Abstract
Purpose:
To assess the impact of adjunctive ultrasound guided diffuse optical tomography (US-guided DOT) on BI-RADS assessment in women undergoing US-guided breast biopsy.
Method:
This prospective study enrolled women referred for US-guided breast biopsy between 3/5/2019 and 3/19/2020. Participants underwent US-guided DOT immediately before biopsy. The US-guided DOT acquisition generated average maximum total hemoglobin (HbT) spatial maps and quantitative HbT values. Four radiologists blinded to histopathology assessed conventional imaging (CI) to assign a CI BI-RADS assessment and then integrated DOT information in assigning a CI&DOT BI-RADS assessment. HbT was compared between benign and malignant lesions using an ANOVA test and Tukey’s test. Benign biopsies were tabulated, deeming BI-RADS ≥ 4A as positive. Reader agreement was assessed.
Results:
Among 61 included women (mean age 48 years), biopsy demonstrated 15 (24.6%) malignant and 46 (75.4%) benign lesions. Mean HbT was 55.3 ± 22.6 μM in benign lesions versus 85.4 ± 15.6 μM in cancers (p < .001). HbT threshold of 78.5 μM achieved sensitivity 80% (12/15) and specificity 89% (41/46) for malignancy. Across readers and patients, 197 pairs of CI BI-RADS and CI&DOT BI-RADS assessments were assigned. Adjunctive US-guided DOT achieved a net decrease in 23.5% (31/132) of suspicious (CI BI-RADS ≥ 4A) assessments of benign lesions (34 correct downgrades and 3 incorrect upgrades). 38.3% (31/81) of 4A assessments were appropriately downgraded. No cancer was downgraded to a non-actionable assessment. Interreader agreement analysis demonstrated kappa = 0.48–0.53 for CI BI-RADS and kappa = 0.28–0.44 for CI&DOT BI-RADS.
Conclusions:
Integration of US-guided DOT information achieved a 23.5% reduction in suspicious BI-RADS assessments for benign lesions. Larger studies are warranted, with attention to improved reader agreement.
Keywords: Breast cancer, Breast ultrasound, Diffuse optical tomography, Optical imaging, Predictive value
1. Introduction
Malignant and benign breast disease represent a heterogeneous spectrum of disorders with overlapping imaging characteristics. For example, a benign fibroadenoma typically appears on both mammography and ultrasound as an oval mass with circumscribed margins, but these imaging features may also be observed in mucinous and medullary carcinoma and occasionally high grade triple receptor negative invasive carcinoma. This overlap in imaging characteristics influences radiologists to recommend biopsy and thereby contributes to a low positive predictive value (PPV) of biopsy recommended (PPV2) and PPV of biopsy performed (PPV3). In 2017, the Breast Cancer Surveillance Consortium reported a PPV2 of 27.5% [1]. Most suspicious imaging abnormalities recommended for biopsy are of low to moderate suspicion (BI-RADS subcategories 4A or 4B), associated with PPV3 of 7.6% and 22.3%, respectively [2]. Of note the PPV of diagnostic breast US varies worldwide and is often higher than demonstrated in the United States. Studies from Europe, South America and South Korea report PPV of breast US of 39%, 29%, and 51% (BIRADS 4 overall), 20%, 6%, and 26% (BI-RADS 4A), 41.5%, 25.4%, and 83% (BI-RADS 4B), and 74%, 81%, and 91% (BI-RADS 4C) respectively [3–5].
Optical imaging with diffuse optical tomography (DOT) transmits near infra-red light at specific wavelengths selectively absorbed by water, fat, oxygenated and deoxygenated hemoglobin. Average maximum total hemoglobin concentration (HbT), the sum of oxygenated and deoxygenated hemoglobin average maximum concentrations, provides information on tissue vascularity distinct from that captured by color Doppler. Since malignant lesions generally have higher HbT, this variable may help differentiate benign from malignant breast lesions [6–9]. More current DOT systems use mammography, ultrasound (US), or MRI to improve accuracy of optical measurements [10–17]. Combined conventional and optical imaging provides complementary diagnostic data involving structure and function. In the case of contrast enhanced MRI, which already provides functional data, the addition of optical imaging may improve specificity and overall accuracy [18].
US-guided DOT is conducive to clinical practice because it is safe, portable, inexpensive, and able to use preexisting commercial US equipment. A commercial US transducer is fitted within a hand-held optical probe to create a co-registered US-guided DOT device. The co-registered US device localizes the lesion, and the optical system images tumor hemoglobin content. US-guided DOT has shown promise in differentiating breast cancer from benign abnormalities [12–14,19]. However, to our knowledge, US-guided DOT has not been assessed under prospective clinical conditions. It is possible that information obtained from US-guided DOT could be used to refine BI-RADS assessments to achieve better patient selection for biopsy.
This study aimed to assess the impact of adjunctive US-guided DOT on BI-RADS assessment in women undergoing US-guided breast biopsy.
2. Methods
2.1. Patients
This prospective study was approved by the local IRB and was HIPAA compliant. On a weekly basis from March 5, 2019 through March 19, 2020, a non-author clinical research assistant (CRA) reviewed records of women ≥ 18 years old scheduled for US-guided core needle biopsy or diagnostic cyst aspiration at a single institution. The CRA excluded patients on the basis of pregnancy, history of breast augmentation, or prior radiation treatment to the breast, and forwarded a list of eligible patients to one of two breast radiologists, who reviewed the patients’ imaging to confirm lesion eligibility. Potential participants were excluded if the lesion was located within 5 mm of the skin, beneath the nipple-areolar complex, at the site of recent (<6 months) biopsy, or in the same region (typically within the same quadrant) as additional imaging abnormalities, or if having an abnormality in the mirror image location of the contralateral breast, (including mastectomy). Exclusions at the patient and lesion levels were due to the possibility of altered vascularity in the setting of preexisting conditions, e.g. pregnancy or prior surgical or radiation treatment. Further exclusions were made to ensure the accuracy of optical data transmission and reconstruction, which may be limited by optical absorption by the nipple areolar complex or effected by superficial location. Additional lesions within the ipsilateral area of interest or contralateral reference sites were excluded to prevent confounding optical signatures unrelated to the lesion undergoing biopsy and to preserve the adequacy of the reference data. The CRA approached eligible patients to assess their willingness to participate, resulting in additional patient exclusions. Written informed consent was obtained on the day of biopsy for all included patients.
