Breast-Conserving Surgery Margin Guidance Using Micro-Computed Tomography: Challenges When Imaging Radiodense Resection Specimens

Abstract

Background.

Breast-conserving surgery (BCS) is an integral component of early-stage breast cancer treatment, but costly reexcision procedures are common due to the high prevalence of cancer-positive margins on primary resections. A need exists to develop and evaluate improved methods of margin assessment to detect positive margins intraoperatively.

Methods.

A prospective trial was conducted through which micro-computed tomography (micro-CT) with radiological interpretation by three independent readers was evaluated for BCS margin assessment. Results were compared to standard-of-care intraoperative margin assessment (i.e., specimen palpation and radiography [abbreviated SIA]) for detecting cancer-positive margins.

Results.

Six hundred margins from 100 patients were analyzed. Twenty-one margins in 14 patients were pathologically positive. On analysis at the specimen-level, SIA yielded a sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of 42.9%, 76.7%, 23.1%, and 89.2%, respectively. SIA correctly identified six of 14 margin-positive cases with a 23.5% false positive rate (FPR). Micro-CT readers achieved sensitivity, specificity, PPV, and NPV ranges of 35.7–50.0%, 55.8–68.6%, 15.6–15.8%, and 86.8–87.3%, respectively. Micro-CT readers correctly identified five to seven of 14 margin-positive cases with an FPR range of 31.4–44.2%. If micro-CT scanning had been combined with SIA, up to three additional margin-positive specimens would have been identified.

Discussion.

Micro-CT identified a similar proportion of margin-positive cases as standard specimen palpation and radiography, but due to difficulty distinguishing between radiodense fibroglandular tissue and cancer, resulted in a higher proportion of false positive margin assessments.

Breast cancer is the second-leading cause of cancer death among women in the United States.¹ For early-stage cancers, breast-conserving surgery (BCS) combined with radiation therapy is the most common course of treatment.² BCS involves local excision of breast cancer with a surrounding margin of healthy tissue, while preserving the shape and appearance of the breast. Achieving margins free of cancer is important, because a twofold increase in ipsilateral breast cancer recurrence is associated with positive margins.^3–10 The success of BCS hinges on the ability of the surgical team to assess resection margins thoroughly and rapidly during the initial procedure to avoid costly reexcision procedures that are associated with increased patient distress, worse cosmetic outcomes,^11,12 and increased medical costs.^13–18

The overall reexcision rate due to positive margins is 17–19% in the United States based on recent, large studies that encompass thousands of BCS patients.^19,20 Furthermore, Chakedis et al. tracked the BCS reexcision rates of 55 surgeons between 2016 and 2020 at one integrated healthcare consortium and found that the individual surgeon reexcision rate varied from 7.8% to 36.8%, demonstrating substantial intersurgeon variability.²⁰ A need exists for improving intraoperative margin assessment, such that positive margin rates, and thus reexcision procedures, become less frequent and the high variability in performance between providers is reduced.

Current methods for intraoperative margin assessment in BCS include visual inspection and palpation of the resected tissue by the surgeon, meticulous recording of resection anatomical orientation using surgical ink and/or sutures, projection specimen radiography, and at few high volume medical centers, intraoperative pathological processing.^21–23 Although intraoperative pathological processing provides high diagnostic accuracy (sensitivities of 73–91%),^24–27 these methods are resource-intensive and time-consuming (~30 min per case).^24,26 The surgical technique of routine circumferential cavity shaving reduces reexcision rates substantially (rates of 8–10% with versus 21–24% without routine cavity shaving),^28–30 but the technique necessitates the removal of increased volumes of tissue, raising cosmetic concerns,²⁶ and adds to procedural costs.²⁹ X-ray-based projection (i.e., two-dimensional), specimen radiography is a mainstay for intraoperative margin assessment. However, overlapping radiodensities of tumor and intervening breast tissues obscure lesion borders and therefore limit assessment of the entire margin. (Typically, specimen radiography is used to visualize the coronal plane of the resected tissue. In this orientation, the anterior and posterior margins are not clearly resolved.) While specimen radiography is useful for verifying the removal of the primary lesion, some studies suggest it is ineffective for margin assessment specifically,^31–33 highlighting the need for other intraoperative margin assessment techniques.

