Differing benefits of artificial intelligence-based computer-aided diagnosis for breast US according to workflow and experience level

Purpose@#This study evaluated how artificial intelligence-based computer-assisted diagnosis (AICAD) for breast ultrasonography (US) influences diagnostic performance and agreement between radiologists with varying experience levels in different workflows. @*Methods@#Images of 492 breast lesions (200 malignant and 292 benign masses) in 472 women taken from April 2017 to June 2018 were included. Six radiologists (three inexperienced [<1 year of experience] and three experienced [10-15 years of experience]) individually reviewed US images with and without the aid of AI-CAD, first sequentially and then simultaneously. Diagnostic performance and interobserver agreement were calculated and compared between radiologists and AI-CAD. @*Results@#After implementing AI-CAD, the specificity, positive predictive value (PPV), and accuracy significantly improved, regardless of experience and workflow (all P<0.001, respectively). The overall area under the receiver operating characteristic curve significantly increased in simultaneous reading, but only for inexperienced radiologists. The agreement for Breast Imaging Reporting and Database System (BI-RADS) descriptors generally increased when AI-CAD was used (κ=0.29-0.63 to 0.35-0.73). Inexperienced radiologists tended to concede to AI-CAD results more easily than experienced radiologists, especially in simultaneous reading (P<0.001). The conversion rates for final assessment changes from BI-RADS 2 or 3 to BI-RADS higher than 4a or vice versa were also significantly higher in simultaneous reading than sequential reading (overall, 15.8% and 6.2%, respectively; P<0.001) for both inexperienced and experienced radiologists. @*Conclusion@#Using AI-CAD to interpret breast US improved the specificity, PPV, and accuracy of radiologists regardless of experience level. AI-CAD may work better in simultaneous reading to improve diagnostic performance and agreement between radiologists, especially for inexperienced radiologists.

Application of machine learning to ultrasound images to differentiate follicular neoplasms of the thyroid gland

Purpose@#This study was conducted to evaluate the diagnostic performance of machine learning in differentiating follicular adenoma from carcinoma using preoperative ultrasonography (US). @*Methods@#In this retrospective study, preoperative US images of 348 nodules from 340 patients were collected from two tertiary referral hospitals. Two experienced radiologists independently reviewed each image and categorized the nodules according to the 2015 American Thyroid Association guideline. Categorization of a nodule as highly suspicious was considered a positive diagnosis for malignancy. The nodules were manually segmented, and 96 radiomic features were extracted from each region of interest. Ten significant features were selected and used as final input variables in our in-house developed classifier models based on an artificial neural network (ANN) and support vector machine (SVM). The diagnostic performance of radiologists and both classifier models was calculated and compared. @*Results@#In total, 252 nodules from 245 patients were confirmed as follicular adenoma and 96 nodules from 95 patients were diagnosed as follicular carcinoma. As measures of diagnostic performance, the average sensitivity, specificity, and accuracy of the two experienced radiologists in discriminating follicular adenoma from carcinoma on preoperative US images were 24.0%, 84.0%, and 64.8%, respectively. The sensitivity, specificity, and accuracy of the ANN and SVM-based models were 32.3%, 90.1%, and 74.1% and 41.7%, 79.4%, and 69.0%, respectively. The kappa value of the two radiologists was 0.076, corresponding to slight agreement. @*Conclusion@#Machine learning-based classifier models may aid in discriminating follicular adenoma from carcinoma using preoperative US.

Scoring System to Stratify Malignancy Risks for Mammographic Microcalcifications Based on Breast Imaging Reporting and Data System 5th Edition Descriptors

