Ultrasound features for prediction of long-term outcomes of women with primary breast cancer <20 mm.

Background: Some women die despite the favorable prognosis of small breast cancers. Breast ultrasound features may reflect pathological and biological characteristics of a breast tumor. This study aimed to explore whether ultrasound features could identify small breast cancers with poor outcomes. Methods: This retrospective study examined confirmed breast cancers with a size of <20 mm diagnosed in our hospital between 02/2008 and 08/2019. Clinicopathological and ultrasound features were compared between alive and deceased breast cancer patients. Survival was analyzed using the Kaplan-Meier curves. Multivariable Cox proportional hazards models were used to examine the factors associated with breast cancer-specific survival (BCSS) and disease-free survival (DFS). Results: Among the 790 patients, the median follow-up was 3.5 years. The deceased group showed higher frequencies of spiculated (36.7% vs. 11.2%, P<0.001), anti-parallel orientation (43.3% vs. 15.4%, P<0.001), and spiculated morphology combined with anti-parallel orientation (30.0% vs. 2.4%, P<0.001). Among 27 patients with spiculated morphology and anti-parallel orientation, nine cancer-specific deaths and 11 recurrences occurred, for a 5-year BCSS of 77.8% and DFS of 66.7%, while 21 breast-cancer deaths and 41 recurrences occurred among the remaining patients with higher 5-year BCSS (97.8%, P<0.001) and DFS (95.4%, P<0.001). Spiculated and anti-parallel orientation (HR=7.45, 95%CI: 3.26-17.00; HR=6.42, 95%CI: 3.19-12.93), age ≥55 years (HR=5.94, 95%CI: 2.24-15.72; HR=1.98, 95%CI: 1.11-3.54), and lymph nodes metastasis (HR=3.99, 95%CI: 1.89-8.43; HR=2.99, 95%CI: 1.71-5.23) were independently associated with poor BCSS and DFS. Conclusions: Spiculated and anti-parallel orientation at ultrasound are associated with poor BCSS and DFS in patients with primary breast cancer <20 mm.

Deep learning radiomics of ultrasonography for differentiating sclerosing adenosis from breast cancer.

OBJECTIVES: The purpose of our study is to present a method combining radiomics with deep learning and clinical data for improved differential diagnosis of sclerosing adenosis (SA)and breast cancer (BC). METHODS: A total of 97 patients with SA and 100 patients with BC were included in this study. The best model for classification was selected from among four different convolutional neural network (CNN) models, including Vgg16, Resnet18, Resnet50, and Desenet121. The intra-/inter-class correlation coefficient and least absolute shrinkage and selection operator method were used for radiomics feature selection. The clinical features selected were patient age and nodule size. The overall accuracy, sensitivity, specificity, Youden index, positive predictive value, negative predictive value, and area under curve (AUC) value were calculated for comparison of diagnostic efficacy. RESULTS: All the CNN models combined with radiomics and clinical data were significantly superior to CNN models only. The Desenet121+radiomics+clinical data model showed the best classification performance with an accuracy of 86.80%, sensitivity of 87.60%, specificity of 86.20% and AUC of 0.915, which was better than that of the CNN model only, which had an accuracy of 85.23%, sensitivity of 85.48%, specificity of 85.02%, and AUC of 0.870. In comparison, the diagnostic accuracy, sensitivity, specificity, and AUC value for breast radiologists were 72.08%, 100%, 43.30%, and 0.716, respectively. CONCLUSIONS: A combination of the CNN-radiomics model and clinical data could be a helpful auxiliary diagnostic tool for distinguishing between SA and BC.

Multi-parameter ultrasonography-based predictive model for breast cancer diagnosis.

