Prospective study of AI-assisted prediction of breast malignancies in physical health examinations: role of off-the-shelf AI software and comparison to radiologist performance.

Objective: In physical health examinations, breast sonography is a commonly used imaging method, but it can lead to repeated exams and unnecessary biopsy due to discrepancies among radiologists and health centers. This study explores the role of off-the-shelf artificial intelligence (AI) software in assisting radiologists to classify incidentally found breast masses in two health centers. Methods: Female patients undergoing breast ultrasound examinations with incidentally discovered breast masses were categorized according to the 5th edition of the Breast Imaging Reporting and Data System (BI-RADS), with categories 3 to 5 included in this study. The examinations were conducted at two municipal health centers from May 2021 to May 2023.The final pathological results from surgical resection or biopsy served as the gold standard for comparison. Ultrasonographic images were obtained in longitudinal and transverse sections, and two junior radiologists and one senior radiologist independently assessed the images without knowing the pathological findings. The BI-RADS classification was adjusted following AI assistance, and diagnostic performance was compared using receiver operating characteristic curves. Results: A total of 196 patients with 202 breast masses were included in the study, with pathological results confirming 107 benign and 95 malignant masses. The receiver operating characteristic curve showed that experienced breast radiologists had higher diagnostic performance in BI-RADS classification than junior radiologists, similar to AI classification (AUC = 0.936, 0.806, 0.896, and 0.950, p < 0.05). The AI software improved the accuracy, sensitivity, and negative predictive value of the adjusted BI-RADS classification for the junior radiologists' group (p< 0.05), while no difference was observed in the senior radiologist group. Furthermore, AI increased the negative predictive value for BI-RADS 4a masses and the positive predictive value for 4b masses among radiologists (p < 0.05). AI enhances the sensitivity of invasive breast cancer detection more effectively than ductal carcinoma in situ and rare subtypes of breast cancer. Conclusions: The AI software enhances diagnostic efficiency for breast masses, reducing the performance gap between junior and senior radiologists, particularly for BI-RADS 4a and 4b masses. This improvement reduces unnecessary repeat examinations and biopsies, optimizing medical resource utilization and enhancing overall diagnostic effectiveness.

Development of a Multi-Parametric ultrasonography nomogram for prediction of invasiveness in ductal carcinoma in situ.

OBJECTIVE: To investigate the independent risk variables associated with the potential invasiveness of ductal carcinoma in situ (DCIS) on multi-parametric ultrasonography, and further construct a nomogram for risk assessment. METHODS: Consecutive patients from January 2017 to December 2022 who were suspected of having ductal carcinoma in situ (DCIS) based on magnetic resonance imaging or mammography were prospectively enrolled. Histopathological findings after surgical resection served as the gold standard. Grayscale ultrasound, Doppler ultrasound, shear wave elastography (SWE), and contrast-enhanced ultrasound (CEUS) examinations were preoperative performed. Binary logistic regression was used for multifactorial analysis to identify independent risk factors from multi-parametric ultrasonography. The correlation between independent risk factors and pathological prognostic markers was analyzed. The predictive efficacy of DCIS associated with invasiveness was assessed by logistic analysis, and a nomogram was established. RESULTS: A total of 250 DCIS lesions were enrolled from 249 patients, comprising 85 pure DCIS and 165 DCIS with invasion (DCIS-IDC), of which 41 exhibited micro-invasion. The multivariate analysis identified independent risk factors for DCIS with invasion on multi-parametric ultrasonography, including image size (>2cm), Doppler ultrasound RI (≥0.72), SWE's Emax (≥66.4 kPa), hyper-enhancement, centripetal enhancement, increased surrounding vessel, and no contrast agent retention on CEUS. These factors correlated with histological grade, Ki-67, and human epidermal growth factor receptor 2 (HER2) (P < 0.1). The multi-parametric ultrasound approach demonstrated good predictive performance (sensitivity 89.7 %, specificity 73.8 %, AUC 0.903), surpassing single US modality or combinations with SWE or CEUS modalities. Utilizing these factors, a predictive nomogram achieved a respectable performance (AUC of 0.889) for predicting DCIS with invasion. Additionally, a separate nomogram for predicting DCIS with micro-invasion, incorporating independent risk factors such as RI (≥0.72), SWE's Emax (≥65.2 kPa), and centripetal enhancement, demonstrated an AUC of 0.867. CONCLUSION: Multi-parametric ultrasonography demonstrates good discriminatory ability in predicting both DCIS with invasion and micro-invasion through the analysis of lesion morphology, stiffness, neovascular architecture, and perfusion. The use of a nomogram based on ultrasonographic images offers an intuitive and effective method for assessing the risk of invasion in DCIS. Although the nomogram is not currently considered a clinically applicable diagnostic tool due to its AUC being below the threshold of 0.9, further research and development are anticipated to yield positive outcomes and enhance its viability for clinical utilization.