2.2. Diffuse optical tomography prototype and image reconstruction
The DOT prototype used in this study has been described [20]. In brief, the DOT probe has a clamshell design (containing the DOT components) that opens to accommodate a commercial US transducer (Fig. 1). The footprint of the probe is approximately 9 cm. The DOT probe body has nine light “emitter” fibers embedded at different positions in one half, and 14 light “detector” fibers within the other half. Distances between source and detector positions on the DOT probe range from 3 cm to 7.5 cm. Four source laser diodes produce wavelengths at 740 nm, 785 nm, 808 nm and 830 nm that are switched sequentially to the nine light emitter fibers. Using a previously validated algorithm [21], each acquired data set is reconstructed at four wavelengths and analyzed across wavelengths to determine quantitative hemoglobin values and display hemoglobin spatial map distributions. The resulting hemoglobin spatial distribution map typically includes seven slices reconstructed at 0.5 cm increments starting from 0.2 to 0.5 cm below the skin surface. Hemoglobin maximum for each data set is derived from the slices containing a sonographically visible lesion. Eight to ten datasets are acquired for each lesion site and an average HbT is computed from the maxima of all lesion data sets. Image reconstruction is performed offline and takes approximately 20 min. The engineer performing data reconstruction was blinded to pathology results.
Fig. 1.
Photographs of the diffuse optical tomography (DOT) probe in the open position with a commercial ultrasound transducer (yellow arrow) in place (A), and in the closed position looking at the probe surface (B). White arrow (B) represents 1 of 9 light emission optical fibers, blue arrowhead (B) represents 1 of 14 light detection optical fibers.
2.3. Image acquisition
US and US-guided DOT examinations were obtained immediately before biopsy using one of four commercial US units equipped with SL-15–4 transducers (Aixplorer, SuperSonic Imagine Inc., Aix-en-Provence, FR) and a 4th generation DOT system described above [20]. The US examination was performed by a breast US technologist supervised by one of five breast imaging radiologists participating in the study. In the United States it is common practice for a dedicated ultrasound technologist to perform the initial US exam. Images obtained by the technologist are reviewed by the supervising radiologist who may then choose to perform additional scanning. In this study, the role of the supervising radiologist was to ensure that the lesion of interest was evaluated and to confirm technical adequacy of the exam and inclusion of requisite images (described below). The five radiologists participating in the study are hereafter referred to as R1, R2, R3, R4, and R5, with 21, 32, 25, 10, and 3 years of breast US experience, respectively. R3 and R4 had 2.5 years, and R5 one year, of experience with US-guided DOT, although they had not previously performed DOT interpretations; R1 and R2 had no prior experience in US-guided DOT. The radiologist assigned to supervise the US exam preceding US-guided DOT differed from the radiologist who performed the subsequent US-guided intervention to prevent procedure related feedback, such as resolution of a complicated cyst during biopsy or aspiration, from influencing DOT interpretation. The biopsy radiologist did not interpret DOT data of their biopsy patients to preserve blinding of pathology results.
The breast US was acquired in approximately 5 min and was used to localize the lesion of interest for subsequent US-guided DOT and to systematically characterize the abnormality. Three pairs of orthogonal images were obtained (gray scale, gray scale with measurements, and power Doppler). The supervising radiologist recorded conventional US descriptors of the index lesion based on the BI-RADS 5th edition [22]. After exam completion, the commercial US transducer was inserted into the DOT probe. The DOT engineer was directed to the target lesion and then performed US-guided DOT of the index lesion and of the mirror image location of the contralateral breast. The mirror image contralateral breast served as a normal reference for calculating background absorption and scattering coefficients using the standard fitting algorithm of measured amplitude and phase vs. source-detector distance at the contralateral side [23]. The background values were used in the image reconstruction algorithm of the index lesion. The depth of the contralateral reference scan was matched to the depth of the lesion scan. Multiple DOT datasets were acquired of the lesion (8–10 datasets) and the contralateral site (5–8 datasets). The DOT acquisition took < 5 s per dataset, and the entire US-guided DOT examination took approximately 5–10 min.
2.4. Exam interpretation
Interpretation of conventional breast imaging (CI), (i.e. mammography and US), and DOT data was performed by four study radiologists (R1, R2, R3, and R4) who underwent a training session prior to DOT interpretation (see Supplementary Methods). Reading occurred as sequential or batch reads; batch reading sessions occurred at three separate timepoints (see Supplementary Methods). Readers were blinded to histopathology outcome and were precluded from interpreting patients for whom they performed biopsy. Readers were provided the patient’s clinical and breast imaging information, including standard of care imaging reports. The reader first re-interpreted the CI examinations, i.e. US and mammography (when performed), and gave a CI BI-RADS assessment [22]. BI-RADS assessments from the standard of care exam report were not used. The optical data were then reviewed and an integrated CI&DOT BI-RADS assessment was rendered. Change in the BI-RADS assessment (upgrade, no change, or downgrade) due to DOT was recorded. Readers were provided quantitative HbT values and evaluated HbT spatial map features qualitatively, prioritizing quantitative data over qualitative data. Readers were guided by previously published HbT thresholds of benign (HbT < 50 μM) or malignant (HbT ≥ 80 μM) lesions [12,14]. In general, when conventional imaging demonstrated a low to moderate suspicion lesion and HbT was below 50 μM the lesion was downgraded. Lesions with HbT equal to or above 80 μM were typically upgraded or unchanged based on their CI appearance. BI-RADS assessment of lesions with HbT in the 50–79 μM range were usually unchanged, with infrequent downgrading or upgrading based on a combination of the CI appearance and whether the HbT was in the lower or upper part of the intermediate range. When evaluating HbT spatial maps, the reader decided if there was an increase in HbT above an expected low level of HbT at the site of interest, termed “background”. For cases exceeding “background”, the reader described three HbT spatial map features: shape (“Oval” or “Not Oval”), location of highest HbT (“Central” or “Peripheral”), and ring distribution of HbT, i.e. the spacing of concentric HbT rings (“Tight”, “Loose”, or “Uncertain”) (Fig. 2).
Fig. 2.
Single representative level from spatial maps of average maximum total hemoglobin concentration (HbT) of three different patients highlighting the HbT spatial map features. A. 40-year-old woman with fat necrosis. Spatial map shows background HbT only. B. 50-year-old woman with intermediate grade invasive ductal carcinoma. Spatial map shows high HbT (113 μM), oval shape, central location of highest HbT (arrow), and tight ring distribution (bracket). C. 40-year-old woman with fibroadenomatoid change. Spatial map shows intermediate HbT (72 μM) with “not oval” shape, peripheral location of highest HbT (arrowheads), and diffuse ring distribution (bracket).
