Deep learning based ultrasound analysis facilitates precise distinction between parotid pleomorphic adenoma and Warthin tumor.

Background: Pleomorphic adenoma (PA), often with the benign-like imaging appearances similar to Warthin tumor (WT), however, is a potentially malignant tumor with a high recurrence rate. It is worse that pathological fine-needle aspiration cytology (FNAC) is difficult to distinguish PA and WT for inexperienced pathologists. This study employed deep learning (DL) technology, which effectively utilized ultrasound images, to provide a reliable approach for discriminating PA from WT. Methods: 488 surgically confirmed patients, including 266 with PA and 222 with WT, were enrolled in this study. Two experienced ultrasound physicians independently evaluated all images to differentiate between PA and WT. The diagnostic performance of preoperative FNAC was also evaluated. During the DL study, all ultrasound images were randomly divided into training (70%), validation (20%), and test (10%) sets. Furthermore, ultrasound images that could not be diagnosed by FNAC were also randomly allocated to training (60%), validation (20%), and test (20%) sets. Five DL models were developed to classify ultrasound images as PA or WT. The robustness of these models was assessed using five-fold cross-validation. The Gradient-weighted Class Activation Mapping (Grad-CAM) technique was employed to visualize the region of interest in the DL models. Results: In Grad-CAM analysis, the DL models accurately identified the mass as the region of interest. The area under the receiver operating characteristic curve (AUROC) of the two ultrasound physicians were 0.351 and 0.598, and FNAC achieved an AUROC of only 0.721. Meanwhile, for DL models, the AUROC value for discriminating between PA and WT in the test set was from 0.828 to 0.908. ResNet50 demonstrated the optimal performance with an AUROC of 0.908, an accuracy of 0.833, a sensitivity of 0.736, and a specificity of 0.904. In the test set of cases that FNAC failed to provide a diagnosis, DenseNet121 demonstrated the optimal performance with an AUROC of 0.897, an accuracy of 0.806, a sensitivity of 0.789, and a specificity of 0.824. Conclusion: For the discrimination of PA and WT, DL models are superior to ultrasound and FNAC, thereby facilitating surgeons in making informed decisions regarding the most appropriate surgical approach.

Application of ultrasound-based radiomics models of breast masses to predict invasive components of encapsulated papillary carcinoma.

Background: Axillary lymph node (ALN) metastasis is seen in encapsulated papillary carcinoma (EPC), mostly with an invasive component (INV). Radiomics can offer more information beyond subjective grayscale and color Doppler ultrasound (US) image interpretation. This study aimed to develop radiomics models for predicting an INV of EPC in the breast based on US images. Methods: This study retrospectively enrolled 105 patients (107 masses) with a pathological diagnosis of EPC from January 2016 to April 2021, and all masses had preoperative US images. Of the 107 masses, 64 were randomized to a training set and 43 to a test set. US and clinical features were analyzed to identify features associated with INVs. Then, based on the manually segmented US images to obtain radiomics features, the models to predict INVs were built with 5 ensemble machine learning classifiers. We estimated the performance of the predictive models using accuracy, the area under the receiver operating characteristic (ROC) curve (AUC), sensitivity, and specificity. Results: The mean age was 63.71 years (range, 31 to 85 years); the mean size of tumors was 23.40 mm (range, 9 to 120 mm). Among all clinical and US features, only shape was statistically different between EPC with INVs and those without (P<0.05). In this study, the models based on Random Under Sampling (RUS) Boost, Random Forest, XGBoost, AdaBoost, and Easy Ensemble methods had good performance, among which RUS Boost had the best performance with an AUC of 0.875 [95% confidence interval (CI): 0.750-0.974] in the test set. Conclusions: Radiomics prediction models were effective in predicting the INV of EPC, whereas clinical and US features demonstrated relatively decreased predictive utility.

Differentiation between Phyllodes Tumors and Fibroadenomas through Breast Ultrasound: Deep-Learning Model Outperforms Ultrasound Physicians.