OBJECTIVE: To develop a scoring system stratifying the malignancy risk of mammographic microcalcifications using the 5th edition of the Breast Imaging Reporting and Data System (BI-RADS).MATERIALS AND METHODS: One hundred ninety-four lesions with microcalcifications for which surgical excision was performed were independently reviewed by two radiologists according to the 5th edition of BI-RADS. Each category's positive predictive value (PPV) was calculated and a scoring system was developed using multivariate logistic regression. The scores for benign and malignant lesions or BI-RADS categories were compared using an independent t test or by ANOVA. The area under the receiver operating characteristic curve (AUROC) was assessed to determine the discriminatory ability of the scoring system. Our scoring system was validated using an external dataset.RESULTS: After excision, 69 lesions were malignant (36%). The PPV of BI-RADS descriptors and categories for calcification showed significant differences. Using the developed scoring system, mean scores for benign and malignant lesions or BI-RADS categories were significantly different (p < 0.001). The AUROC of our scoring system was 0.874 (95% confidence interval, 0.840–0.909) and the PPV of each BI-RADS category determined by the scoring system was as follows: category 3 (0%), 4A (6.8%), 4B (19.0%), 4C (68.2%), and 5 (100%). The validation set showed an AUROC of 0.905 and PPVs of 0%, 8.3%, 11.9%, 68.3%, and 94.7% for categories 3, 4A, 4B, 4C, and 5, respectively.CONCLUSION: A scoring system based on BI-RADS morphology and distribution descriptors could be used to stratify the malignancy risk of mammographic microcalcifications.

Identification of Preoperative Magnetic Resonance Imaging Features Associated with Positive Resection Margins in Breast Cancer: A Retrospective Study

OBJECTIVE: To determine which preoperative breast magnetic resonance imaging (MRI) findings and clinicopathologic features are associated with positive resection margins at the time of breast-conserving surgery (BCS) in patients with breast cancer. MATERIALS AND METHODS: We reviewed preoperative breast MRI and clinicopathologic features of 120 patients (mean age, 53.3 years; age range, 27–79 years) with breast cancer who had undergone BCS in 2015. Tumor size on MRI, multifocality, patterns of enhancing lesions (mass without non-mass enhancement [NME] vs. NME with or without mass), mass characteristics (shape, margin, internal enhancement characteristics), NME (distribution, internal enhancement patterns), and breast parenchymal enhancement (BPE; weak, strong) were analyzed. We also evaluated age, tumor size, histology, lymphovascular invasion, T stage, N stage, and hormonal receptors. Univariate and multivariate logistic regression analyses were used to determine the correlation between clinicopathological features, MRI findings, and positive resection margins. RESULTS: In univariate analysis, tumor size on MRI, multifocality, NME with or without mass, and segmental distribution of NME were correlated with positive resection margins. Among the clinicopathological factors, tumor size of the invasive breast cancer and in situ components were significantly correlated with a positive resection margin. Multivariate analysis revealed that NME with or without mass was an independent predictor of positive resection margins (odds ratio [OR] = 7.00; p < 0.001). Strong BPE was a weak predictor of positive resection margins (OR = 2.59; p = 0.076). CONCLUSION: Non-mass enhancement with or without mass is significantly associated with a positive resection margin in patients with breast cancer. In patients with NME, segmental distribution was significantly correlated with positive resection margins.

Ex Vivo Shear-Wave Elastography of Axillary Lymph Nodes to Predict Nodal Metastasis in Patients with Primary Breast Cancer / 한국유방암학회지

PURPOSE: There is still a clinical need to easily evaluate the metastatic status of lymph nodes during breast cancer surgery. We hypothesized that ex vivo shear-wave elastography (SWE) would predict precisely the presence of metastasis in the excised lymph nodes. METHODS: A total of 63 patients who underwent breast cancer surgery were prospectively enrolled in this study from May 2014 to April 2015. The excised axillary lymph nodes were examined using ex vivo SWE. Metastatic status was confirmed based on the final histopathological diagnosis of the permanent section. Lymph node characteristics and elasticity values measured by ex vivo SWE were assessed for possible association with nodal metastasis. RESULTS: A total of 274 lymph nodes, harvested from 63 patients, were examined using ex vivo SWE. The data obtained from 228 of these nodes from 55 patients were included in the analysis. Results showed that 187 lymph nodes (82.0%) were nonmetastatic and 41 lymph nodes (18.0%) were metastatic. There was significant difference between metastatic and nonmetastatic nodes with respect to the mean (45.4 kPa and 17.7 kPa, p<0.001) and maximum (55.3 kPa and 23.2 kPa, p<0.001) stiffness. The elasticity ratio was higher in the metastatic nodes (4.36 and 1.57, p<0.001). Metastatic nodes were significantly larger than nonmetastatic nodes (mean size, 10.5 mm and 7.5 mm, p<0.001). The size of metastatic nodes and nodal stiffness were correlated (correlation coefficient of mean stiffness, r=0.553). The area under curve of mean stiffness, maximum stiffness, and elasticity ratio were 0.794, 0.802, and 0.831, respectively. CONCLUSION: Ex vivo SWE may be a feasible method to predict axillary lymph node metastasis intraoperatively in patients undergoing breast cancer surgery.