2.5. Histopathology
Histopathologic data were extracted from pathology reports. One breast pathologist with 9 years of experience retrospectively reviewed the US-guided core needle biopsy and related surgical results. Patients with benign histopathology were categorized according to the primary histopathology noted in the report. Benign histopathology was further categorized as non-proliferative, proliferative, and fibroadenoma-like (listed in Table 1). These three histologic groups have been shown to have different HbT content [12,14]) and may lead to different treatment decisions. Cancer included invasive carcinoma and ductal carcinoma in situ and was evaluated for histologic subtype, grade (ranging from 1 to 3), and Nottingham score (ranging from 1 to 9).
Table 1.
Age, maximum total hemoglobin concentration (HbT), and histopathology of 61 patients who underwent diffuse optical tomography summarized by histology subgroup.
| Group (n) | Age (y) Mean (range) | HbT (μM) mean ± SD | Histologic diagnosis on biopsy |
|---|---|---|---|
|
| |||
| Cancer (15) | 48 (27–79) | 85.4 ± 15.6 (reference) | Invasive ductal carcinoma (9) Invasive lobular carcinoma (2) Solid papillary carcinoma/invasive ductal carcinoma (1) Invasive papillary carcinoma (1) Invasive mammary carcinoma** (1) Ductal carcinoma in situ (1) |
| Proliferative (8) | 40 (29–57) | 70.3 ± 24.7 (+p = .32) | Sclerosing adenosis (2) Florid usual ductal hyperplasia (1) Usual ductal hyperplasia (1) Radial scar (1) Intraductal papilloma (1) Papillary apocrine metaplasia (2) |
| Fibroadenoma-like (16) | 42 (18–70) | 59.6 ± 18.8 (+p = .004) | Fibroadenoma (12) Fibroadenomatoid change (2) Fibroepithelial lesion* (2) |
| Non-proliferative (22) | 50 (26–72) | 47.9 ± 21.5 (+p < .001) |
Fibrosis (6) Pseudoangiomatous stromal hyperplasia (4) Cystic apocrine metaplasia (4) Cyst (4) Fibrocystic disease (2) Fat necrosis (2) |
Both lesions exhibited imaging and clinical features consistent with fibroadenoma.
Invasive mammary carcinoma – invasive epithelial breast cancer is present but the underlying phenotype, invasive ductal carcinoma vs. invasive lobular carcinoma, is uncertain.
adjusted p value.
2.6. Statistical analysis
HbT was compared between histologic groups using the one way ANOVA test followed by a post-hoc Tukey’s honest test of significance. The Wilcoxon rank sum test was used to compare lesion size between benign and malignant lesions, and Spearman coefficient used to assess the correlation of lesion size with HbT. We used DeLong’s method to test whether the area under the curve (AUC) of HbT differed from the AUC of lesion size [24]. We derived the optimal cutoff point at which the (sensitivity, specificity) pair has the shortest distance to the perfect classification (sensitivity = 1, specificity = 1) on the ROC curve for HbT.
An analysis of BI-RADS assessment was performed using a single read per radiologist. If a reader provided multiple reads for a given patient, then the read from the most recent session was used. BI-RADS 2 and 3 assessments were considered test negative; BI-RADS 4A, 4B, 4C, and 5 assessments were considered test positive.
HbT shape, location of highest HbT, and HbT ring distribution pooled data from R1, R2, and R3 were compared between benign and malignant lesions using Fisher’s exact test.
For interreader agreement, Fleiss’s Kappa coefficient was calculated for CI BI-RADS, CI&DOT BI-RADS and HbT spatial map features across all readers using the R package “psych”, performed separately for batch sessions 1 and 2. Cohen’s Kappa statistic was calculated for intra-reader agreement for each reader (R1, R2, and R3) between batch session 1 and 2, with 95% CIs derived by 1000 bootstrapping samples; the average Cohen’s Kappa across readers was computed. Two different kappa statistics were used, as Cohen’s kappa assesses agreement between two readers and Fleiss’ kappa applies to more than two readers.
The logistic mixed effects model was applied across batch sessions 1 and 2 to model the binary outcome of cancer versus non-cancer with the fixed effect of each variable (CI BI-RADS, CI&DOT BIRADS, HbT shape, location of highest HbT, and HbT ring distribution), the fixed effect of reading session and reader, and the random effect of patients, considering the correlation of reads on the same patient’s examination. Logistic regression coefficients, standard error (SE), and Wald p values were derived [25].
3. Results
A total of 127 eligible patients were approached for study participation. Of these, 64 declined to participate; a total of 63 patients agreed to participate and provided informed consent. Two patients were subsequently excluded from analysis (one whose biopsy was canceled, and one with a technically deficient DOT scan) (Fig. 3). The remaining 61 patients (mean age 48 years, range 18–79 years) with 61 lesions constituted the study group and underwent US and US-guided DOT imaging, and subsequent US-guided intervention [core needle biopsy (59), diagnostic cyst aspiration (2)]. Biopsy demonstrated 15 malignant lesions (24.6%), and 46 benign lesions (75.4%) [22 non-proliferative, 16 fibroadenoma-like, and 8 proliferative]. Cancers were significantly larger than benign lesions [median size 1.9 cm with interquartile range (IQR) 1.6–2.6 cm vs 1.3 cm (IQR 1.0–2.0 cm), respectively; p = .02).
Fig. 3.
Flow chart of patient sample. DOT = diffuse optical tomography; US = ultrasound.
The mean HbT of cancer was significantly higher than the mean HbT of all benign lesions (85.4 μM vs 55.3 μM, p < .001), and was also significantly higher than the mean HbT of benign non-proliferative lesions (85.4 μM vs 47.9 μM, p < .001) and the mean HbT of fibroadenoma-like lesions (85.4 μM vs 59.6 μM p = .004) (Table 1). However, the mean HbT of cancer was not significantly different than the mean HbT of benign proliferative lesions (85.4 μM vs 70.3 μM p = .32) (Table 1). 80% (12/15) of malignant lesions and 11% (5/46) of benign lesions demonstrated HbT > 80 μM (Table 2, Table 3). A typical example of the DOT features of a malignant lesion is demonstrated (see Fig. 4).
Table 2.