A multitude of novel technologies are under investigation for improved margin assessment during BCS, including intraoperative ultrasonography, radiofrequency spectroscopy, bioimpedance spectroscopy, microresolution computed tomography (micro-CT), optical coherence tomography, ex vivo magnetic resonance imaging, photoacoustic microscopy, and fluorescence imaging.³⁴ These technologies are at various stages of preclinical and clinical evaluation. One or more of these techniques may be adopted in BCS for improved margin assessment.

Several recent studies have evaluated micro-CT for intraoperative margin assessment during BCS, given that the volumetric, X-ray-based imaging modality can resolve the entire margin of a resected specimen. These studies have reported a wide range of specimen-level performances for detecting positive margins: sensitivities of 38–93%, specificities of 53–100%, positive predictive values (PPVs) of 22–100%, and negative predictive values (NPVs) of 79–90%, with sample sizes ranging from 30 to 173 specimens.^35–39 Additionally, a recent, 200-patient, micro-CT imaging study by Kulkarni et al. reported margin-level sensitivities of 91–94%, specificities of 81–85%, PPVs of 25–30%, and NPVs of 99–99%.⁴⁰ Micro-CT provides rapid, volumetric scanning of an entire BCS resection with reconstruction resolutions typically within 45–150 μm³ voxels.^35,38 The objective of the present study was to evaluate the performance of micro-CT as a tool for detecting positive margins during initial BCS procedures performed at an outpatient surgery clinic where only margin palpation and specimen radiography are used for intraoperative margin assessment, and targeted cavity shaves are only performed if specimen palpation and/or radiography are deemed positive by the surgical team. The principal contribution of this study to the existing literature is highlighting the limitations of micro-CT when imaging radiodense resection specimens.

METHODS

Study Design

The study was a prospective, single-armed, observational trial that enrolled 117 patients. It was conducted at the Dartmouth Heath Outpatient Surgery Center in Lebanon, New Hampshire, between February 2020 and January 2022 and was approved by the Dartmouth Health Institutional Review Board. Each enrolled patient signed an informed consent form. Inclusion criteria were women aged 18 years or older, histologic diagnosis of invasive breast cancer based on pre-surgical core biopsy, and the desire for BCS. Patients with an expected BCS specimen size greater than the specimen holder (i.e., >10 cm × 10 cm × 5 cm) were excluded from the study. The imaging system was stationed in the surgical suite within 100 feet of the operating room (OR).

Each BCS was performed by a single surgical oncologist with more than 30 years of experience (RJB). All eligible patients who presented to the study surgical oncologist between the dates of February 7, 2020 and January 12, 2022 were approached about the study, and 66% (117/178) of eligible patients consented to study enrollment. Immediately after completing standard-of-care tissue resection, the surgeon palpated the six anatomical margins of the specimen (i.e., superficial, deep, cranial, caudal, medial, and lateral), assessing each tissue edge for potential cancer involvement. The whole, uninked specimen was placed in a specimen holder, oriented to correspond with labels on the holder plates, taken from the OR, and imaged in the surgical suite by a nonclinical member of the study team. One micro-CT scan was performed, generating a volumetric map of X-ray attenuation. After imaging, the specimen was returned to the OR, and the six margins of the specimen were inked with different colors by the surgeon. The specimen was then placed on a specimen tray with a radiosensitive, alphanumeric grid, and a single two-dimensional projection specimen radiograph was taken, according to standard-of-care. Each specimen radiograph was interpreted jointly by the surgeon and by a breast radiologist who called into the OR. Consensus between the surgeon and radiologist led to targeted cavity shaves if one or more margins were suspected of cancer involvement. The tissue was then transported to the Gross Pathology Laboratory for standard histopathological processing. Micro-CT scanning (including transport time) took <10 min per case. All clinical staff members were blinded to the micro-CT imaging results; micro-CT imaging did not play a role in surgical decision making.

The micro-CT imaging system used in this study was an upright In Vivo Imaging System SpectrumCT (PerkinElmer, Hopkington, MA). Micro-CT settings reported previously and known to provide reasonable contrast, high-resolution, and rapid scanning were used (i.e., 50 kVp, 1 mA, 720 projections, 100 ms/exposure, 150 μm³ voxels, less than 4 minutes for acquisition and reconstruction).³⁵ Each BCS resection was placed in a specimen holder comprised of two optically clear acrlyic plates with anatomical labels to maintain tissue orientation, and the plates were held together by rubber clamps.