Shear-wave elastography in breast ultrasonography: the state of the art

Shear-wave elastography (SWE) is a recently developed ultrasound technique that can visualize and measure tissue elasticity. In breast ultrasonography, SWE has been shown to be useful for differentiating benign breast lesions from malignant breast lesions, and it has been suggested that SWE enhances the diagnostic performance of ultrasonography, potentially improving the specificity of conventional ultrasonography using the Breast Imaging Reporting and Data System criteria. More recently, not only has SWE been proven useful for the diagnosis of breast cancer, but has also been shown to provide valuable information that can be used as a preoperative predictor of the prognosis or response to chemotherapy.

Thyroid nodules with nondiagnostic results on repeat fine-needle aspiration biopsy: which nodules should be considered for repeat biopsy or surgery rather than follow-up?

PURPOSE: The goal of this study was to assess the clinicopathologic and ultrasonographic features of thyroid nodules with nondiagnostic results on repeat ultrasonography (US)-guided fineneedle aspiration biopsy (FNAB) according to size and the number of suspicious findings and to determine the proper management of nodules with consecutive nondiagnostic results. METHODS: This retrospective study included 297 nodules with nondiagnostic results on repeat FNAB that were evaluated by US over the course of at least 12 months of follow-up, a follow-up biopsy, or an operation. We compared clinical and US variables between benign and malignant nodules in thyroid nodules with repeat nondiagnostic results. RESULTS: The comparison of benign and malignant nodules with repeat nondiagnostic results revealed that age, marked hypoechogenicity, irregular or microlobulated margins, microcalcifications, and nonparallel shape were significantly associated with malignancy. Multivariate logistic regression analysis in malignant nodules revealed that microcalcifications and irregular or microlobulated margins were independently associated with malignancy. Among them, only irregular or microlobulated margins were independently significant as a predictor of malignancy in repeatedly nondiagnostic nodules measuring >10 mm. Using receiver operating characteristic analysis, the best cutoff value for the "number of suspicious findings" between benign and malignant nodules was three in nodules of all sizes, three in nodules measuring ≤10 mm, and two in nodules measuring >10 mm. CONCLUSION: Irregular or microlobulated margins may be the most frequent US features in repeatedly nondiagnostic nodules >10 mm. The presence of "two or more suspicious findings" can be used as the cutoff for distinguishing benign and malignant nodules.

Validation of the fifth edition BI-RADS ultrasound lexicon with comparison of fourth and fifth edition diagnostic performance using video clips

PURPOSE: The aim of this study was to evaluate the positive predictive value (PPV) and the diagnostic performance of the ultrasonographic descriptors in the fifth edition of BI-RADS, comparing with the fourth edition using video clips. METHODS: From September 2013 to July 2014, 80 breast masses in 74 women (mean age, 47.5±10.7 years) from five institutions of the Korean Society of Breast Imaging were included. Two radiologists individually reviewed the static and video images and analyzed the images according to the fourth and fifth edition of BI-RADS. The PPV of each descriptor was calculated and diagnostic performances between the fourth and fifth editions were compared. RESULTS: Of the 80 breast masses, 51 (63.8%) were benign and 29 (36.2%) were malignant. Suspicious ultrasonographic features such as irregular shape, non-parallel orientation, angular or spiculated margins, and combined posterior features showed higher PPV in both editions (all P0.05). The area under the receiver operating characteristics curve was higher in the fourth edition (0.708 to 0.690), without significance (P=0.416). CONCLUSION: The fifth edition of the BI-RADS ultrasound lexicon showed comparable performance to the fourth edition and can be useful in the differential diagnosis of breast masses using ultrasonography.