Characteristics of 15 patients with cancer. CI BI-RADS and CI&DOT BI-RADS assessments are provided for all four readers (R1, R2, R3, and R4) and left blank when not available for a given patient for a given reader.
| Patient # | Age (y) | CI BI-RADS R1 R2 R3 R4 | CI&DOT BI-RADS R1 R2 R3 R4 | HbT (μM) | Size on US (cm) | Histologya | Grade | Size on Histology (cm)a | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
| ||||||||||||||
| 002 | 27 | 4C | 4C | 4C | 4B | 4C | 4C | 4C | 4C | 90.4 | 2.3 | IDC | 3 (9/9) | 1.8 |
| 005 | 44 | 5 | 5 | 5 | 5 | 5 | 5 | 5 | 5 | 85.6 | 1.8 | IDC | 2 (6/9) | NAC |
| 010 | 58 | 5 | 5 | 5 | 5 | 5 | 5 | 5 | 5 | 80.6 | 4.1 | IDC | 3 (9/9) | NAC |
| 020 | 78 | 5 | 5 | 5 | 5 | 5 | 5 | 5 | 5 | 89.1 | 1.6 | IDC | 3 (9/9) | 1.5 |
| 022 | 73 | 5 | 5 | 5 | 5 | 5 | 4C | 4C | 5 | 64.0 | 3.4 | ILC | 3 (9/9) | 2.2 |
| 023 | 69 | 5 | 5 | 5 | 5 | 5 | 5 | 5 | 5 | 81.3 | 1.5 | SPC/IDC | 1 (5/9) | 1.5b |
| 024 | 72 | 4B | 4B | 4C | 4C | 4B | 4B | 4B | 4A | 56.2 | 1.2 | IPC | 2 (6/9) | 0.6 |
| 032 | 51 | 4C | 4C | 4C | 5 | 5 | 5 | 113.0 | 1.6 | IDC | 2 (6/9) | 1.6 | ||
| 033 | 59 | 4C | 4C | 4C | 4C | 4C | 4C | 69.3 | 0.9 | IDC | 2 (6/9) | 1.4 | ||
| 034 | 69 | 4C | 5 | 5 | 4C | 5 | 5 | 81.0 | 4.6 | IMC | 3 (8/9) | * | ||
| 036 | 35 | 5 | 5 | 5 | 5 | 5 | 5 | 80.9 | 1.7 | IDC | 1 (5/9) | * | ||
| 039 | 56 | 5 | 5 | 5 | 5 | 105.5 | 2.6 | ILC | 3 (9/9) | 3.2b | ||||
| 051 | 68 | 4C | 4C | 4B | 4C | 4C | 4C | 91.2 | 1.9 | DCIS | NA | 1.5 | ||
| 057 | 61 | 5 | 5 | 5 | 5 | 107.9 | 2.4 | IDC | 3 (9/9) | NAC | ||||
| 063 | 78 | 4C | 5 | 4C | 5 | 85.0 | 2.6 | IDC | 1 (5/9) | NAC | ||||
Based on surgical specimen following resection if available, and otherwise on biopsy.
Multi-focal or multi-centric disease, size listed corresponds to the largest malignant deposit.
Surgery either performed at outside institution or not performed due to disease progression.
CI = conventional imaging; DOT = diffuse optical tomography; HbT = average maximum total hemoglobin concentration; IDC = invasive ductal carcinoma; ILC = invasive lobular carcinoma; SPC = solid papillary carcinoma; IPC = invasive papillary carcinoma; IMC = invasive mammary carcinoma; DCIS = ductal carcinoma in situ; NAC = Neoadjuvant Chemotherapy; US = ultrasound. NA = not applicable.
Table 3.
Characteristics of 5 patients with a benign diagnosis and with average maximum total hemoglobin concentration (HbT) ≥ 80 μM.
| Patient # | Age (y) | Histologic diagnosis | HbT (μM) |
|---|---|---|---|
|
| |||
| 025 | 39 | Fibroadenoma | 92.0 |
| 027 | 42 | Fibroadenoma | 84.8 |
| 040 | 57 | Intraductal papilloma / ADH | 105.5 |
| 046 | 49 | Papillary apocrine metaplasia | 80.5 |
| 061 | 43 | Complex sclerosing lesion | 87.7 |
ADH = atypical ductal hyperplasia.
Fig. 4.
Anti-radial ultrasound (US) image and average maximum total hemoglobin concentration (HbT) spatial distribution map of a 50-year-old woman (patient 32) with a 1.6 cm stage 1 intermediate grade invasive ductal carcinoma. US shows an irregular hypoechoic mass with indistinct margins (arrow). US-guided diffuse optical tomography (DOT) reveals HbT of 113 μM with oval shape, central high HbT, and tight ring distribution. All three readers upgraded assessment from BI-RADS 4C on conventional imaging (CI) to BI-RADS 5 based on CI and DOT.
We performed additional analyses due to concern that lesion size rather than pathology type was accounting for HbT differences between benign and malignant groups. Lesion size and HbT were only weakly correlated (r = 0.21). The AUC for predicting cancer trended higher for HbT, (AUC = 0.86, 95% CI: 0.76–0.97), than the AUC for lesion size, (AUC = 0.71, 95% CI: 0.57–0.84) (p = .06). For HbT, the optimal cutoff point on the ROC curve of the sensitivity, specificity pair was 78.5 μM. This resulted in a specificity = 0.89, sensitivity = 0.80, negative predictive value = 0.93, and positive predictive value = 0.71.
The BI-RADS assessment analysis included 197 reads on 61 patients (51 by R1, 55 by R2, 57 by R3, 34 by R4), corresponding with 197 assessments for each of CI BI-RADS and CI&DOT BI-RADS. Two reads were inadvertently performed by the biopsy radiologist; one occurred four months after biopsy by R3 and one > 6 months after biopsy by R4. Study reads were missing in two patients, including one by R1 and one by R2.
The agreement analysis was conducted on the initial 35 patients. Interreader agreement included 157 reads, 95 from the first batch session (30, 33, and 32 by R1, R2, and R3 respectively), and 62 from the second batch session (21, 21, and 20 by R1, R2, and R3, respectively). The intrareader agreement analysis included 120 reads on 20 patients per reader from batch sessions 1 (60) and 2 (60), 20 per session by each of R1, R2 and R3.