Board-certified breast radiologists, blinded to the pathological margin analysis, independently interpreted each specimen micro-CT scan (three readers with an average of 27 years of experience: RAZ, RMDA, MTH). Each radiological read was performed postoperatively and took approximately five minutes. Readers were allowed to access patient history, preoperative images, and preoperative biopsy data. For every margin in the micro-CT scan, a status was recorded (i.e., negative or positive for tumor on the margin).

Micro-CT radiological interpretation was performed using 3D Slicer (v4.11),⁴¹ a free and open-source medical image analysis software program. Each radiologist was trained to use the software interface before the study. Each scan was visualized with consistent contrast and with three orthogonal planes displayed simultaneously. Scans were analyzed based on tissue radiodensity (i.e., image intensity) and macroscale morphology (i.e., structural features). Final margin statuses, based on (1) standard-of-care intraoperative margin assessment (i.e., specimen palpation and radiography [abbreviated SIA]), and (2) individual radiologists’ readings of the specimen micro-CT, were compared to those provided in final pathology reports confirmed by a board-certified, breast pathologist with more than 30 years of experience (WAW). Pathologically, a positive margin was defined as “tumor on ink,” according to Society of Surgical Oncology–American Society for Radiation Oncology consensus guidelines for invasive breast cancer.¹⁰

Analysis was only performed on primary lumpectomy specimens. Targeted cavity shaves were guided only by SIA. If a specimen (plus any cavity shaves) yielded one or more positive margins as determined by histopathology, the patient underwent further breast surgery. Imaging of cavity shaves and reexcision specimens was beyond the scope of this study.

Statistical Analysis

All statistical analyses were performed using MATLAB (v2021a, Mathworks, Natick, MA, USA). The primary study endpoint was the ability of a trained breast radiologist to assess the presence of malignant tissue on the margins of primary BCS resections using micro-CT, relative to “gold standard” histopathological examination. Performance metrics of sensitivity, specificity, PPV, and NPV with 95% Clopper–Pearson confidence intervals, quantified on the margin-level (n = 600) and on the specimen-level (n = 100), were reported for each intraoperative margin assessment method. Fleiss’ kappa was used to evaluate inter-reader agreement between the three sets of micro-CT readings.

For specimen-level performance, the following rules applied: first, a true negative reading required that a technique correctly identified all margins as negative; second, a false positive reading resulted if all margins were pathologically negative, but a technique erroneously reported ≥1 positive margins; third, a true positive reading resulted if ≥1 margins were pathologically positive, and a technique correctly identified ≥1 involved margins as positive; and fourth, a false negative reading resulted if ≥1 pathologically positive margins were missed by a technique.

RESULTS

Analysis was performed on 100 primary BCS resections from 100 of 117 enrolled patients (Figure 1). All six anatomical margins were analyzed on each specimen. Thus, 600 margin-level measurements and 100 specimen-level measurements were made. Demographic data are summarized in Table 1. The mean patient age was 62.9 (range 38–91) years, and 93% of patients were of white/non-Hispanic ethnicity (representative of the Dartmouth Health catchment area). Twenty-five patients (25%) had invasive ductal carcinoma (IDC), 51 patients (51%) had IDC plus ductal carcinoma in situ (DCIS), 22 patients (22%) had invasive lobular carcinoma (ILC), and two patients (2%) had ILC plus pleomorphic lobular carcinoma in situ (LCIS). Seventy-six percent of resections utilized a wire for tumor localization, whereas the remainder of resections relied on tumor palpation.

FIG. 1.

Enrolled patients

TABLE 1.

Demographic data

Parameter	Value
Age (yr), n (%)
Mean (STD)	62.9 (10.9)
<40	1 (1.0)
40–50	10 (10.0)
50–60	25 (25.0)
60–70	40 (40.0)
>70	24 (24.0)
Race/ethnicity, n (%)
White/non-Hispanic	93 (93.0)
White/Hispanic or Latina	2 (2.0)
White/unknown	3 (3.0)
American Indian or Alaska native	1 (1.0)
Unknown	1 (1.0)
Body mass index (kg/m2), n (%)
Mean (STD)	28.8 (6.3)
18.5–25	35 (35.0)
25–30	27 (27.0)
>30	38 (38.0)
Wire-localized lumpectomy, n (%)
No	24 (24.0)
Yes	76 (76.0)
Tumor type, n (%)
IDC	25 (25.0)
IDC + DCIS	51 (51.0)
ILC	22 (22.0)
ILC + pleomorphic LCIS	2 (2.0)
Pathology largest tumor diameter (cm)
Mean (STD)	1.8 (1.5)
Minimum	0.2
Maximum	9.0
Pathology report specimen volume (cm3)*
Mean (STD)	111 (65)
Minimum	21
Maximum	371
Pathology report specimen weight (g)
Mean (STD)	57 (31)
Minimum	12
Maximum	207