The preoperative differentiation of breast phyllodes tumors (PTs) from fibroadenomas (FAs) plays a critical role in identifying an appropriate surgical treatment. Although several imaging modalities are available, reliable differentiation between PT and FA remains a great challenge for radiologists in clinical work. Artificial intelligence (AI)-assisted diagnosis has shown promise in distinguishing PT from FA. However, a very small sample size was adopted in previous studies. In this work, we retrospectively enrolled 656 breast tumors (372 FAs and 284 PTs) with 1945 ultrasound images in total. Two experienced ultrasound physicians independently evaluated the ultrasound images. Meanwhile, three deep-learning models (i.e., ResNet, VGG, and GoogLeNet) were applied to classify FAs and PTs. The robustness of the models was evaluated by fivefold cross validation. The performance of each model was assessed by using the receiver operating characteristic (ROC) curve. The area under the curve (AUC), accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were also calculated. Among the three models, the ResNet model yielded the highest AUC value, of 0.91, with an accuracy value of 95.3%, a sensitivity value of 96.2%, and a specificity value of 94.7% in the testing data set. In contrast, the two physicians yielded an average AUC value of 0.69, an accuracy value of 70.7%, a sensitivity value of 54.4%, and a specificity value of 53.2%. Our findings indicate that the diagnostic performance of deep learning is better than that of physicians in the distinction of PTs from FAs. This further suggests that AI is a valuable tool for aiding clinical diagnosis, thereby advancing precision therapy.

Survival outcome assessment for triple-negative breast cancer: a nomogram analysis based on integrated clinicopathological, sonographic, and mammographic characteristics.

OBJECTIVE: This study aimed to incorporate clinicopathological, sonographic, and mammographic characteristics to construct and validate a nomogram model for predicting disease-free survival (DFS) in patients with triple-negative breast cancer (TNBC). METHODS: Patients diagnosed with TNBC at our institution between 2011 and 2015 were retrospectively evaluated. A nomogram model was generated based on clinicopathological, sonographic, and mammographic variables that were associated with 1-, 3-, and 5-year DFS determined by multivariate logistic regression analysis in the training set. The nomogram model was validated according to the concordance index (C-index) and calibration curves in the validation set. RESULTS: A total of 636 TNBC patients were enrolled and divided into training cohort (n = 446) and validation cohort (n = 190). Clinical factors including tumor size > 2 cm, axillary dissection, presence of LVI, and sonographic features such as angular/spiculated margins, posterior acoustic shadows, and presence of suspicious lymph nodes on preoperative US showed a tendency towards worse DFS. The multivariate analysis showed that no adjuvant chemotherapy (HR = 6.7, 95% CI: 2.6, 17.5, p < 0.0005), higher axillary tumor burden (HR = 2.7, 95% CI: 1.0, 7.1, p = 0.045), and ≥ 3 malignant features on ultrasound (HR = 2.4, CI: 1.1, 5.0, p = 0.021) were identified as independent prognostic factors associated with poorer DFS outcomes. In the nomogram, the C-index was 0.693 for the training cohort and 0.694 for the validation cohort. The calibration plots also exhibited excellent consistency between the nomogram-predicted and actual survival probabilities in both the training and validation cohorts. CONCLUSIONS: Clinical variables and sonographic features were correlated with the prognosis of TNBCs. The nomogram model based on three variables including no adjuvant chemotherapy, higher axillary tumor load, and more malignant sonographic features showed good predictive performance for poor survival outcomes of TNBC. KEY POINTS: • The absence of adjuvant chemotherapy, heavy axillary tumor load, and malignant-like sonographic features can predict DFS in patients with TNBC. • Mammographic features of TNBC could not predict the survival outcomes of patients with TNBC. • The nomogram integrating clinicopathological and sonographic characteristics is a reliable predictive model for the prognostic outcome of TNBC.

Prediction for pathological and immunohistochemical characteristics of triple-negative invasive breast carcinomas: the performance comparison between quantitative and qualitative sonographic feature analysis.

