Diagnostic performance of tomosynthesis, digital mammography and a dedicated digital specimen radiography system versus pathological assessment of excised breast lesions

Breast conserving surgery (BCS) is an established treatment modality for early breast cancer, offering better aesthetic results and less morbidity, without compromising survival, compared with radical mastectomy.^1,2

A clear resection margin after surgical excision is associated with a reduced risk of local recurrence. Positive margins are associated with a 2-fold increased risk of ipsilateral recurrence.³ This risk is not eliminated by radiotherapy, systemic chemotherapy or endocrine therapy. Therefore, to achieve the best local disease-free survival, a negative margin must be achieved during surgery and confirmed by the final microscopic assessment of the excised tissue.

The published reoperation rates for patients with early stage breast cancer vary considerably, between 10% and 60%, depending on the treating centre and the surgeon’s practice, with an average rate of ~20%.⁴ To reduce the rate of reoperation in patients with non-palpable lesions, intraoperative assessment is performed to confirm adequate removal of the detected lesions and margin. The resected sample is imaged, most commonly by digital mammography, while the patient is under general anaesthesia. The images are analysed and information about the resection margins is given to the surgeon. Proper assessment of the margins reduces the need for reoperation, the cost of hospital stay and the subsequent psychological or cosmetic impact on the patient. Failure to achieve negative margins usually results in re-excision or mastectomy.⁵

Digital breast tomosynthesis eliminates tissue superimposition and provides a clearer view of dense breasts because it provides three-dimensional (3D) images unlike mammography, which provides two-dimensional (2D) images.⁶ Specimen tomosynthesis was previously shown to be superior to digital mammography for depicting excised lesions and evaluation of resection margins.^{7,8,9,10,11,12,13} Mobile, dedicated digital specimen radiography is a rapid method for intraoperative specimen assessment that can image the excised specimen in the operating room, thus avoiding the need to send the specimen to the radiology department. This shortens the duration of anaesthesia, reducing morbidity and mortality, and decreases operating room occupancy, thus reducing costs, while still providing comparable results to digital mammography.^14,15,16

The aim of this study was to directly compare the diagnostic performance of full-field digital mammography (FFDM), digital breast tomosynthesis and a dedicated digital specimen radiography system (SRS) for the evaluation of resected breast lesions in consecutive patients, and to compare the margin status of resected lesions versus the final pathological report.

This study was undertaken as part of the continuous improvement, quality control and internal validation of modern surgical specimen imaging technologies at the Breast Unit at Kuopio University Hospital (Kuopio, Finland). During the study period, clinical decisions concerning lesion removal and margin status were made by experienced breast radiologists and surgeons and based on all available images.

The detailed analyses and intertechique comparisons described here were performed retrospectively and did not affect patient management. The Chair of the hospital district waived the need to obtain written informed consent from the patients owing to the retrospective nature of the analyses (Approval: FinMargins 5063573; 508/2021). All clinical investigations were conducted according to the relevant guidelines and the principles expressed in the Declaration of Helsinki.

Lesion localisation and surgery

The surgical procedure was planned individually according to the patient’s preference, tumour size, tumour location, clinical findings, breast shape and breast size. All patients were evaluated at multidisciplinary meetings, at least twice, pre- and postoperatively.

Non-palpable tumours were localised pre-operatively using at least one guidewire (breast localization needle Duo, SOMATEX® Medical Technologies GmbH) under ultrasound or stereotactic guidance. Two-view mammography (CC and lateromedial) was routinely performed to confirm the position of each lesion relative to the guidewire, and the location of the lesion was ink-marked on the skin, including supine MRI-guided localisation projections.¹⁸ Tumours were excised en bloc from the subcutaneous area to the muscle, and the overlying skin was removed in patients with superficial lesions to achieve a healthy macroscopic surgical margin of ≥1 cm, in accordance with national guidelines, and hence achieve microscopically negative margins (defined as no “tumour-on-ink”). In ductal carcinoma in-situ (DCIS), the need for reoperation is evaluated in a multidisciplinary meeting whenever DCIS margins are less than 2 mm. Intraoperatively, the specimens were placed on a Styrofoam slab, fixed with wooden sticks, and the location of the excision was anatomically marked. Metallic clips were placed directly on the specimen to indicate the orientation. The fascia posterior to the tumour was removed and fixed aside if removed separately (Figure 1,2). The specimen was then placed in a plastic container and immediately transported to the Breast Radiology Unit.