In the BI-RADS assessment (Table 4), the false-positive rate was 67% (132/197) for CI BI-RADS and 51% (101/197) for CI&DOT BI-RADS. The use of US-guided DOT resulted in a net decrease in 31 of 132 (23.5%) false positive CI BI-RADS assessments. Thirty-four suspicious assessments of benign lesions were correctly downgraded from CI BI-RADS ≥ 4A to CI&DOT BI-RADS 2 (n = 18) or 3 (n = 16). Three probably benign assessments (CI BI-RADS 3) were incorrectly upgraded to CI&DOT BI-RADS 4A. 38.3% (31/81) of BI-RADS 4A assessments and 5.9% (3/51) of BI-RADS 4B assessments were appropriately downgraded. The 34 downgrades occurred in 19 unique patients and are summarized in Table 5. Fig. 5 shows a representative example of a correctly downgraded benign lesion.
Table 4.
Comparison of readers’ BI-RADS assessments of the index lesion based on conventional imaging (CI) and integrated CI and diffuse optical tomography (DOT), stratified by whether the lesion was benign or malignant on biopsy, and by whether the BI-RADS assessment indicated follow-up (2 or 3) or biopsy (≥4).
| Reader | CI | CI & DOT | ||||||
|---|---|---|---|---|---|---|---|---|
|
|
|
|||||||
| Benign | Malignant | Benign | Malignant | |||||
|
|
|
|
|
|||||
| 2 or 3 | ≥4 | 2 or 3 | ≥4 | 2 or 3 | ≥4 | 2 or 3 | ≥4 | |
|
| ||||||||
| All | 17 | 132 | 0 | 48 | 48 | 101 | 0 | 48 |
| 1 | 1 | 40 | 0 | 10 | 6 | 35 | 0 | 10 |
| 2 | 5 | 37 | 0 | 13 | 12 | 30 | 0 | 13 |
| 3 | 3 | 39 | 0 | 15 | 17 | 25 | 0 | 15 |
| 4 | 8 | 16 | 0 | 10 | 13 | 11 | 0 | 10 |
Table 5.
Characteristics of 19 patients with at least one reading in which a BI-RADS ≥ 4A assessment on conventional imaging was downgraded to a non-actionable BI-RADS through integration of US-guided diffuse optical tomography (DOT).
| Patient # (Age) | Primary Diagnosis | Finding Typea | Echo Patterna | Shapea | Marginsa | Post. Featurea | Intern. Vasc.a | Size on US (cm)c | HbT (μM) | Change in BI-RADS due to DOT (# of readers) |
|---|---|---|---|---|---|---|---|---|---|---|
|
| ||||||||||
| 001 (67) | Fat Necrosis | Mass | Hetero. | Irregular | Indistinct | No PF | Yes | 0.6 | 21 | 4A to 3 (1) 4A to 2 (1) |
| 003 (44) | Fibrosis | Mass | Complex Cystic | Oval | Circum. | Enhance | No | 1.4 | 49 | 4A to 3 (1) |
| 008 (26) | Fibrosis | Asym. | NA | NA | NA | NA | No | 2.2 | 34 | 4A to 3 (1) 4A to 2 (2) |
| 009 (69) | Fat necrosis | Mass | Hyper. | Round | Indistinct | Shadow | No | 0.5 | 19 | 4B to 3 (1) 4A to 3 (1) 4A to 2 (1) |
| 012 (43) | Cystic apocrine metaplasia | Mass | Complex Cystic | Oval | Indistinct | No PF | No | 1.0 | 70 | 4A to 3 (1) |
| 014 (40) | Stromal fibrosis | Mass | Hypo. | Oval | Indistinct & angularb | No PF | No | 1.2 | 28 | 4A to 2 (1) |
| 015 (52) | FAC | Mass | Hetero. | Oval | Circum. | Comb. | Yes | 4.3 | 42 | 4A to 2 (1) |
| 016 (37) | Fibroadenoma | Mass | Hypo. | Oval | Circum. | Enhance | Yes | 2.5 | 45 | 4A to 3 (1) |
| 017 (52) | Cyst aspirated | Mass | Hypo. | Oval | Circum. | Enhance | No | 2.7 | 38 | 4A to 2 (2) |
| 021 (52) | Intraductal papillary apocrine metaplasia | Mass | Complex Cystic | Oval | Circum. | Enhance | Yes | 1.0 | 36 | 4B to 2 (1) 4A to 3 (1) 4A to 2 (1) |
| 028 (18) | Fibroadenoma | Mass | Hypo. | Oval | Circum. | No PF | Yes | 4.5 | 36 | 4A to 2 (1) |
| 029 (37) | Stromal fibrosis | Mass | Hetero. | Irregular | Indistinct & Angularb | Shadow | Yes | 1.2 | 22 | 4B to 3 (1) 4A to 2 (1) |
| 041 (48) | PASH | Mass | Complex Cystic | Irregular | Circum. | Comb. | No | 1.5 | 14 | 4A to 3 (2) 4A to 2 (1) |
| 047 (51) | Fibroadenoma | mass | Hypo. | Oval | Circum. | Enhance | Yes | 1.2 | 19 | 4A to 3 (2) 4A to 2 (1) |
| 054 (45) | PASH | Asym. | NA | NA | NA | NA | No | 1.9 | 48 | 4A to 3 (2) |
| 058 (61) | Microcysts | Mass | Anechoic | Irregular | Circum. | Enhance | No | 0.8 | 14 | 4A to 2 (1) |
| 059 (58) | Microcysts with apocrine metaplasia | Mass | Hypo. | Oval | Circum. | Enhance | No | 1.1 | 30 | 4A to 2 (2) |
| 060 (53) | PASH | Mass | Hypo. | Oval | Circum. | No PF | No | 2.0 | 41 | 4A to 2 (1) |
| 062 (52) | Columnar cell change | Mass | Hypo. | Irregular | Indistinct | Shadow | No | 1.9 | 65 | 4A to 3 (1) |
based on conventional US descriptors (BI-RADS 5th edition).
multiple margin descriptors were attributed.
maximum diameter measured.
Post. = Posterior; HbT = average maximum total Hemoglobin concentration; Hetero. = Heterogeneous; No PF = No Posterior Features; Circum. = circumscribed; Enhance = Enhancement; Complex Cystic = Complex Cystic and Solid; Asym. = asymmetry; NA = not applicable; Hyper. = Hyperechoic; Shadow = Shadowing; Hypo. = hypoechoic; Comb. = Combined Pattern; PASH = pseudoangiomatous stromal hyperplasia; US = ultrasound.
Fig. 5.