Lymphangiogenesis in Breast Cancer Correlates with Matrix Stiffness on Shear-Wave Elastography

PURPOSE: To correlate tumor stiffness and lymphangiogenesis in breast cancer and to find its clinical implications. MATERIALS AND METHODS: A total of 140 breast cancer patients were evaluated. Tumor stiffness was quantitatively measured by shear-wave elastography in preoperative ultrasound examination, calculated as mean elasticity value (kPa). Slides of resected breast cancer specimens were reviewed for most fibrotic area associated with tumor. D2-40 immunohistochemical staining was applied for fibrotic areas to detect the lymphatic spaces. Microlymphatic density, tumor stiffness, and clinicopathologic data were analyzed. RESULTS: Higher elasticity value was associated with invasive size of tumor, microlymphatic density, histologic grade 3, absence of extensive intraductal component, presence of axillary lymph node metastasis, and Ki-67 labeling index (LI) in univariate regression analysis, and associated with Ki-67 LI and axillary lymph node metastasis in multivariate regression analysis. Microlymphatic density was associated histologic grade 3, mean elasticity value, and Ki-67 LI in univariate regression analysis. In multivariate regression analysis, microlymphatic density was correlated with mean elasticity value. CONCLUSION: In breast cancer, tumor stiffness correlates with lymphangiogenesis and poor prognostic factors.

Factors Influencing the Background Parenchymal Enhancement in Follow-Up Breast MRI after Adjuvant Endocrine Therapy

PURPOSE: To investigate factors influencing the evaluation of background parenchymal enhancement (BPE) at follow-up breast magnetic resonance imaging (MRI) after adjuvant endocrine therapy. MATERIALS AND METHODS: One hundred twelve women with breast cancer and MRI of the contralateral unaffected breast before and after endocrine therapy were identified. Two readers in consensus performed blinded side-by-side comparison of BPE (minimal, mild, moderate, and marked) before and after therapy with categorical scales. Age, body mass index, menopausal status, treatment regimen (selective estrogen receptor modulator or aromatase inhibitor), chemotherapy, follow-up duration, BPE at baseline MRI, MRI field strength before and after therapy, and recurrence were analyzed for their influences on decreased BPE. RESULTS: Younger age, premenopausal status, treatment with selective estrogen receptor modulator, MRI field strength, and moderate or marked baseline BPE were significantly associated with decreased BPE. In multivariate analysis, MRI field strength and baseline BPE showed a significant association. CONCLUSION: MRI field strength and baseline BPE before and after therapy were associated with decreased BPE at post-therapy, follow-up MRI.

Comparison of hormonal receptor and HER2 status between ultrasound-guided 14-gauge core needle biopsy and surgery in breast cancer patients

PURPOSE: To evaluate the concordance of estrogen receptor, progesterone receptor, and human epidermal growth factor receptor 2 (HER2) statuses between ultrasound (US)-guided 14-gauge core needle biopsy (CNB) and surgery and to analyze whether the clinicopathological and imaging features including those from mammography and ultrasonography can predict the concordance in breast cancer patients. METHODS: The concordance of receptor status between CNB and surgery was assessed for 55 breast cancers in 55 women who underwent CNB before treatment. The clinicopathological and imaging features and the concordance rates were compared between the non-neoadjuvant chemotherapy (non-NAC) group and the NAC group according to the initial treatment. The concordance rates were analyzed according to the clinicopathological and imaging features, by using the chi-square or Fisher exact test and McNemar test for the categorical and the independent t-test for continuous variables. RESULTS: Among 55 women, 22 women (40%) were part of the non-NAC group and 33 women (60%) were part of the NAC group. The concordance rates were 0.86-1.00 in the non-NAC group and 0.76-0.88 in the NAC group. In all three receptors, the difference in the concordance rate between the two groups was not significant. In the NAC group, the absence of axillary lymph node metastasis (1.00, P=0.02) and visibility of cancer on mammography (0.93, P=0.04) showed the higher concordance of the HER2 status. CONCLUSION: Concordance of the receptor status between surgery and US-guided 14-gauge CNB was feasible in breast cancer patients. The absence of axillary lymph node metastasis after NAC and the visibility of cancer on mammography prior to NAC may be helpful for predicting the concordance of HER2 in breast cancer patients.