OBJECTIVE: Sonographic features are associated with pathological and immunohistochemical characteristics of triple-negative breast cancer (TNBC). To predict the biological property of TNBC, the performance using quantitative high-throughput sonographic feature analysis was compared with that using qualitative feature assessment. METHODS: We retrospectively reviewed ultrasound images, clinical, pathological, and immunohistochemical (IHC) data of 252 female TNBC patients. All patients were subgrouped according to the histological grade, Ki67 expression level, and human epidermal growth factor receptor 2 (HER2) score. Qualitative sonographic feature assessment included shape, margin, posterior acoustic pattern, and calcification referring to the Breast Imaging Reporting and Data System (BI-RADS). Quantitative sonographic features were acquired based on the computer-aided radiomics analysis. Breast cancer masses were manually segmented from the surrounding breast tissues. For each ultrasound image, 1688 radiomics features of 7 feature classes were extracted. The principal component analysis (PCA), least absolute shrinkage and selection operator (LASSO), and support vector machine (SVM) were used to determine the high-throughput radiomics features that were highly correlated to biological properties. The performance using both quantitative and qualitative sonographic features to predict biological properties of TNBC was represented by the area under the receiver operating characteristic curve (AUC). RESULTS: In the qualitative assessment, regular tumor shape, no angular or spiculated margin, posterior acoustic enhancement, and no calcification were used as the independent sonographic features for TNBC. Using the combination of these four features to predict the histological grade, Ki67, HER2, axillary lymph node metastasis (ALNM), and lymphovascular invasion (LVI), the AUC was 0.673, 0.680, 0.651, 0.587, and 0.566, respectively. The number of high-throughput features that closely correlated with biological properties was 34 for histological grade (AUC 0.942), 27 for Ki67 (AUC 0.732), 25 for HER2 (AUC 0.730), 34 for ALNM (AUC 0.804), and 34 for LVI (AUC 0.795). CONCLUSION: High-throughput quantitative sonographic features are superior to traditional qualitative ultrasound features in predicting the biological behavior of TNBC. KEY POINTS: • Sonographic appearances of TNBCs showed a great variety in accordance with its biological and clinical characteristics. • Both qualitative and quantitative sonographic features of TNBCs are associated with tumor biological characteristics. • The quantitative high-throughput feature analysis is superior to two-dimensional sonographic feature assessment in predicting tumor biological property.

A Non-Contact Fault Diagnosis Method for Bearings and Gears Based on Generalized Matrix Norm Sparse Filtering.

Fault diagnosis of mechanical equipment is mainly based on the contact measurement and analysis of vibration signals. In some special working conditions, the non-contact fault diagnosis method represented by the measurement of acoustic signals can make up for the lack of contact testing. However, its engineering application value is greatly restricted due to the low signal-to-noise ratio (SNR) of the acoustic signal. To solve this deficiency, a novel fault diagnosis method based on the generalized matrix norm sparse filtering (GMNSF) is proposed in this paper. Specially, the generalized matrix norm is introduced into the sparse filtering to seek the optimal sparse feature distribution to overcome the defect of low SNR of acoustic signals. Firstly, the collected acoustic signals are randomly overlapped to form the sample fragment data set. Then, three constraints are imposed on the multi-period data set by the GMNSF model to extract the sparse features in the sample. Finally, softmax is used to as a classifier to categorize different fault types. The diagnostic performance of the proposed method is verified by the bearing and planetary gear datasets. Results show that the GMNSF model has good feature extraction ability performance and anti-noise ability than other traditional methods.

Ultrasonographic appearance of triple-negative invasive breast carcinoma is associated with novel molecular subtypes based on transcriptomic analysis.