Predictive value of contrast-enhanced ultrasonography and ultrasound elastography for management of BI-RADS category 4 nonpalpable breast masses.

PURPOSE: The objective of this study was to investigate the independent risk factors and associated predictive values of contrast-enhanced ultrasound (CEUS), shear wave elastography (SWE), and strain elastography (SE) for high-risk lesions (HRL) and malignant tumors (MT) among nonpalpable breast masses classified as BI-RADS category 4 on conventional ultrasound. METHODS: This prospective study involved consecutively admitted patients with breast tumors from January 2018, aiming to explore the management of BI-RADS category 4 breast tumors using CEUS and elastography. We conducted a retrospective review of patient data, focusing on those with a history of a nonpalpable mass as the primary complaint. Pathologic findings after surgical resection served as the gold standard. The CEUS arterial-phase indices were analyzed using contrast agent arrival-time parametric imaging processing mode, while quantitative and qualitative indices were examined on ES images. Independent risk factors were identified through binary logistic regression multifactorial analysis. The predictive efficacy of different modalities was compared using a receiver operating characteristics curve. Subsequently, a nomogram for predicting the risk of HRL/MT was established based on a multifactorial logistic regression model. RESULTS: A total of 146 breast masses from 146 patients were included, comprising 80 benign tumors, 12 HRLs, and 54 MTs based on the final pathology. There was no significant difference in pathologic size between the benign and HRL/MT groups [8.00(6.25,10.00) vs. 9.00(6.00,10.00), P = 0.506]. The diagnostic efficacy of US plus CEUS exceeded that of US plus SWE/SE for BI-RADS 4 nonpalpable masses, with an AUC of 0.954 compared to 0.798/0.741 (P ï¼œ 0.001). Further stratified analysis revealed a more pronounced improvement for reclassification of BI-RADS 4a masses (AUC: 0.943 vs. 0.762/0.675, P ï¼œ 0.001) than BI-RADS 4b (AUC:0.950 vs. 0.885/0.796, P＞0.05) with the assistance of CEUS than SWE/SE. Employing downgrade CEUS strategies resulted in negative predictive values ranging from 95.2 % to 100.0 % for BI-RADS 4a and 4b masses. Conversely, using upgrade nomogram strategies, which included the independent predictive risk factors of irregular enhanced shape, poor defined enhanced margin, earlier enhanced time, increased surrounding vessels, and presence of contrast agent retention, the diagnostic performance achieved an AUC of 0.947 with good calibration. CONCLUSION: After investigating the potential of CEUS and ES in improving risk assessment and diagnostic accuracy for nonpalpable BI-RADS category 4 breast masses, it is evident that CEUS has a more significant impact on enhancing classification compared to ES, particularly for BI-RADS 4a subgroup masses. This finding suggests that CEUS may offer greater benefits in improving risk assessment and diagnostic accuracy for this specific subgroup of breast masses.

Value of an expanded range of lesions on contrast-enhanced ultrasound for the diagnosis of hypervascular breast masses.