Radial ultrasound image and average maximum total hemoglobin concentration (HbT) spatial distribution map of a 26-year-old woman (patient 008) with biopsy proven stromal fibrosis. The patient presented with a palpable abnormality corresponding to a sonographic hypoechoic area. Diffuse optical tomography (DOT) revealed HbT of 33.7 μM, below level of < 50 μM that is indicative of benign lesions. Three of four readers downgraded the lesion from a BI-RADs 4A on conventional imaging (CI) to BI-RADS 2 or 3 based on CI and DOT; one reader maintained BI-RADS 4A assessment.
With regard to malignant lesions, 6 assessments in 3 patients were appropriately upgraded from CI BI-RADS 4B to CI&DOT BI-RADS 4C (n = 2) and from CI BI-RADS 4C to CI&DOT BI-RADS 5 (n = 4) (Table 2, Fig. 4). Four assessments in two patients were wrongly downgraded to lower suspicion assessments: CI BI-RADS 5 to CI&DOT BI-RADS 4C (n = 2), and CI BI-RADS 4C to CI&DOT BI-RADS 4B (n = 1) and CI&DOT BI-RADS 4A (n = 1) (Table 2). Both of these patients had HbT < 80 μM and somewhat uncommon malignant pathology. Subject 022 had HbT of 64 μM and a diagnosis of invasive lobular carcinoma, while subject 024 had HbT of 56 μM and combined solid papillary carcinoma and invasive ductal carcinoma. No malignant lesion was downgraded to an assessment that did not have an associated biopsy recommendation.
HbT shape was not-oval in 57.1% (16/28) of cancers versus 46.8% (22/47) of benign lesions (p = 0.5). Location of highest HbT was peripheral in 28.6% (8/28) of cancers versus 42.6% (20/27) of benign lesions (p = 0.3). HbT ring distribution was tight in 38.5% (10/26) of cancers versus 6.5% (3/46) of benign lesions (p = 0.001).
Based on logistic mixed effects model fitting, CI BI-RADS, CI&DOT BI-RADS, and HbT ring distribution were positively associated with malignant pathology (all with p <0.01), while HbT shape (p = 0.93) and location of highest HbT (p = .97) were not.
For CI-BI-RADS, interreader agreement was moderate, kappa(κ) = 0.53 (95% CI: 0.24–0.73) batch session 1 and κ = 0.48 (95% CI: 0.15–0.70) batch session 2. This compared with moderate, κ = 0.44 (95% CI: 0.23 to 0.60), and fair agreement, κ = 0.28 (95% CI: 0.03 to 0.48) for CI&DOT BI-RADS. HbT spatial map feature agreement was: κ = 0.50 (95% CI: 0.32 to 0.66) for HbT shape, κ = 0.33 (95% CI: 0.11 to 0.55) for location of highest HbT, and κ = 0.43 (95% CI: 0.18 to 0.60) for HbT ring distribution.
Intrareader agreement had kappa values for R1, R2, and R3 of 0.67, 0.50, and 0.53 (average 0.57) for CI BI-RADS; of 0.57, 0.37, and 0.62 (average 0.52) for CI&DOT BI-RADS; 0.25, 0.71, and 0.88 (average 0.63) for HbT shape; − 0.02, 0.27, and 0.60 (average 0.31) for location of highest HbT; and 1.0, 1.0, and 1.0 (average 1.0) for HbT ring distribution.
4. Discussion
This prospective study highlights the potential of adjunctive US-guided DOT to correctly downgrade benign lesions with low to moderate suspicion assessments. There is no expectation for downgrading high suspicion (category 4C or 5) abnormalities, but these account for < 15% of positive diagnostic assessments [2]. In one large study, 90% (125,447/139,533) of positive diagnostic assessments were BI-RADS 4 and of those subcategorized, 55.6% (23,258/41,841) were BI-RADS 4A and 31.8% (13,302/41,841) were BI-RADS 4B, with PPV3 of 7.6% and 22% respectively [2]. In a public health context, a net 23.5% reduction in biopsy recommendations could translate to substantial benefits in societal costs and improved resource allocation, and spare those patients the anxiety, costs, and potential morbidity of percutaneous biopsy.
The US-DOT technology is appealing because it has no known health risks and few barriers to clinical implementation. The DOT probe can be made to accommodate most of the commercially available US transducers. The technology is inexpensive, has a rapid acquisition time, and can be performed by any operator with breast US experience. It provides complimentary functional information on lesion vascular content akin to contrast-enhanced MRI and contrast-enhanced mammography but without the need for IV injection and risk of potential adverse events related to the contrast agent.
These results compare favorably with other adjunctive breast US technologies. In the multi-institutional breast elastography trial (BE1 trial) using a threshold of 30kPA maximum, 33.1% (58/175) of benign low suspicion lesions could have been downgraded without false negative results [26]. Using US-guided DOT, 38.3% of low suspicion lesions were appropriately downgraded without downgrading of malignancy to a non-biopsy recommendation. However, unlike US-guided DOT, the accuracy of shearwave elastography is operator dependent, and relies on the proper angle of insonation, degree of breast contact, and region of interest placement [27], which limits its clinical application. In the pivotal trial of optoacoustic imaging, 43.3% (1433/3310) of BI-RADS 4A assessments were correctly downgraded to BI-RADS 2 or 3, but 22.3% (114/511) of malignant masses were incorrectly downgraded to non-actionable assessments [28]. In another study of optoacoustic imaging combined with conventional US, 41.1% (60/146) of benign masses with BI-RADS 4A and 4B assessments were correctly downgraded to Bi-RADS 3, but 4.5% (3/67) of malignant masses were incorrectly downgraded [29].
While readers relied primarily on quantitative HbT values, HbT spatial maps were also qualitatively assessed. Not-oval shape, central location of highest HbT, and tight ring distribution of HbT were significantly more common in cancers, although tight ring distribution of HbT was the only feature independently associated with malignancy. The findings suggest a contributory role of HbT ring distribution in DOT interpretation.
Interreader agreement was somewhat worse for CI&DOT BI-RADS than CI BI-RADS. Nevertheless it was comparable (κ = 0.28–0.44) to interobserver agreement for BI-RADS assessment categories using US (κ = 0.30) [30]. Further studies are warranted of approaches to help optimize radiologists’ reproducibility in DOT evaluation.