Figure 1

Figure 2

The final analyses included 204 patients with a mean age of 62.5 years (range 33–95 years). The patient characteristics, histological diagnosis and surgical procedures are presented in Table 1. The majority of lesions were treated by wide local excision (158/216, 73.1%), and oncoplastic conservative breast resection was performed for more than a quarter of lesions (58/216, 26.9%). The mammographic features of the lesions are presented in Table 2.

Regarding interobserver agreement, the ICC was high for lesion visibility (0.787) and diagnostic certainty (0.684) with tomosynthesis. Similar results were observed for SRS and FFDM, with ICCs of 0.742 and 0.804 for lesion visibility and 0.671 and 0.683 for diagnostic certainty, respectively.

As shown in Table 3, both observers felt that tomosynthesis was the best performing imaging modality in 76.9% of cases (Figure 2,3). SRS was the least favoured method, chosen only once (0.5%) by observer 1 and six times by observer 2 (2.8%).

Figure 3

Table 4 presents the results reported by both observers. For both observers, tomosynthesis provided significantly better lesion visibility than SRS and FFDM, which translated into a significantly greater diagnostic certainty. Moreover, tomosynthesis was superior to the other two methods for identifying spiculations and calcifications. The high performance of tomosynthesis was not affected by peritumoral density (p = 0.851).

All three methods showed comparable results for estimating the diameters of the excised lesions (Table 5). Observer 1, who was more experienced, estimated the lesion sizes with greater accuracy relative to the final pathological report than observer 2, who was less experienced and tended to overestimate the lesion sizes using all three methods. The Pearson’s correlation coefficient for tomosynthesis relative to the final pathology was greater than those for SRS and FFDM. The Bland-Altman plots constructed using the lesion diameters measured by the three imaging methods relative to the final pathological report are shown in Figure 4 for both observers. The plots illustrate the greater accuracy of observer 1 compared with the less-experienced observer 2, as well as the superior agreement of tomosynthesis to the final pathological report. The plots also indicate that the differences in measurements increase with increasing lesion diameter for each imaging modality.

Figure 4

Of 204 patients included in this study, only one underwent reoperation owing to inadequate margins. Small invasive ductal carcinoma foci were associated with a 6.5 × 5.5 cm DCIS close to three margins. Upon re-resection, a 2.5 cm residual grade 3 DCIS was found. Intraoperative findings revealed multicentric disease in one patient that was confirmed by pathological assessment of frozen section, and mastectomy was performed. The margins of that specimen were found to be macroscopically and microscopically adequate. One patient with papilloma had a small, microscopic extension to the edge of the specimen. In addition, nine cancers (one DCIS, one invasive lobular and seven invasive ductal cancers) that were not visible on preoperative imaging were found on the marginal resection specimens removed by the surgeons for cosmetic or additional marginal purposes; all nine had negative margins.

Intraoperative assessment of resected specimens is particularly important for successful BCS. To the best of our knowledge, this study is the first to compare the performance of three imaging modalities performed simultaneously in the same patients. Of the three modalities, digital tomosynthesis was superior for visualising lesions, calcifications and spiculations in the majority of specimens, and therefore provided the best confirmation of complete removal of the target lesion. Moreover, both observers, with different levels of experience, felt that tomosynthesis was the superior imaging modality for the majority of cases and reported greater certainty of diagnosis compared with the use of FFDM and SRS.

The cross-sectional capability of tomosynthesis reduces the effect of breast tissue superimposition and therefore helps to delineate the tumour margins. In this study, tomosynthesis was superior to the other imaging modalities, regardless of the fact that the majority of our patients had fatty breasts (BI-RADS density A or B) and approximately two-thirds of the lesions had low peritumoral tissue densities. This may reflect the better image quality of tomosynthesis compared with 2D techniques. Our results are in concordance with those of prior studies.^12,13,19

The underperformance of SRS is noticeable because it did not visualise one in five lesions. Prior studies showed that using mobile SRS in an operating theatre reduced the duration and cost of surgery significantly^9,14,15,16, at the expense of inferior image quality.¹⁴ However, a mobile SRS equipped with tomosynthesis exhibited greater accuracy than standard mammography and reduced the rate of re-excision.¹¹