BACKGROUND: Various sonographic features of triple-negative invasive breast carcinomas (TNBC) expected to be associated with the molecular subtypes based on transcriptomic analysis were examined. The effects of clinical, sonographic, pathological, and molecular features on survival outcome was also studied. METHODS: One hundred and fourteen patients with breast cancer with negative expression of estrogen receptor (ER), progesterone receptor (PR), and human epidermal receptor 2 (HER2) were included in our retrospective study. Based on the transcriptomic profiles, four stable clusters named immunomodulatory (IM), luminal androgen receptor (LAR), mesenchymal-like (MES), and basal-like and immune-suppressed (BLIS) were identified. Ultrasound (US) images were reviewed by two US physicians according to Breast Imaging Reporting and Data System (BI-RADS). Multivariate Cox regression was used to determine the variables associated with recurrence-free survival (RFS) and overall survival (OS). RESULTS: There were 21 IM, 18 LAR, 36 MES, and 39 BLIS cases. The four molecular subtypes showed significant differences in terms of tumor shape (P=0.008) and posterior acoustic pattern (P=0.028). Compared with the subtypes LAR and MES, the IM and BLIS subtypes had higher probability of presenting benign-like sonographic features, such as regular shape, no angular/spiculated margin, and posterior acoustic enhancement (P<0.05). The independent risk factors for RFS events and death were axillary lymph node metastasis (P<0.05) and BLIS subtype (P<0.05). BLIS subtype showed worse OS than other subtypes (log rank P=0.05). TNBCs with benign sonographic features tended to have less death events (3.3% vs. 15.2%, P=0.088). CONCLUSIONS: Sonographic appearance of TNBCs is associated with transcriptome-based molecular subtypes, and tends to correlate with the survival outcome.

Tumor Proliferation and Invasiveness Derived From Ultrasound Appearances of Invasive Breast Cancers: Moving Beyond the Routine Differential Diagnosis.

OBJECTIVES: To investigate the correlation between ultrasound (US) appearances of invasive breast cancers and tumor proliferation and invasiveness measured according to the histologic grade, Ki-67 expression, axillary lymph node metastasis (ALNM), and lymphovascular invasion (LVI). METHODS: This study evaluated 676 patients who underwent primary surgical treatment of invasive breast cancers. The preoperative US reports and postoperative pathologic and immunohistochemical results of the patients were retrospectively reviewed. Ultrasound characteristics were evaluated according to the American College of Radiology Breast Imaging Reporting and Data System (BI-RADS) lexicon. Logistic regression analyses were used to identify independent predictive US features that were correlated with tumor proliferation and invasiveness of breast cancers. Odds ratios (ORs) were calculated. RESULTS: Posterior acoustic enhancement and calcifications on US images were independent predictive factors of a higher histologic grade and a higher Ki-67 level (OR, 1.69-6.54; P < .05). Meanwhile, a noncircumscribed margin (OR, 2.61; P < .05) and posterior acoustic shadow (OR, 1.62; P < .05) were independent predictors of ALNM. An irregular shape (OR, 2.13; P < .05) and calcifications (OR, 1.69; P < .05) were independent risk factors for LVI. Infiltrative breast cancers scored as BI-RADS category 5 had higher probability to be associated with ALNM (OR, 3.33; P < .0005) and LVI (OR, 2.87; P < .0005). CONCLUSIONS: Ultrasound features of invasieve breast cancers might have a predictive value for tumor proliferation and invasiveness. The US features correlated with a high cellular proliferation rate were different from those associated with ALNM. The tumor shape, margin, posterior acoustic pattern, and calcifications at US are suggested to be considered by clinicians when making clinical decisions.

Author Correction: Triple-negative invasive breast carcinoma: the association between the sonographic appearances with clinicopathological feature.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Sonographic Features of Triple-Negative Breast Carcinomas Are Correlated With mRNA-lncRNA Signatures and Risk of Tumor Recurrence.

BACKGROUND: To determine a correlation between mRNA and lncRNA signatures, sonographic features, and risk of recurrence in triple-negative breast cancers (TNBC). METHODS: We retrospectively reviewed the data from 114 TNBC patients having undergone transcriptome analysis. The risk of tumor recurrence was determined based on the correlation between transcriptome profiles and recurrence-free survival. Ultrasound (US) features were described according to the Breast Imaging Reporting and Data System. Multivariate logistic regression analysis determined the correlation between US features and risk of recurrence. The predictive value of sonographic features in determining tumor recurrence was analyzed using receiver operating characteristic curves. RESULTS: Three mRNAs (CHRDL1, FCGR1A, and RSAD2) and two lncRNAs (HIF1A-AS2 and AK124454) were correlated with recurrence-free survival in patients with TNBC. Among the three mRNAs, two were upregulated (FCGR1A and RSAD2) and one was downregulated (CHRDL1) in TNBCs. LncRNAs HIF1A-AS2 and AK124454 were upregulated in TNBCs. Based on these signatures, an integrated mRNA-lncRNA model was established using Cox regression analysis to determine the risk of tumor recurrence. Benign-like sonographic features, such as regular shape, circumscribed margin, posterior acoustic enhancement, and no calcifications, were associated with HIF1A-AS2 expression and high risk of tumor recurrence (P<0.05). Malignant-like features, such as irregular shape, uncircumscribed margin, no posterior acoustic enhancement, and calcifications, were correlated with CHRDL1 expression and low risk of tumor recurrence (P<0.05). CONCLUSIONS: Sonographic features and mRNA-lncRNA signatures in TNBCs represent the risk of tumor recurrence. Taken together, US may be a promising technique in determining the prognosis of patients with TNBC.