IDC invasive ductal carcinoma; DCIS ductal carcinoma in situ; ILC invasive lobular carcinoma; LCIS lobular carcinoma in situ; STD one standard deviation

^*Standard-of-care pathological processing involved approximating the height, width, and depth of each resection with a ruler. Resection volume was approximated using the ruler measurements and assuming a cuboid geometry

BCS resection characteristics are summarized in Table 2. Of the 600 margins assessed in the trial, 3.5% (21/600) were confirmed positive for cancer by histopathology. Positive margins were mainly due to high-grade IDC (28.6%, 6/21) and intermediate-grade ILC (23.8%, 5/21) lesions. The anatomical margins responsible for the greatest share of positives were deep and cranial, each being responsible for 23.8% (5/21) of positive margins recorded. Fourteen primary resections (14% of all cases) contained at least one positive margin per histopathology. Positive margins were found in 11.8% (9/76) of ductal carcinoma (IDC ± DCIS) cases and 20.8% (5/24) of lobular carcinoma (ILC ± LCIS) cases. Positive margins were found in 13.2% (10/76) of nonpalpable, wire-localized cases and 16.7% (4/24) of palpable cases. SIA assigned one or more margins as positive in 26 patients, all leading to targeted cavity shaves during initial procedures. In three of the 14 margin-positive cases confirmed by histopathology, the primary specimen margin was positive, but the targeted cavity shave(s) taken resulted in a negative margin. Thus, 11 patients (11%) had positive margins after their initial surgery and underwent further breast surgery (i.e., one mastectomy, ten lumpectomy reexcisions).

TABLE 2.

Margin status characteristics

Parameter	Value
Histologic assessment, n (%)	Margin-level results	Specimen-level results*
Cancer-negative	579 (96.5)	86 (86.0)
Cancer-positive	21 (3.5)	14 (14.0)
Pathology
IDC	12 (57.1)	7 (50.0)
DCIS	3 (14.3)	2 (14.3)
ILC	5 (23.8)	4 (28.6)
Pleomorphic LCIS	1 (4.8)	1 (7.1)
Anatomical margin
Superficial	3 (14.3)	–
Deep	5 (23.8)	–
Cranial	5 (23.8)	–
Caudal	2 (9.5)	–
Medial	3 (14.3)	–
Lateral	3 (14.3)	–
Results based on tumor type, n (%)
IDC ± DCIS cases only	456 (76.0)	76 (76.0)
Cancer-negative	441 (96.7)	67 (88.2)
Cancer-positive	15 (3.3)	9 (11.8)
ILC ± LCIS cases only	144 (24.0)	24 (24.0)
Cancer-negative	138 (95.8)	19 (79.2)
Cancer-positive	6 (4.2)	5 (20.8)
Results based on surgical technique, n (%)
Wire localization	456 (76.0)	76 (76.0)
Cancer-negative	443 (97.2)	66 (86.8)
Cancer-positive	13 (2.9)	10 (13.2)
Palpation	144 (24.0)	24 (24.0)
Cancer-negative	136 (94.4)	20 (83.3)
Cancer-positive	8 (5.6)	4 (16.7)
Intraoperative targeted shave(s) taken, n (%)
No	–	74 (74)
Yes	–	26 (26)
Re-excision procedure performed, n (%)
No	–	89 (89)
Yes	–	11 (11)

IDC invasive ductal carcinoma; DCIS ductal carcinoma in situ; ILC invasive lobular carcinoma; LCIS lobular carcinoma in situ

^*“Cancer-positive” specimen is one that has ≤1 positive margins

Performances of SIA and micro-CT-based margin assessment are reported in Table 3. Margin-level performance by SIA yielded a sensitivity, specificity, PPV, and NPV of 33.3%, 95.9%, 22.6%, and 97.5%, respectively. SIA correctly identified seven of 21 positive margins (true positives) with a false positive rate (FPR) of 4.1%. Meanwhile, radiological interpretation of micro-CT yielded margin-level sensitivity, specificity, PPV, and NPV ranging from 23.8–42.9%, 87.7–89.8%, 7.8–11.3%, and 97.0–97.7%, respectively. Micro-CT readers correctly identified five to nine of 21 positive margins with an FPR ranging from 10.2 to 12.3%. Inter-reader agreement evaluating micro-CT scans for margin involvement was “fair” on a margin-level (three readers; Fleiss’ κ = 0.36; 95% confidence interval 0.34–0.37).