The reoperation rate after BCS varied markedly in earlier studies. Tumour-related factors that may influence decisions regarding reoperation include focality, presence of DCIS and tumour size⁴, and non-tumour-related factors include inadequate assessment of the extent of macroscopic disease at diagnosis, inaccurate impalpable disease localisation and limited use of intraoperative specimen radiography.^4,5 BCS is not specifically limited based on cut-off values for tumour size; instead, surgeons should balance their decision between the assessed tumour size and the total breast volume. More intraoperative tissue sampling − such as shaving the resection margins after lumpectomy in certain situations or oncoplastic BCS techniques − may help reduce the rate of reoperation.^5,20 In this study, the reoperation rate was low, which is presumably multifactorial. All preoperative imaging findings were re-evaluated by specialist breast radiologists and multidisciplinary specialists. Every effort was made to evaluate the tumour extent preoperatively and to transfer these findings to the surgical position by ink-marking the skin. Furthermore, all procedures were standardised, including tumour localisation, macroscopic resection margins of ≥1 cm, en bloc resection for anterior and posterior margins, structured specimen orientation, specimen fixation, imaging and structured histopathological reporting. All of these factors might help to reduce positive margins.⁴

The efficacy of specimen mammography for margin assessment is not yet well established. Laws et al. reported that the use of any margin assessment technique did not improve margin status compared with guidewire localisation alone.²¹ According to a meta-analysis, specimen radiography to assess the surgical margin had lower sensitivity than frozen sections (53% vs. 86%, respectively).²² However, frozen section is a resource-demanding procedure, is not always readily available and it might prolong the duration and cost of surgery. By contrast, intraoperative imaging is simple, rapid and readily available. In our opinion, its lower accuracy is due mainly to the inability to comprehensively evaluate the microscopic tumour extension from the target lesion. Mammography tends to underestimate the size of DCIS and, although the sensitivity of specimen imaging is higher for invasive cancers, it is lower for DCIS.²³ A greater resection margin threshold may reduce the risk of missing a positive margin but increases unnecessary resection of healthy tissue.²³ Mazouni et al. determined the sensitivity and specificity of different radiological threshold values (1, 5 and 10 mm), and found that the 10 mm threshold value had the highest sensitivity (75%).²⁴ Britton et al. reported that a maximum distance of ≥11 mm from the lesion to the specimen edge was associated with a 77% likelihood of having a clear final histological margin.²⁵ Leung et al. reported that a 15 mm radiological margin showed the highest combination of sensitivity and specificity for predicting a positive margin.²⁶ DCIS is often associated with invasive cancer and, in this study, half of the specimens included DCIS components. Therefore, the consistent wide macroscopic resection margins in this study presumably contributed to the microscopically clear margins.

It is difficult to directly compare the results of studies assessing the clinical value of specimen radiography because of marked heterogeneity in the study designs and inclusion criteria, as well as the methodology and terminology used, thus making comparisons inconclusive.²⁷ Inconsistencies may be due to different imaging protocols, specimen compression, selective inclusion of patients with different stages of cancer, inclusion of mainly invasive or DCIS patients and the definitions of the outcome measures.^27,28 In this study, we sought to include all consecutive patients treated at our institution. We excluded only those patients who underwent BCS after neoadjuvant therapy because some of these patients only have residual microscopic disease or marking clips. Therefore, the patient population in this study closely represents clinical practice at a specialist tertiary hospital.

This study has limitations to consider. This was a single-centre study and the analysis was performed retrospectively. The observers had varying years of experience, which may contribute to the interobserver variability and is consistent with a previous report.²⁹ Furthermore, we could not perform more extensive analysis of the diagnostic accuracy of each method in the evaluation of margins owing to complete primary resection of the lesions without positive margins for malignant lesions. Regardless of these limitations, we analysed a cohort of consecutive patients, which is consistent with and representative of actual clinical practice. Moreover, we included 216 specimens from 204 patients, a considerably larger cohort than most of the related studies reported to date. We also performed three different imaging modalities for each specimen, which allowed us to directly compare radiographs obtained in the same orientation for all imaging methods and thus remove some potential sources of error.

In conclusion, tomosynthesis was superior to SRS and FFDM for the detection and evaluation of target breast lesions, and detected spiculations and calcifications. Therefore, tomosynthesis was more reliable than other intraoperative imaging modalities for evaluating complete excision of breast lesions.