]. Additionally, we found that lesion signal intensity in T2WI was a helpful indicative of its nature, as most of malignant lesions were hypointense in T2WI, opposing most benign lesions which displayed high signal intensity. However, if the malignant lesion had cystic necrosis, hemorrhagic changes, fatty component, mucinous, myxoid, or edematous stroma, it could display high T2 signal intensity mimicking benign pathologies [22]. Moreover, if a benign lesion was abundant with fibrosis or inflammatory changes, e.g., fibrocystic changes, epithelial hyperplasia, granulomatous inflammations, papilloma, fibroadenoma, and sclerosing adenosis, it might exhibit T2 hypointensity and low ADC values simulating radiologically the malignant lesion [23]. Special attention should also be paid to the suppressed lesions in STIR sequence, knowing this sequence is not fat selective and the signal suppression counts on T1 relaxation of tissue. Subsequently, differentiating hemorrhagic versus fatty lesions on STIR images solely could be problematic [24]. Our study also confirmed the established malignancy features namely the ill-defined borders. Yet, counting solely on this manifestation could be misleading, as there was overlapping between benign and malignant lesions, i.e., 25% of benign lesions had ill-defined margins while 22% of pathologically proven malignant lesions had well-defined margin radiologically (Figs. 4 and 5). It had also been concluded by others that this sign was statistically significant but still unreliable in differentiating between benign and malignant pathologies [25].
Our results were concordant with multiple studies. Belli et al. [20] had evaluated the DWI and STIR in comparison to lesion histopathology by two observers. For rater 1, STIR and DWI specificity was 99.3% and 95.7%; for rater 2, it was 99.3% and 96.4%. STIR and DWI sensitivity was 76.5% and 76.5% for rater 1, and 77.5% and 77.6% for rater 2. They had reported that UE-MRI had overall good diagnostic performance especially with lesions ≥ 1 cm in size, regardless its location with very good interobserver agreement for STIR and DWI. They concluded that UE-MRI can be employed where contrast injection is contraindicated. Another similar study was done by Telegrafo et al. [26]. They compared the UE-MRI—including the T2, STIR WI, and DWI with background body signal suppression (DWIBS)—with the DCE-MRI and the histological findings. They reported that breast UE-MRI is an effective alternative tool to CE-MRI to evaluate breast lesions with no statistically significant difference between them. They also mentioned that STIR and DWIBS were beneficial in lesion detection while T2WI and ADC values in lesion characterization. Their calculated sensitivity, specificity, diagnostic accuracy, and PPV and NPV values for the UE-MRI was 94%, 79%, 86%, 79%, and 94%, respectively; for DCE-MRI, it was 98%, 83%, 90%, 84%, and 98%, respectively. In another research, Kul et al. [27] found that UE-MRI (combining T2 and DWI)—instead of immediate biopsy—may lower the negative biopsy rates in low-risk breast masses (namely B-RADS 3 and 4A). They reviewed UE-MRI images and categorized the lesions into benign and malignant then compared that into the histology results. They found out that UE-MRI had 91% specificity and 99% NPV for detection of breast cancer. A further research was carried out by Baltzer et al. [25]. Two observers assessed the UE-MRI diagnostic performance in comparison to DCE-MRI. It was revealed that there was no significant difference between the two methods and observers. They also found that lesion visibility was less in UE-MRI resulting in more false results, but lesion size measurement was very close in all sequences. Trimboli et al. [28] also evaluated the combined STIR, T2WI, and DWI sequences in comparison to pathology and negative follow-up by two blinded radiologists. They deduced that UE-MRI protocol had sensitivity and specificity of 76–78% and 90% in cancer detection respectively with nearly matched interreader agreement. Furthermore, Moschetta et al. [29] reported that T2WI and STIR sequences can be a valid tool for spotting occult inflammatory breast cancer. In comparison to biopsies results, these sequences’ calculated sensitivity, specificity, diagnostic accuracy, and PPV and NPV values was 86%, 100%, 96%, 100%, and 94%.
The limitation of our study is that the malignant encountered lesions were nearly double the benign ones. This could be attributed to our inclusive criteria, by which we were adherent; the included lesions must have their histopathology results available. Since evidently benign lesions would not be biopsied, subsequently fewer benign ones had been included.