TABLE 3.

Margin-level and specimen-level assessment results and performance metrics with 95% Clopper–Pearson confidence intervals for all 100 initial resections

Parameter	Value
Margin-level results	Path+, Image+ (TP)	Path−, Image− (TN)	Path−, Image+ (FP)	Path+, Image− (FN)
SIA	7	555	24	14
Micro-CT Reader 1	6	514	65	15
Micro-CT Reader 2	5	520	59	16
Micro-CT Reader 3	9	508	71	12
Margin-level performance, n (95% CI)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
SIA	33.3 (14.6–57.0)	95.9 (93.9–97.3)	22.6 (12.4–37.5)	97.5 (96.7–98.2)
Micro-CT Reader 1	28.6 (11.3–52.2)	88.8 (85.9–91.2)	8.5 (4.3–15.9)	97.2 (96.3–97.8)
Micro-CT Reader 2	23.8 (8.2–47.2)	89.8 (87.0–92.1)	7.8 (3.7–15.9)	97.0 (96.2–97.6)
Micro-CT Reader 3	42.9 (21.8–66.0)	87.7 (84.8–90.3)	11.3 (6.9–17.9)	97.7 (96.7–98.4)
Specimen-level results	Path+, Image+ (TP)	Path−, Image− (TN)	Path−, Image+ (FP)	Path+, Image− (FN)
SIA	6	66	20	8
Micro-CT Reader 1	6	54	32	8
Micro-CT Reader 2	5	59	27	9
Micro-CT Reader 3	7	48	38	7
Specimen-level performance, n (95% CI)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
SIA	42.9 (17.7–71.1)	76.7 (66.4–85.2)	23.1 (12.8–38.1)	89.2 (83.8–93.0)
Micro-CT Reader 1	42.9 (17.7–71.1)	62.8 (51.7–73.0)	15.8 (8.8–26.7)	87.1 (80.7–91.6)
Micro-CT Reader 2	35.7 (12.8–64.9)	68.6 (57.7–78.2)	15.6 (7.9–28.6)	86.8 (81.2–90.9)
Micro-CT Reader 3	50.0 (23.0–77.0)	55.8 (44.7–66.5)	15.6 (9.4–24.7)	87.3 (79.7–92.3)

Path pathology; Image intraoperative margin assessment technique; TP true positive; TN true negative; FP false positive; FN false negative; SIA standard-of-care intraoperative margin assessment; CI confidence interval; PPV positive predictive value; NPV negative predictive value

On analysis at the specimen-level, SIA yielded a sensitivity, specificity, PPV, and NPV of 42.9%, 76.7%, 23.1%, and 89.2%, respectively. SIA correctly identified six of 14 margin-positive cases with an FPR of 23.5%. Meanwhile, micro-CT readers achieved sensitivity, specificity, PPV, and NPV ranges of 35.7–50.0%, 55.8–68.6%, 15.6–15.8%, and 86.8–87.3%, respectively. Micro-CT readers correctly identified five to seven of 14 margin-positive cases but suffered from an FPR range of 31.4–44.2%. Inter-reader agreement evaluating micro-CT scans for margin involvement was “moderate” on a specimen-level (three readers; Fleiss’ κ = 0.45; 95% confidence interval 0.42–0.48).