Objectives: To develop, validate, and evaluate a predictive model for breast cancer diagnosis using conventional ultrasonography (US), shear wave elastography (SWE), and contrast-enhanced US (CEUS). Materials and methods: This retrospective study included 674 patients with 674 breast lesions. The data, a main and an independent datasets, were divided into three cohorts. Cohort 1 (80% of the main dataset; n = 448) was analyzed by logistic regression analysis to identify risk factors and establish the predictive model. The area under the receiver operating characteristic curve (AUC) was analyzed in Cohort 2 (20% of the main dataset; n = 119) to validate and in Cohort 3 (the independent dataset; n = 107) to evaluate the predictive model. Results: Multivariable regression analysis revealed nine independent breast cancer risk factors, including age > 40 years; ill-defined margin, heterogeneity, rich blood flow, and abnormal axillary lymph nodes on US; enhanced area enlargement, contrast agent retention, and irregular shape on CEUS; mean SWE higher than the cutoff value (P < 0.05 for all). The diagnostic performance of the model was good, with AUC values of 0.847, 0.857, and 0.774 for Cohorts 1, 2, and 3, respectively. The model increased the diagnostic specificity (from 31% to 81.3% and 7.3% to 73.1% in cohorts 2 and 3, respectively) without a significant loss in sensitivity (from 100.0% to 90.1% and 100.0% to 81.8% in cohorts 2 and 3, respectively). Conclusion: The multi-parameter US-based model showed good performance in breast cancer diagnosis, improving specificity without a significant loss in sensitivity. Using the model could reduce unnecessary biopsies and guide clinical diagnosis and treatment.

Preoperative Non-Invasive Prediction of Breast Cancer Molecular Subtypes With a Deep Convolutional Neural Network on Ultrasound Images.

Purpose: This study aimed to develop a deep convolutional neural network (DCNN) model to classify molecular subtypes of breast cancer from ultrasound (US) images together with clinical information. Methods: A total of 1,012 breast cancer patients with 2,284 US images (center 1) were collected as the main cohort for training and internal testing. Another cohort of 117 breast cancer cases with 153 US images (center 2) was used as the external testing cohort. Patients were grouped according to thresholds of nodule sizes of 20 mm and age of 50 years. The DCNN models were constructed based on US images and the clinical information to predict the molecular subtypes of breast cancer. A Breast Imaging-Reporting and Data System (BI-RADS) lexicon model was built on the same data based on morphological and clinical description parameters for diagnostic performance comparison. The diagnostic performance was assessed through the accuracy, sensitivity, specificity, Youden's index (YI), and area under the receiver operating characteristic curve (AUC). Results: Our DCNN model achieved better diagnostic performance than the BI-RADS lexicon model in differentiating molecular subtypes of breast cancer in both the main cohort and external testing cohort (all p < 0.001). In the main cohort, when classifying luminal A from non-luminal A subtypes, our model obtained an AUC of 0.776 (95% CI, 0.649-0.885) for patients older than 50 years and 0.818 (95% CI, 0.726-0.902) for those with tumor sizes ≤20 mm. For young patients ≤50 years, the AUC value of our model for detecting triple-negative breast cancer was 0.712 (95% CI, 0.538-0.874). In the external testing cohort, when classifying luminal A from non-luminal A subtypes for patients older than 50 years, our DCNN model achieved an AUC of 0.686 (95% CI, 0.567-0.806). Conclusions: We employed a DCNN model to predict the molecular subtypes of breast cancer based on US images. Our model can be valuable depending on the patient's age and nodule sizes.

A Convolutional Neural Network Based on Ultrasound Images of Primary Breast Masses: Prediction of Lymph-Node Metastasis in Collaboration With Classification of Benign and Malignant Tumors.

Purpose: A convolutional neural network (CNN) can perform well in either of two independent tasks [classification and axillary lymph-node metastasis (ALNM) prediction] based on breast ultrasound (US) images. This study is aimed to investigate the feasibility of performing the two tasks simultaneously. Methods: We developed a multi-task CNN model based on a self-built dataset containing 5911 breast US images from 2131 patients. A hierarchical loss (HL) function was designed to relate the two tasks. Sensitivity, specificity, accuracy, precision, F1-score, and analyses of receiver operating characteristic (ROC) curves and heatmaps were calculated. A radiomics model was built by the PyRadiomics package. Results: The sensitivity, specificity and area under the ROC curve (AUC) of our CNN model for classification and ALNM tasks were 83.5%, 71.6%, 0.878 and 76.9%, 78.3%, 0.836, respectively. The inconsistency error of ALNM prediction corrected by HL function decreased from 7.5% to 4.2%. Predictive ability of the CNN model for ALNM burden (≥3 or ≥4) was 77.3%, 62.7%, and 0.752, and 66.6%, 76.8%, and 0.768, respectively, for sensitivity, specificity and AUC. Conclusion: The proposed multi-task CNN model highlights its novelty in simultaneously distinguishing breast lesions and indicating nodal burden through US, which is valuable for "personalized" treatment.