Shear-wave elastography for breast masses: local shear wave speed (m/sec) versus Young modulus (kPa)

PURPOSE: To evaluate and compare the performance of shear-wave elastography (SWE) for breast masses using the local shear wave speed (m/sec) vs. Young modulus (kPa). METHODS: A total of 130 breast lesions in 123 women who underwent SWE before ultrasound-guided core needle biopsy or surgical excision were included. With the region-of-interest placed over the stiffest areas of the lesion on SWE, the quantitative mean, maximum, and standard deviation (SD) of the elasticity values were measured in kPa and m/sec for each lesion. The SD was also measured with the region-of-interest including the whole breast lesion (wSD). The area under the receiver operating characteristic curve (AUC), sensitivity, and specificity of each elasticity value measured in kPa and m/sec were compared. RESULTS: Of the 130 lesions, 49 (37.7%) were malignant and 81 (62.3%) were benign. The AUCs for the mean, maximum, and SD of the elasticity values using kPa and m/sec did not differ significantly: mean, 0.974 vs. 0.974; maximum, 0.960 vs. 0.976; SD, 0.916 vs. 0.916. However, the AUC for wSD showed a significant difference: 0.964 (kPa) vs. 0.960 (m/sec) (P=0.036). There was no significant difference in the sensitivity and specificity of the mean, maximum, and wSD of the elasticity values. However, the specificity of the SD was significantly different between the two different measurements: 95.1% (kPa) vs. 87.7% (m/sec) (P=0.031). CONCLUSION: The quantitative elasticity values measured in kPa and m/sec on SWE showed good diagnostic performance. The specificity of the SD and AUC of the wSD measured in kPa were significantly higher than those measured in m/sec.

Practice guideline for the performance of breast ultrasound elastography

Ultrasound (US) elastography is a valuable imaging technique for tissue characterization. Two main types of elastography, strain and shear-wave, are commonly used to image breast tissue. The use of elastography is expected to increase, particularly with the increased use of US for breast screening. Recently, the US elastographic features of breast masses have been incorporated into the 2nd edition of the Breast Imaging Reporting and Data System (BI-RADS) US lexicon as associated findings. This review suggests practical guidelines for breast US elastography in consensus with the Korean Breast Elastography Study Group, which was formed in August 2013 to perform a multicenter prospective study on the use of elastography for US breast screening. This article is focused on the role of elastography in combination with B-mode US for the evaluation of breast masses. Practical tips for adequate data acquisition and the interpretation of elastography results are also presented.

Thyroid Nodules with Macrocalcification: Sonographic Findings Predictive of Malignancy

PURPOSE: To analyze which sonographic features of thyroid nodules with macrocalcifications were predictable of thyroid malignancy. MATERIALS AND METHODS: We reviewed sonographic findings of 854 macrocalcified thyroid nodules in patients who underwent fine needle aspiration biopsy between December 2009 and January 2011. There were 171 non-diagnostic aspirations, 34 nodules with category 3, 4, 5 based on Bethesda system, which were not confirmed by surgery, and these nodules were excluded from the analysis. Sonographic characteristics of the macrocalcifications including its thickness, interruption, and existence of soft tissue rim outside the macrocalcification were analyzed. Other sonographic characteristics of nodules such as shape, margin, composition, echo pattern, vascularity, and underlying parenchymal echogenicity were also evaluated. The correlation of sonographic features with cytopathologic results and the diagnostic performance of sonographic features for the prediction of malignancy were analyzed. RESULTS: Among 649 nodules, 179 (27.6%) nodules were malignant and 470 (72.4%) nodules were benign. Among the features of the macrocalcification, interruption, irregular thickness, or the presence of soft tissue outside calcification rim were associated with malignancy (p<0.001). A high sensitivity and negative predictive values for the prediction of malignancy was found in sonographic characteristics of irregular thickness (92.2% and 91.0%, respectively) and the presence of soft tissue (88.5% and 88.8%, respectively). CONCLUSION: Sonographic characteristics of macrocalcification such as interruption, irregular thickness and the presence of soft tissue rim were associated with malignancy in thyroid nodules with macrocalcifications.