DISCUSSION

Intraoperative margin assessment in BCS is an ongoing clinical problem, as the overall reexcision rate due to positive margins is near 20% today.^19,20 Emerging technologies for improved margin assessment are at various stages of development and evaluation. We report on the results of micro-CT coupled with radiological interpretation for detecting positive margins on primary BCS resections. Micro-CT-based margin assessment offered similar performance to SIA (i.e., specimen palpation and radiography) for detecting positive margins but at the cost of higher false positive rates. Notably, our specimen-level micro-CT results are within the ranges of previously published, micro-CT margin assessment study results, except for PPV: specimen-level sensitivities of 38–93%, specificities of 53–100%, PPVs of 22–100%, and NPVs of 79–90%.^35–39

The wide range of reported micro-CT BCS margin assessment performances is notable, and little if any discussion in previous studies has systematically addressed these differences in performance. The total number of positive margins in this study was relatively low (21/600 margins from 14/100 specimens) compared to other studies. For example, DiCorpo et al. reported micro-CT BCS margin assessment results based on 114 margin-positive specimens and only 59 margin-negative specimens (66% of specimens were margin-positive).³⁸ Also, Kulkarni et al. reported micro-CT margin assessment results based on 57 margin-positive specimens from a total of 200 (29% of specimens were margin-positive).⁴⁰ It is possible that studies with a higher proportion of positive margins had a greater amount of cancer at the margin, making it easier to detect. Differences in micro-CT radiological interpretation could be the result of radiologist experience, level of training and experience interacting with volumetric specimen scans, and how the volumetric data are displayed and analyzed. Furthermore, the duration of radiological interpretation is inconsistent between published studies, ranging from less than 5 minutes^36,37,40 to “many hours,”³⁸ a factor that very likely impacts reported margin assessment performances while also impacting the relevance of micro-CT scanning for intraoperative use during BCS. Another consideration is the quality of the micro-CT scanning between studies. The added benefit of micro-CT systems capable of higher resolution scanning (i.e., on the order of 10 vs. 100 μm³ reconstruction voxel size) is unclear but may contribute to differences in reported margin assessment performances. Importantly, the present study did not use a preselected patient cohort predisposed to have an elevated positive margin rate (a so-called “enriched” population). Rather, the results presented herein represent a high percentage (66%) of all patients with invasive cancer seen in the routine practice of a surgical oncologist and are thus generalizable to common practice.

The primary limitation of micro-CT scanning for margin assessment is related to analyzing specimens with substantial fibroglandular tissue components. Radiological assessments noted that several cases were difficult to evaluate given the complexity of intervening fibroglandular tissues, which presented with similar radiodensity to tumor tissues. Figure 2 shows representative specimens with radiodense features extending to the margin. Of the six specimens shown in the figure, only one specimen had margins confirmed positive for cancer by histopathology (Patient 042, red arrows). Our study radiologists frequently elected to call radiodense features on the margin “positive,” leading to the high false positive rates reported by micro-CT readings relative to SIA (Table 3). This finding highlights a key challenge of micro-CT, because X-ray imaging is known to provide poor soft-tissue contrast and suggests that micro-CT scanning could be complemented by an adjuvant modality for improved margin assessment in these complex cases.^38,39,42

FIG. 2.

Representative specimens (each row) with intertwined tumor and fibroglandular tissues making radiological interpretation difficult. The columns from left to right (red, green, then yellow) show intersecting transverse, coronal, and sagittal planes in each scan. The specimen in row (a) was confirmed to have four positive margins by histopathology (red arrows), while all other specimens in rows (b)–(f) had only pathologically negative margins. IDC invasive ductal carcinoma; DCIS ductal carcinoma in situ; ILC invasive lobular carcinoma

Radiomics is a medical image-analysis technique that involves quantifying a wide array of image features that can be related to underlying pathophysiology.⁴³ A compelling potential use of three-dimensional radiomic analysis could be to quantify the “fibroglandular complexity” of each whole BCS resection using micro-CT imagery, thereby filtering specimens requiring additional imaging or analysis for accurate margin assessment. Machine learning approaches, e.g., using radiomics^44–46 or deep learning,^47,48 can perform tumor prognostication and segmentation tasks using volumetric medical image data. However, these quantitative analyses benefit from data that contain few if any artifacts. In this study, 76% of cases involved a surgical guidewire (Table 1), and it was standard-of-care to have a surgical clip placed in the primary lesion preoperatively, at the time of the diagnostic core biopsy. Radiological assessments noted difficulty interpreting many scans due to the presence of beam hardening artifacts from these localization tools, particularly the surgical guide wires (Fig. 3). The presence of these artifacts limited the potential value of computational analysis of our micro-CT data, beyond direct image interpretation by clinical radiologists. Other common localization tools in BCS, such as radiofrequency identification tags (e.g., LOCalizer⁴⁹), infrared reflectors (e.g., Savi Scout⁵⁰), and magnetic seeds (e.g., Magseed⁵¹), all contain metal components, and as such, are anticipated to introduce artifacts in micro-CT imagery similar to those created by surgical clips. The use of nonwire localization techniques, especially noninvasive techniques, such as intraoperative ultrasound,⁵² may be advantageous when coupled with micro-CT-based margin assessment. Future work could involve using dual-energy CT⁵³ or deep-learning techniques^54,55 for metal artifact reduction to enable computational, quantitative analysis. It also is feasible to consider micro-CT specimen scanning after removal of the surgical guide wire or other localization tools to circumvent these artifacts.