The Association Between Ultrasound Features and Biological Properties of Invasive Breast Carcinoma Is Modified by Age, Tumor Size, and the Preoperative Axilla Status.

OBJECTIVES: To investigate the value of ultrasound (US) feature-based models in predicting the proliferation and invasiveness of invasive breast cancer (IBC) and to compare the performance of models based solely on US features with models that combined US features, patient age, tumor size, and axilla status from US. METHODS: With ethical approval, 746 patients with a pathologic diagnosis of IBC were reviewed for preoperative clinical, US, and postoperative pathologic data. The proliferation and invasiveness properties of the IBC included the histologic grade and Ki-67 status and lymphovascular invasion (LVI) and axillary lymph node metastasis (ALNM), respectively. Logistic regression analyses were used to identify independent risk factors for tumor proliferation and invasiveness. RESULTS: Posterior echo enhancement, calcification, a tumor size larger than 2 cm, and suspicion of ALNM from axillary US were independent risk factors for a high histologic grade and high Ki-67 expression of IBC (P < .05). A posterior echo shadow, patient age younger than 45 years, and suspicious findings on axillary US imaging were independent variables for predicting the presence of LVI and ALNM in IBC (P < .05). Calcification was the independent factor for predicting LVI (P = .013). The predictive performance of the combined models was improved compared with the US feature-based models, with a higher accuracy rate and negative predictive value. The area under curve of the combined models was also significantly higher than that of the single models (P < .05). CONCLUSIONS: Compared with the US feature-based models, the combined models yielded better predictive performance. This may provide a more robust model to predict the tumor biological properties of IBC before surgery.

Clinicopathologic and Ultrasound Variables Associated With a Heavy Axillary Nodal Tumor Burden in Invasive Breast Carcinoma.

OBJECTIVES: To identify clinicopathologic and ultrasound (US) variables that were associated with a heavy nodal tumor burden, which was defined as 3 or more lymph nodes involved with metastasis to the axilla after invasive breast carcinoma. METHODS: With ethical approval, 621 patients with a pathologic diagnosis of invasive breast carcinoma were retrospectively analyzed for clinical, pathologic, and US data. Pathologic findings were ascertained by the final paraffin pathologic analysis. Ultrasound characteristics were evaluated on the basis of the American College of Radiology's Breast Imaging Reporting and Data System (BI-RADS). Univariate and multivariate logistic regression analyses were used to assess the clinicopathologic and US variables that were associated with a heavy nodal tumor burden at the axilla. RESULTS: There were 107 cases (17.2%) of invasive breast carcinoma with a heavy tumor burden at the axilla. The independent clinicopathologic variables for a heavy tumor burden at the axilla included a tumor size of 2 to 5 cm (odds ratio [OR], 1.86; P = .036), the presence of lymphovascular invasion (OR, 23.52; P < .001), the presence of papillary invasion (OR, 2.93; P = .043), and a non-triple-negative subtype (OR, 2.34; P = .04). The independent US features of breast tumors that were associated with a heavy tumor burden at the axilla included BI-RADS category 5 (OR, 5.50; P = .024) and a posterior acoustic shadow (OR, 1.94; P = .024). CONCLUSIONS: A large tumor size, lymphovascular invasion, papillary invasion, and a non-triple-negative subtype on the pathologic analysis as well as BI-RADS category 5 and a posterior acoustic shadow on a US assessment were associated with a heavy nodal tumor burden at the axilla. These US characteristics of the primary breast carcinoma might provide additional information to axillary US for the prediction of axillary nodal tumor loads.