Thyroid nodules with Bethesda System IV Cytology: Can Ultrasonography Differentiate Malignancy from Benign?

PURPOSE: The purpose of this study is to identify clinical and imaging parameters that can be used in differentiation of benign versus malignancy of preoperative FNA diagnosis of Bethesda system IV nodules. MATERIALS AND METHODS: We analyzed clinical, ultrasonographic, and CT findings of 28 thyroid nodules with Bethesda system IV cytology on FNA, which were proven as follicular or Hurthle cell neoplasms on surgical pathology. RESULTS: No statistically significant differences according to age, sex, and ultrasonographic parameters, including echogenicity, margin, calcification, shape, cystic component, and degree of vascularity and enhancement on CT were observed between benign and malignant follicular neoplasms. Only the lesion size was significantly different (p<0.05). CONCLUSION: The size of follicular neoplasm is predictive of malignancy. If a thyroid nodule with the Bethesda IV cytology is larger than 24.5 mm, there will be a greater probability of malignancy.

Current Trends in Breast Ultrasonography

In the last 30 years, technical improvements have directly contributed to expanding sonographic breast imaging applications into the most important adjunctive imaging modality for breast evaluation. Most of these advances have related to improvements in sonographic gray-scale image resolution and contrast, but some applications such as color Doppler imaging, elastography, optical imaging, or three-dimensional ultrasounds have resulted in sonographic information that is uniquely different from grayscale imaging. The current spectrum of new breast sonographic techniques not only offers information uniquely different from gray-scale imaging but also involves hardware advances that affect the method of image production. In this article, we discuss the current trends in breast ultrasonography focusing on the advances to further improve accuracy for breast lesion diagnosis.

Concordant or Discordant? Imaging-Pathology Correlation in a Sonography-Guided Core Needle Biopsy of a Breast Lesion

An imaging-guided core needle biopsy has been proven to be reliable and accurate for the diagnosis of both benign and malignant diseases of the breast, and has replaced surgical biopsy. However, the possibility of a false-negative biopsy still remains. Imaging-pathology correlation is of critical importance in imaging-guided breast biopsies to detect such a possible sampling error and avoid a delay in diagnosis. We will review five possible categories and corresponding management after performing an imaging-pathology correlation in a sonography-guided core needle biopsy of a breast lesion, as well as illustrate the selected images for each category in conjunction with the pathologic finding. Radiologists should be familiar with the imaging features of various breast pathologies and be able to appropriately correlate imaging findings with pathologic results after a core needle biopsy.

Supplementary Screening Sonography in Mammographically Dense Breast: Pros and Cons

Sonography is an attractive supplement to mammography in breast cancer screening because it is relatively inexpensive, requires no contrast-medium injection, is well tolerated by patients, and is widely available for equipment as compared with MRI. Sonography has been especially valuable for women with mammographically dense breast because it has consistently been able to detect a substantial number of cancers at an early stage. Despite these findings, breast sonography has known limitations as a screening tool; operator-dependence, the shortage of skilled operators, the inability to detect microcalcifications, and substantially higher false-positive rates than mammography. Further study of screening sonography is still ongoing and is expected to help establish the role of screening sonography.

Clear Cell Hidradenoma of the Axilla: a Case Report with Literature Review

Clear cell hidradenoma is an uncommon benign skin appendageal tumor that typically involves the dermal layer of the head, face, and extremities. The breast is a rare site for this lesion, with only two documented cases, which were determined based on mammogram and sonogram findings. We present a case of clear cell hidradenoma of the axillary tail with radiological findings and a literature review.

Secretory Carcinoma of the Breast: A Case Report

A secretory carcinoma of the breast is a rare, but clinically and histologically a distinct variant of the invasive ductal carcinoma, which has a slow growth pattern and a favorable prognosis. Few studies have reported the radiologic findings involved in a secretory carcinoma of the breast. We report the imaging findings of a secretory carcinoma of the breast in a 48-year-old woman, which were similar to those of a benign mass found on a mammography and sonography.