Ultrasonic multimodality imaging features and the classification value of nonpuerperal mastitis.

OBJECTIVE: To explore the value of ultrasonic multimodality imaging for characterizing nonpuerperal mastitis (NPM) lesions and feasibility of distinguishing different subtypes. METHODS: Thirty-eight NPM lesions were assessed using conventional ultrasonography (US), strain elastography (SE), and contrast-enhanced ultrasound (CEUS). The lesions were confirmed pathologically and classified as granulomatous lobular mastitis (GLM), plasma cell mastitis (PCM), or nonspecific mastitis (NSM). Furthermore, diagnostic indicators were evaluated. The diagnostic performances of the modalities were compared using the area under the receiver operating characteristic curve (AUC). RESULTS: The overall morphological features on US differed significantly between the GLM and PCM groups (p = 0.002). Lesion size (≤10 mm) (p = 0.003) and mean SE score (p = 0.001) differed significantly between the PCM and NSM groups. The frequent NPM characteristic on CEUS was hyperenhancement with (or without) increased lesion size; intergroup differences were not significant. Breast Imaging Reporting and Data System > 3 was considered to indicate malignancy; accordingly, the accuracy of US alone, US with CEUS, and US with SE was 10.5%, 21.1%, and 65.8%, respectively. Moreover, the AUC for US with SE for classifying GLM and PCM was 0.616. CONCLUSION: CEUS cannot accurately classify NPM subtypes, while US and SE are valuable for classification.

The association between conventional ultrasound and contrast-enhanced ultrasound appearances and pathological features in small breast cancer.

OBJECTIVE: To investigate the association between ultrasound appearances and pathological features in small breast cancer. MATERIALS AND METHODS: A total of 186 small breast cancers in 186 patients were analyzed in this retrospective study from January 2015 to December 2019 according to pathological results. Forty-seven cases of axillary lymph node metastasis were found. All patients underwent radical axillary surgery following conventional ultrasound (US) and contrast-enhanced ultrasound (CEUS) examinations. The association between ultrasound appearances and pathological features was analyzed using univariate distributions and multivariate analysis. Then, a logistic regression model was established using the pathological diagnosis of lymph node metastasis and biochemical indicators as the dependent variable and the ultrasound appearances as independent variables. RESULTS: In small breast cancer, risk factors of axillary lymph node metastasis were crab claw-like enhancement on CEUS and abnormal axillary lymph nodes on US. The logistic regression model was established as follows: (axillary lymph node metastasis) = 1.100×(crab claw-like enhancement of CEUS) + 2.749×(abnormal axillary lymph nodes of US) -5.790. In addition, irregular shape on CEUS and posterior echo attenuation on US were risk factors for both positive estrogen receptor and progesterone receptor expression, whereas calcification on US was a risk factor for positive Her-2 expression. A specific relationship could be found using the following logistic models: (positive ER expression) = 1.367×(irregular shape of CEUS) + 1.441×(posterior echo attenuation of US) -5.668; (positive PR expression) = 1.265×(irregular shape of CEUS) + 1.136×(posterior echo attenuation of US) -4.320; (positive Her-2 expression) = 1.658×(calcification of US) -0.896. CONCLUSION: Logistic models were established to provide significant value for the prediction of pre-operative lymph node metastasis and positive biochemical indicators, which may guide clinical treatment.

LMO2 upregulation due to AR deactivation in cancer-associated fibroblasts induces non-cell-autonomous growth of prostate cancer after androgen deprivation.

The androgen receptor (AR) is expressed in prostate fibroblasts in addition to normal prostate epithelial cells and prostate cancer (PCa) cells. Moreover, AR activation in fibroblasts dramatically influences prostate cancer (PCa) cell behavior. Androgen deprivation leads to deregulation of AR downstream target genes in both fibroblasts and PCa cells. Here, we identified LIM domain only 2 (LMO2) as an AR target gene in prostate fibroblasts using ChIP-seq and revealed that LMO2 can be repressed directly by AR through binding to androgen response elements (AREs), which results in LMO2 overexpression after AR deactivation due to normal prostate fibroblasts to cancer-associated fibroblasts (CAFs) transformation or androgen deprivation therapy. Next, we investigated the mechanisms of LMO2 overexpression in fibroblasts and the role of this event in non-cell-autonomous promotion of PCa cells growth in the androgen-independent manner through paracrine release of IL-11 and FGF-9. Collectively, our data suggest that AR deactivation deregulates LMO2 expression in prostate fibroblasts, which induces castration resistance in PCa cells non-cell-autonomously through IL-11 and FGF-9.