FIG. 3.

Representative specimens (each row) with beam-hardening artifacts from the surgical guidewire (a–c) and surgical clip (a–b) that impacted radiological interpretation of specific margins (yellow arrows). The columns from left to right (red, green, then yellow) show intersecting transverse, coronal, and sagittal planes in each scan. Only the cranial margin of the specimen in row (a) was confirmed to be positive by histopathology, while all margins of the specimens in rows (b) and (c) were confirmed to be pathologically negative. IDC invasive ductal carcinoma; DCIS ductal carcinoma in situ

Micro-CT scanning in this study performed similarly to SIA for identifying cancer-positive BCS resections. SIA correctly identified six margin-positive cases, while micro-CT scanning with radiological interpretation correctly identified five to seven margin-positive cases (Table 3). Micro-CT-based assessment identified some margin-positive cases that were missed by standard-of-care assessment. As such, if micro-CT-based margin assessment by a single reader had been combined with SIA, up to three additional margin-positive specimens (up to five positive margins) could have been identified during initial BCS procedures.

Micro-CT Reader 3 identified half (7/14) of all margin-positive cases in the study (Table 3). If a targeted cavity shave had been performed for every micro-CT-based positive reading, Reader 3’s interpretation also would have necessitated false positive cavity shaves in 38 cases. In contrast, SIA correctly identified six of 14 margin-positive cases with only 20 false positive cases. The trend of increased false positive readings is consistent for all micro-CT readers relative to SIA, both on a margin-level and on a specimen-level.

A noteworthy observation is that 24% of cases in the trial were lobular in origin (i.e., ILC ± LCIS), which is more than twice the rate of ILC among all cases of invasive breast cancer nationwide (~10%)⁵⁶ and is substantially higher than lobular carcinoma rates reported in other recent micro-CT margin assessment studies (0–10%).^36–38,40 The morphology of ILC (i.e., a single cell growth pattern) is known to make radiological detection and extent of disease evaluation difficult.⁵⁷ Specimen-level false negative and false positive readings occurred in 12.5% (3/24) and 16.7–20.8% (4/24–5/24) of lobular carcinoma cases, respectively (Supplemental Table 1). Meanwhile, false negative and false positive readings occurred in 5.3–7.9% (4/76–6/76) and 30.3–44.7% (23/76–34/76) of ductal carcinoma cases, respectively (Supplemental Table 2). A higher portion of lobular carcinoma cases yielded false negative readings relative to ductal carcinoma cases. However, ductal carcinoma cases were responsible for relatively more false positive readings.

Additional studies are needed to better capture the effectiveness of micro-CT scanning for intraoperative margin assessment during BCS. These studies should focus on reporting institutional standard-of-care techniques for intraoperative margin assessment, along with details related to the quality of micro-CT scanning, how and when radiological interpretation is performed, and the duration allowed for image interpretation, so that differences between published BCS specimen datasets and reported margin assessment performances can be better understood and characterized. A limitation of this study was that only specimens with an invasive cancer component were analyzed, although DCIS-only cases are responsible for a significant portion of positive margins in BCS today.⁵⁸ Future studies could investigate the performance of micro-CT for evaluating the margins of DCIS-only cases. Furthermore, this study involved a single surgical oncologist practicing at one medical institution. Additional studies investigating micro-CT-based margin assessment by multiple surgeons at multiple institutions are encouraged (e.g., the recent study by Kulkarni et al. involved a total of seven surgeons at two institutions⁴⁰).

Micro-CT scanning alone is ineffective for margin assessment in complex cases when radiodense fibroglandular tissue structures obscure the primary lesion. In these cases, adjuvant methods of margin assessment used in combination with micro-CT may provide value. Finally, differences in published performances for micro-CT evaluation of margins are substantial and should be the focus of future studies as the technology is further evaluated for intraoperative margin assessment during BCS.