Background: Breast cancer lesions show an expanded range on contrast-enhanced ultrasound (CEUS). Here, we quantitatively analyze this index to explore its effective cutoff value for distinguishing benign and malignant lesions and the corresponding diagnostic performance and investigate its role in prognostic assessment of malignant lesions. Methods: Consecutive patients who underwent CEUS for breast lesions during the period from September 2017 to June 2019 were included. Original CEUS images were selected, displayed in dual-frame mode, and measured when enhancement of the lesion reached its peak. The longitudinal diameter, transverse diameter, and area of the lesion on the two-dimensional images and the corresponding postenhancement images were measured to calculate six indicators: longitudinal diameter increment, transverse diameter increment, area increment, percent increase in longitudinal diameter, percent increase in transverse diameter, and percent increase in area increment. With postoperative pathology as the gold standard, the cutoff values for distinguishing benign and malignant lesions and the correlations of these indicators with pathological subtypes and pathological grades were evaluated. Results: Malignant lesions showed a more significantly expanded range after enhancement compared to benign lesions, especially in terms of area increase. When the cutoff value of the area increment was set at 0.47 cm2 for distinguishing between benign and malignant lesions, the area under the curve (AUC) was 0.945, and the sensitivity, specificity, accuracy, positive predictive value, and negative predictive value were 90.1%, 91.5%, 90.9%, 87.2%, and 93.5%, respectively. The pathologically measured maximum diameter of malignant masses correlated with the percent increase in transverse diameter, area increment, and percent increase in area increment. The longitudinal diameter increment in the luminal A group was significantly smaller than that in the human epidermal growth factor receptor 2 (HER2)+ group. The percent increase in transverse diameter was helpful for predicting the pathological grade of malignant masses. When the cutoff value of the percent increase in transverse diameter was set at 10.84% for pathological grading, the AUC was 0.623, and the sensitivity was 90.8%. Conclusions: Indicators related to the expanded lesion range on CEUS are helpful in differential diagnosis of benign and malignant lesions and in prognostic assessment of pathological grades.

CNN-based automatic segmentations and radiomics feature reliability on contrast-enhanced ultrasound images for renal tumors.

Objective: To investigate the feasibility and efficiency of automatic segmentation of contrast-enhanced ultrasound (CEUS) images in renal tumors by convolutional neural network (CNN) based models and their further application in radiomic analysis. Materials and methods: From 94 pathologically confirmed renal tumor cases, 3355 CEUS images were extracted and randomly divided into training set (3020 images) and test set (335 images). According to the histological subtypes of renal cell carcinoma, the test set was further split into clear cell renal cell carcinoma (ccRCC) set (225 images), renal angiomyolipoma (AML) set (77 images) and set of other subtypes (33 images). Manual segmentation was the gold standard and serves as ground truth. Seven CNN-based models including DeepLabV3+, UNet, UNet++, UNet3+, SegNet, MultilResUNet and Attention UNet were used for automatic segmentation. Python 3.7.0 and Pyradiomics package 3.0.1 were used for radiomic feature extraction. Performance of all approaches was evaluated by the metrics of mean intersection over union (mIOU), dice similarity coefficient (DSC), precision, and recall. Reliability and reproducibility of radiomics features were evaluated by the Pearson coefficient and the intraclass correlation coefficient (ICC). Results: All seven CNN-based models achieved good performance with the mIOU, DSC, precision and recall ranging between 81.97%-93.04%, 78.67%-92.70%, 93.92%-97.56%, and 85.29%-95.17%, respectively. The average Pearson coefficients ranged from 0.81 to 0.95, and the average ICCs ranged from 0.77 to 0.92. The UNet++ model showed the best performance with the mIOU, DSC, precision and recall of 93.04%, 92.70%, 97.43% and 95.17%, respectively. For ccRCC, AML and other subtypes, the reliability and reproducibility of radiomic analysis derived from automatically segmented CEUS images were excellent, with the average Pearson coefficients of 0.95, 0.96 and 0.96, and the average ICCs for different subtypes were 0.91, 0.93 and 0.94, respectively. Conclusion: This retrospective single-center study showed that the CNN-based models had good performance on automatic segmentation of CEUS images for renal tumors, especially the UNet++ model. The radiomics features extracted from automatically segmented CEUS images were feasible and reliable, and further validation by multi-center research is necessary.