Conventional and contrast-enhanced ultrasound features in sclerosing adenosis and correlation with pathology.

OBJECTIVES: To evaluate the efficacy of conventional ultrasound (US) and contrast-enhanced ultrasound (CEUS) in differential diagnosis of sclerosing adenosis (SA) from malignance and investigate the correlated features with pathology. METHODS: We retrospectively enrolled 103 pathologically confirmed SA. All lesions were evaluated with conventional US while 31 lesions with CEUS. Lesions were divided into SA with or without benign lesions (Group 1, n = 81) and SA with malignancy (Group 2, n = 22). Performance of two methods were analyzed. The ultrasonographic characteristics were compared between two groups with Student's t-test for measurement and chi-squared or Fisher's exact test for count data. RESULTS: There were 22 lesions complicated with malignancy, and the mean age of Group 2 was higher than Group 1 (55.27 vs. 41.57, p < 0.001). The sensitivity, specificity and accuracy of conventional US and CEUS were 95.45%, 46.91%, 57.28% and 100%, 62.5%, 70.97%. Angularity (p < 0.001), spicules (p = 0.023), calcification (p = 0.026) and enlarged scope (p = 0.012) or crab claw-like enhancement (p = 0.008) in CEUS were more frequent detected in SA with malignancy. CONCLUSIONS: Though CEUS showed an improved accuracy, the performance of ultrasound in the diagnosis of SA was limited. Awareness and careful review of the histopathologically related imaging features can be helpful in the diagnosis of SA.

Qualitative analysis of contrast-enhanced ultrasound in the diagnosis of small, TR3-5 benign and malignant thyroid nodules measuring ≤1 cm.

OBJECTIVE: To evaluate the performance of contrast-enhanced ultrasound in the diagnosis of small, solid, TR3-5 benign and malignant thyroid nodules (≤1 cm). METHODS: From January 2016 to March 2018, 185 thyroid nodules from 154 patients who underwent contrast enhanced ultrasound (CEUS) and fine-needle aspiration or thyroidectomy in Shanghai General Hospital were included. The χ2 test was used to compare the CEUS characteristics of benign and malignant thyroid nodules, and the CEUS features of malignant nodules assigned scores. The total score of the CEUS features and the scores of the above nodules were evaluated according to the latest 2017 version of the Thyroid Imaging Reporting and Data System (TI-RADS). The diagnostic performance of the two were compared based on the receiver operating characteristic curves generated for benign and malignant thyroid nodules. RESULTS: The degree, enhancement patterns, boundary, shape, and homogeneity of enhancement in thyroid small solid nodules were significantly different (p＜0.05). No significant differences were seen between benign and malignant thyroid nodules regarding completeness of enhancement and size of enhanced lesions (p＞0.05). The sensitivity, specificity, accuracy, positive predictive value, and negative predictive value of the TI-RADS classification TR5 in diagnosis of malignant nodules were 90.10%, 55.95%, 74.59%, 72.22%, and 82.46%, respectively (area under the curve [AUC]=0.738; 95% confidence interval[CI], 0.663-0.813). The sensitivity, specificity, accuracy, positive predictive value, and negative predictive value of the total score of CEUS qualitative analysis indicators were 86.13%, 89.29%, 87.57%, 90.63%, and 84.27% respectively (AUC = 0.916; 95% CI, 0.871-0.961). CONCLUSION: CEUS qualitative analysis is superior to TI-RADS in evaluating the diagnostic performance of small, solid thyroid nodules. Qualitative analysis of CEUS has a significantly higher specificity for diagnosis of malignant thyroid nodules than TI-RADS. ADVANCES IN KNOWLEDGE: The 2017 version of TI-RADS has recently suggested the malignant stratification of thyroid nodules by ultrasound. In this paper we applied this system and CEUS to evaluate 185 nodules and compare the results with pathological findings to access the diagnostic performance.