This study has limitations. The sample size was restricted to 46 benign lesions and 15 cancers and validation with larger trials is needed. The cancers were relatively large, (median size 1.9 cm), and significantly larger than non-cancers. Non-image guided DOT has been shown to be less predictive in lesions smaller than 0.6 cm; a study of smaller cancers is needed to address US-guided DOT size limitations. Results may not be generalizable, especially given that all readers were experienced breast imagers. Extensive exclusion criteria were applied to ensure that DOT measurements would reflect only the lesion of interest and to ensure technical feasibility. One of the exclusion criteria, normality of the contralateral mirror image site, was not always evaluable because it was not imaged in every patient, which could limit the quality of the reference data. Finally, barriers remain to clinical adoption, including lengthy (~20 min) reconstruction times.
5. Conclusion
In this prospective multi-reader study simulating clinical breast imaging interpretation conditions, the adjunctive use of US-guided DOT resulted in a 23.5% net reduction in suspicious assessments with a 38.3% decrease in BI-RADS 4A assessments, without compromising cancer diagnosis. Among categories of benign lesions, US-guided DOT had its greatest impact in downgrading non-proliferative benign lesions. Additional efforts remain warranted to address radiologists’ reproducibility in application of DOT. Though requiring validation in larger studies, the findings indicate a potential role for adjunctive US-guided DOT in substantially reducing benign breast biopsies.
Supplementary Material
Acknowledgements
The authors appreciate the help of Ruth Holdener and Patricia Fisher, Department of Radiology, Washington University School of Medicine, for patient enrollment and study execution.
Funding:
This study, NCT03842358, was supported by the U.S. National Cancer Institute, R01CA228047. SPP acknowledges funding support from the Foundation for Barnes Jewish Hospital Ronald and Hanna Evens Endowed Chair in Women’s Health.
Abbreviations:
- HbT
average maximum total Hemoglobin concentration
- DOT
Diffuse Optical Tomography
- CI
conventional imaging
Footnotes
CRediT authorship contribution statement
Steven P. Poplack: Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing. Catherine A. Young: Conceptualization, Investigation, Methodology, Project administration, Writing – original draft, Writing – review & editing. Ian S. Hagemann: Investigation, Methodology, Project administration, Resources, Writing – original draft, Writing – review & editing. Jingqin Luo: Conceptualization, Data curation, Formal analysis, Methodology, Writing – original draft, Writing – review & editing. Cheryl R. Herman: Investigation, Writing – review & editing. Kimberly Wiele: Investigation, Writing – review & editing. Shuying Li: Data curation, Formal analysis, Validation, Writing – review & editing. Yifeng Zeng: Data curation, Formal analysis, Validation. Matthew F. Covington: Investigation, Writing – review & editing. Quing Zhu: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Appendix A. Supplementary material
Supplementary data to this article can be found online at https://doi.org/10.1016/j.ejrad.2021.110029.
References
- [1].Sprague BL, Arao RF, Miglioretti DL, Henderson LM, Buist DS, Onega T, Rauscher GH, Lee JM, Tosteson AN, Kerlikowske K, Lehman CD, Breast Cancer Surveillance C, National Performance Benchmarks for Modern Diagnostic Digital Mammography: Update from the Breast Cancer Surveillance Consortium, Radiology 283(1) (2017) 59–69. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [2].Elezaby M, Li G, Bhargavan-Chatfield M, Burnside ES, DeMartini WB, ACR BI-RADS assessment category 4 subdivisions in diagnostic mammography: utilization and outcomes in the national mammography database, Radiology 287 (2) (2018) 416–422. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [3].Varella Miguel Angelo Spinelli, da Cruz Jackson Teixeira, Rauber Andrea, Varella Ivana Santos, Fleck James Freitas, Moreira Luis Fernando, Role of BI-RADS Ultrasound Subcategories 4A to 4C in Predicting Breast Cancer, Clin. Breast Cancer 18 (4) (2018) e507–e511. [DOI] [PubMed] [Google Scholar]
- [4].Lee Hye-Jeong, Kim Eun-Kyung, Min Jung Kim Ji Hyun Youk, Ji Young Lee Dae Ryong Kang, Ki Keun Oh, Observer variability of Breast Imaging Reporting and Data System (BI-RADS) for breast ultrasound, Eur. J. Radiol 65 (2) (2008) 293–298. [DOI] [PubMed] [Google Scholar]
- [5].Elverici E, Barca AN, Aktas H, Ozsoy A, Zengin B, Cavusoglu M, Araz L, Nonpalpable BI-RADS 4 breast lesions: sonographic findings and pathology correlation, Diagn. Interv. Radiol 21 (3) (2015) 189–194. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Tromberg BJ, Cerussi A, Shah N, Compton M, Durkin A, Hsiang D, Butler J, Mehta R, Imaging in breast cancer: diffuse optics in breast cancer: detecting tumors in pre-menopausal women and monitoring neoadjuvant chemotherapy, Breast Cancer Res 7 (6) (2005) 279–285. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [7].Choe Regine, Konecky Soren D., Corlu Alper, Lee Kijoon, Durduran Turgut, Busch David R., Pathak Saurav, Czerniecki Brian J., Tchou Julia, Fraker Douglas L., Angela DeMichele Britton Chance, Arridge Simon R., Schweiger Martin, Culver Joseph P., Schnall Mitchell D., Putt Mary E., Rosen Mark A., Yodh Arjun G., Differentiation of benign and malignant breast tumors by in-vivo three-dimensional parallel-plate diffuse optical tomography, J. Biomed. Opt 14 (2) (2009) 024020, 10.1117/1.3103325. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].Poplack SP, Tosteson TD, Wells WA, Pogue BW, Meaney PM, Hartov A, Kogel CA, Soho SK, Gibson JJ, Paulsen KD, Electromagnetic breast imaging: results of a pilot study in women with abnormal mammograms, Radiology 243 (2) (2007) 350–359. [DOI] [PubMed] [Google Scholar]
- [9].Collettini F, Martin JC, Diekmann F, Fallenberg E, Engelken F, Ponder S, Kroencke TJ, Hamm B, Poellinger A, Diagnostic performance of a Near-Infrared Breast Imaging system as adjunct to mammography versus X-ray mammography alone, Eur. Radiol 22(2) (2012) 350–357. [DOI] [PubMed] [Google Scholar]