Nodule Size Effect on Diagnostic Performance of Ultrasonography and Computed Tomography for Papillary Thyroid Carcinoma.

BACKGROUND: To compare the abilities of ultrasonography (US) and Computed Tomography (CT) to identify calcifications and to predict probability of malignancy for Papillary Thyroid Carcinoma (PTC) and Papillary Thyroid Microcarcinoma (PTMC). METHODS: We reviewed 1008 cases of PTC/PTMC with calcifications reported by pre-operative US, CT, or post-operative pathology. The size of the thyroid nodule was obtained from the US report and the maximum diameter (d) was documented. According to the nodule size (d), the PTC and PTMC groups were each divided into two subgroups, as follows: large PTC group (d ≥ 2 cm), small PTC group (1 cm < d < 2 cm), large PTMC group (0.6 cm ≤ d ≤ 1 cm), and small PTMC group (d < 0.6 cm). RESULTS: In the 1008 patients, the ratio of females to males was 2.29 and the mean age was 40.9 years (standard deviation: 11.7 years). Of the 1008 records, 92.8% were found to have calcifications according to the US report, while 50.4% showed calcifications according to the CT report. This difference between US and CT reports was statistically significant (p < 0.0005). The percentages of US reports showing calcifications were similar for all four PTC and PTMC subgroups (93.7%, 94.3%, 92.1%, and 85.1%, respectively; p = 0.052), while the percentages of CT reports showing calcifications were significantly different among the PTC and PTMC subgroups (62.3%, 52.2%, 45.4%, and 31.3%, respectively; p < 0.0005). As for the prediction of malignancy, US was superior to CT in all four subgroups (large PTC group: 97.1% vs. 54.1%, small PTC group: 94.8% vs. 42.9%, large PTMC group: 97.2% vs. 32.0%, small PTMC group: 95.5% vs. 14.9%; p < 0.0005 for all pairwise comparisons). No significant difference was observed in terms of the ability of US to predict the malignancy of PTC versus PTMC (p = 0.31), while CT showed significant superiority in diagnosing PTC versus PTMC (p < 0.0005). The predictive value of CT for PTC declined as the nodule size decreased (p < 0.05 for all pairwise comparisons). CONCLUSION: Our results showed that US detected calcifications and predicted the malignancy of all nodule sizes of thyroid papillary carcinoma equally well, while the performance of CT declined with the reduction of nodule size.

Triple-negative invasive breast carcinoma: the association between the sonographic appearances with clinicopathological feature.

In this study, we aimed to evaluate the clinical and pathological factors that associated with sonographic appearances of triple-negative (TN) invasive breast carcinoma. With the ethical approval, 560 patients who were pathologically confirmed as invasive breast carcinoma were reviewed for ultrasound, clinical, and pathological data. Logistic regression analysis was used to identify the typical sonographic features for TN invasive breast carcinomas. The effect of clinical and pathological factors on the sonographic features of TN invasive breast carcinoma was studied. There were 104 cases of TN invasive breast carcinoma. The independent sonographic features for the TN subgroup included regular shape (odds ratio, OR = 2.14, p = 0.007), no spiculated/angular margin (OR = 1.93, p = 0.035), posterior acoustic enhancement (OR = 2.14, p = 0.004), and no calcifications (OR = 2.10, p = 0.008). Higher pathological grade was significantly associated with regular tumor shape of TN breast cancer (p = 0.012). Higher Ki67 level was significantly associated with regular tumor shape (p = 0.023) and absence of angular/spiculated margin (p = 0.005). Higher human epidermal growth factor receptor 2 (HER2) score was significantly associated with the presence of calcifications (p = 0.033). We conclude that four sonographic features are associated with TN invasive breast carcinoma. Heterogeneity of sonographic features was associated with the pathological grade, Ki67 proliferation level and HER2 score of TN breast cancers.