Diagnostic efficacy of contrast-enhanced ultrasound for breast lesions of different sizes: a comparative study with magnetic resonance imaging.

OBJECTIVE: This study aimed to compare the diagnostic performance of contrast-enhanced ultrasound (CEUS), MRI, and the combined use of the two modalities for differentiating breast lesions of different sizes. METHODS: A total of 406 patients with 406 solid breast masses detected by conventional ultrasound underwent both CEUS and MRI scans. Histological results were used as reference standards. The lesions were categorized into three groups according to size (Group 1, ≤ 20 mm; Group 2, > 20 mm, Group 3: total lesions). Sensitivity, specificity, accuracy, and receiver operating characteristic (ROC) curve analysis were used to assess the diagnostic performance of these imaging methods for breast lesions. RESULTS: There were 194 benign and 212 malignant breast lesions according to the histological diagnosis. Compared with MRI, CEUS demonstrated similar sensitivity in detecting breast cancer (p = 1.0000 for all) in all the three groups. With regard to specificity, accuracy, and the area under the ROC curve (Az) values, MRI showed a better performance than that shown by CEUS (p ＜0.05 for all), and the combination of the two modalities improved the diagnostic performance of CEUS alone significantly (p ＜0.05 for all) in all the three groups. However, the diagnostic specificity and accuracy of the combined method was not superior to that of MRI alone except for Group 2. CONCLUSION: CEUS demonstrated good sensitivity in detecting breast cancer, and the combined use with MRI can optimize the diagnostic specificity and accuracy in breast cancer prediction. ADVANCES IN KNOWLEDGE: Few studies have compared the diagnostic efficacy of CEUS and MRI, and this study is the first attempt to seek out the diagnostic values for breast lesions of variable sizes (lesions with ≤20 mm and >20 mm).

Incorporation of contrast-enhanced ultrasound in the differential diagnosis for breast lesions with inconsistent results on mammography and conventional ultrasound.

OBJECTIVE: To identify the efficacy of contrast-enhanced ultrasound (CEUS) in re-evaluating masses with inconsistent Breast Imaging Reporting and Data System (BI-RADS) on mammography (MG) and conventional ultrasound (US). MATERIALS AND METHODS: A total of 637 breast lesions were evaluated with MG, US, and CEUS within 6 months and assessed as BI-RADS MG and US. CEUS was used as an additional screening to rerate BI-RADS US according to a five-point system. Lesions were divided into consistent or inconsistent group on the basis of BI-RADS MG and US assessment. The performance of MG, US, and CEUS in the overall and inconsistent group as well as the clinicopathological differences between consistent and inconsistent group were compared using Z test, Mann-Whitney U test, and t-test. RESULTS: The respective AUCs of MG and US were 0.742, 0.843 for overall group and 0.412, 0.789 for inconsistent group. The corresponding values of rerated CEUS BI-RADS were 0.958 and 0.950, which were significantly prior to those of MG and US (p < 0.001). Younger age, negative lymph node status, and dense breast were significantly associated with inconsistent group. CONCLUSION: Incorporation of CEUS to re-evaluate lesions can improve the diagnostic efficacy comparing to MG or US alone especially when disagreement occurred.

Deregulation of ATG9A by impaired AR signaling induces autophagy in prostate stromal fibroblasts and promotes BPH progression.

The activation of androgen receptor (AR) signaling plays an essential role in both prostate stromal cells and epithelial cells during the development of benign prostatic hyperplasia (BPH). Here we demonstrated that androgen ablation after 5α-reductase inhibitor (5-ARI) treatment induced autophagy in prostate stromal fibroblasts inhibiting cell apoptosis. In addition, we found that ATG9A expression was increased after androgen ablation, which facilitated autophagic flux development. Knockdown of ATG9A not only inhibited autophagy notably in prostate stromal fibroblasts, but also reduced the volumes of prostate stromal fibroblast and epithelial cell recombinant grafts in nude mice. In conclusion, our findings suggested that ATG9A upregulation after long-term 5-ARI treatment constitutes a possible mechanism of BPH progression. Thus, combined treatment with 5-ARI and autophagy inhibitory agents would reduce the risk of BPH progression.