- [10].Fang Qianqian, Selb Juliette, Carp Stefan A., Boverman Gregory, Miller Eric L., Brooks Dana H., Moore Richard H., Kopans Daniel B., Boas David A., Combined optical and X-ray tomosynthesis breast imaging, Radiology 258 (1) (2011) 89–97. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [11].Zimmermann Bernhard B., Deng Bin, Singh Bhawana, Martino Mark, Selb Juliette, Fang Qianqian, Sajjadi Amir Y., Cormier Jayne, Moore Richard H., Kopans Daniel B., Boas David A., Saksena Mansi A., Carp Stefan A., Multimodal breast cancer imaging using coregistered dynamic diffuse optical tomography and digital breast tomosynthesis, J. Biomed. Opt 22 (4) (2017) 046008, 10.1117/1.JBO.22.4.046008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [12].Zhu Q, Hegde PU, Ricci A Jr., Kane M, Cronin EB, Ardeshirpour Y, Xu C, Aguirre A, Kurtzman SH, Deckers PJ, Tannenbaum SH, Early-stage invasive breast cancers: potential role of optical tomography with US localization in assisting diagnosis, Radiology 256 (2) (2010) 367–378. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [13].Zhi W, Gu X, Qin J, Yin P, Sheng X, Gao SP, Li Q, Solid breast lesions: clinical experience with US-guided diffuse optical tomography combined with conventional US, Radiology 265 (2) (2012) 371–378. [DOI] [PubMed] [Google Scholar]
- [14].Zhu Quing, Ricci Andrew, Hegde Poornima, Kane Mark, Cronin Edward, Merkulov Alex, Xu Yan, Tavakoli Behnoosh, Tannenbaum Susan, Assessment of Functional Differences in Malignant and Benign Breast Lesions and Improvement of Diagnostic Accuracy by Using US-guided Diffuse Optical Tomography in Conjunction with Conventional US, Radiology 280 (2) (2016) 387–397. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Mastanduno MA, Xu J, El-Ghussein F, Jiang S, Yin H, Zhao Y, Michaelsen KE, Wang K, Ren F, Pogue BW, Paulsen KD, Sensitivity of MRI-guided near-infrared spectroscopy clinical breast exam data and its impact on diagnostic performance, Biomed. Opt. Express 5 (9) (2014) 3103–3115. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [16].Feng J, Xu J, Jiang S, Yin H, Zhao Y, Gui J, Wang K, Lv X, Ren F, Pogue BW, Paulsen KD, Addition of T2-guided optical tomography improves noncontrast breast magnetic resonance imaging diagnosis, Breast Cancer Res. 19 (1) (2017) 117. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [17].Chae EY, Kim HH, Sabir S, Kim Y, Kim H, Yoon S, Ye JC, Cho S, Heo D, Kim KH, Bae YM, Choi YW, Development of digital breast tomosynthesis and diffuse optical tomography fusion imaging for breast cancer detection, Sci. Rep 10 (1) (2020) 13127. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [18].Mastanduno Michael A., Xu Junqing, Fadi El-Ghussein Shudong Jiang, Yin Hong, Zhao Yan, Wang Ke, Ren Fang, Gui Jiang, Pogue Brian W., Paulsen Keith D., MR-Guided Near-Infrared Spectral Tomography Increases Diagnostic Performance of Breast MRI, Clin. Cancer Res 21 (17) (2015) 3906–3912. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [19].Zhu Q, Cronin EB, Currier AA, Vine HS, Huang M, Chen N, Xu C, Benign versus malignant breast masses: optical differentiation with US-guided optical imaging reconstruction, Radiology 237 (1) (2005) 57–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [20].Vavadi H, Mostafa A, Zhou F, Uddin KMS, Althobaiti M, Xu C, Bansal R, Ademuyiwa F, Poplack S, Zhu Q, Compact ultrasound-guided diffuse optical tomography system for breast cancer imaging, J. Biomed. Opt 24 (2) (2018) 1–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [21].Uddin KMS, Mostafa A, Anastasio M, Zhu Q, Two step imaging reconstruction using truncated pseudoinverse as a preliminary estimate in ultrasound guided diffuse optical tomography, Biomed. Opt. Express 8 (12) (2017) 5437–5449. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [22].Mendelson EB, Bohm-Velez M, Berg WA, ACR BI-RADS Ultrasound, ACR BI-RADS Atlas, Breast Imaging Reporting and Data System, American College of Radiology, Reston, VA, 2013, pp. 261–290. [Google Scholar]
- [23].Chen NG, Guo P, Yan S, Piao D, Zhu Q, Simultaneous near-infrared diffusive light and ultrasound imaging, Appl. Opt 40 (34) (2001) 6367–6380. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [24].DeLong ER, DeLong DM, Clarke-Pearson DL, Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach, Biometrics 44 (3) (1988) 837–845. [PubMed] [Google Scholar]
- [25].Bates D, Machler M, Bolker BM, Walker SC, Fitting Linear Mixed-Effects Models Using lme4, J. Statist. Software 67 (1) (2015) 1–48. [Google Scholar]
- [26].Berg WA, Cosgrove DO, Dore CJ, Schafer FK, Svensson WE, Hooley RJ, Ohlinger R, Mendelson EB, Balu-Maestro C, Locatelli M, Tourasse C, Cavanaugh BC, Juhan V, Stavros AT, Tardivon A, Gay J, Henry JP, Cohen-Bacrie C, Investigators BE, Shear-wave elastography improves the specificity of breast US: the BE1 multinational study of 939 masses, Radiology 262 (2) (2012) 435–449. [DOI] [PubMed] [Google Scholar]
- [27].Youk JH, Gweon HM, Son EJ, Shear-wave elastography in breast ultrasonography: the state of the art, Ultrasonography 36 (4) (2017) 300–309. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [28].Neuschler EI, Butler R, Young CA, Barke LD, Bertrand ML, Bohm-Velez M, Destounis S, Donlan P, Grobmyer SR, Katzen J, Kist KA, Lavin PT, Makariou EV, Parris TM, Schilling KJ, Tucker FL, Dogan BE, A pivotal study of optoacoustic imaging to diagnose benign and malignant breast masses: A new evaluation tool for radiologists, Radiology 287 (2) (2018) 398–412. [DOI] [PubMed] [Google Scholar]
- [29].Menezes GLG, Pijnappel RM, Meeuwis C, Bisschops R, Veltman J, Lavin PT, van de Vijver MJ, Mann RM, Downgrading of Breast Masses Suspicious for Cancer by Using Optoacoustic Breast Imaging, Radiology 288(2) (2018) 355–365. [DOI] [PubMed] [Google Scholar]
- [30].Abdullah N, Mesurolle B, El-Khoury M, Kao E, Breast imaging reporting and data system lexicon for US: interobserver agreement for assessment of breast masses, Radiology 252 (3) (2009) 665–672. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.