Breast Tumour Classification Using Ultrasound Elastography with Machine Learning: A Systematic Scoping Review

Ultrasound elastography can quantify stiffness distribution of tissue lesions and complements conventional B-mode ultrasound for breast cancer screening. Recently, the development of computer-aided diagnosis has improved the reliability of the system, whilst the inception of machine learning, such as deep learning, has further extended its power by facilitating automated segmentation and tumour classification. The objective of this review was to summarize application of the machine learning model to ultrasound elastography systems for breast tumour classification. Review databases included PubMed, Web of Science, CINAHL, and EMBASE. Thirteen (n = 13) articles were eligible for review. Shear-wave elastography was investigated in six articles, whereas seven studies focused on strain elastography (5 freehand and 2 Acoustic Radiation Force). Traditional computer vision workflow was common in strain elastography with separated image segmentation, feature extraction, and classifier functions using different algorithm-based methods, neural networks or support vector machines (SVM). Shear-wave elastography often adopts the deep learning model, convolutional neural network (CNN), that integrates functional tasks. All of the reviewed articles achieved sensitivity ³ 80%, while only half of them attained acceptable specificity ³ 95%. Deep learning models did not necessarily perform better than traditional computer vision workflow. Nevertheless, there were inconsistencies and insufficiencies in reporting and calculation, such as the testing dataset, cross-validation, and methods to avoid overfitting. Most of the studies did not report loss or hyperparameters. Future studies may consider using the deep network with an attention layer to locate the targeted object automatically and online training to facilitate efficient re-training for sequential data.

Comparison between Histological Features and Strain Elastographic Characteristics in Canine Mammary Carcinomas

Elastography is a sonographic technique that provides a noninvasive evaluation of the stiffness of a lesion. The objective of this work was to evaluate the accuracy of strain elastography, the most accessible modality in clinical practice, to discriminate between different histological types of malignant mammary neoplasms in the canine species, which can provide complementary information in real time to the diagnosis and thus help in the choice of surgical technique. A total of 34 females with 56 mammary carcinomas were selected and classified into three histological groups according to their aggressiveness. The histological and elastographic characteristics of these malignant tumors were analyzed and compared to evaluate the diagnostic accuracy of strain elastography. Visual score presented a sensitivity of 88.0%, specificity of 58.1%, and accuracy of 71.43% in distinguishing the most aggressive group of carcinomas. The strain ratio had a sensitivity of 84.0%, specificity of 61.1%, and accuracy of 69.64%. On the other hand, intratumoral strain ratio obtained a sensitivity of 71.40% and specificity of 61.90% when intratumoral fibrosis was taken as reference, with an accuracy of 66.07%. Similarly, peritumoral strain ratio was also positively related to fibrosis in the periphery of lesions (p ≤ 0.001), with a sensitivity of 93.80%, specificity of 77.50% and an accuracy of 92.87%. In conclusion, accuracy of this elastographic modality can be a useful method to differentiate more aggressive histological types. Therefore, it represents an additional diagnostic technique useful in the daily clinic thanks to the short time required for the examination, which allows real-time visualization and immediate interpretation of the results.

Repeatability and reproducibility of quantitative cervical strain elastography (E-Cervix) in pregnancy

Strain elastography of the uterine cervix may be useful in the diagnosis and prediction of obstetric complications. The inability to obtain quantitative results, with only the possibility of visual semiquantitative evaluation of the obtained elastograms, has been the limitation of the method thus far. E-Cervix is a software program that uses intrinsic compression to excite tissue and allows the evaluation of quantitative parameters on the basis of pixel distribution in an elastogram. The aim of this study was to assess the repeatability and reproducibility of quantitative cervical strain elastography (E-Cervix) of the uterine cervix and to assess the correlation of the obtained parameters with selected clinical features of patients in the third trimester of pregnancy. In total, 222 patients participated in the study. We assessed 5 ultrasound parameters: elasticity index (ECI), hardness ratio (HR), internal os strain (IOS), external os strain (EOS) and IOS/EOS ratio. Each study was performed according to a predetermined standardized protocol. For all assessed elastographic parameters, we obtained good intra- and interobserver reproducibility. The interclass correlation coefficient (ICC) ranged from 0.77 to 0.838 for intraobserver variability and from 0.771 to 0.826 for interobserver variability. We demonstrated a significant correlation of some obtained elastographic parameters with the basic clinical features of patients, such as age, the number of previous caesarean sections, pregnancy weight and BMI. In each case, the correlation was very low. Quantitative elastographic assessment with the use of E-Cervix is characterized by good repeatability. Some clinical features may affect the value of the parameters obtained. The clinical relevance of this interference requires further investigation.

Added Value of a New Strain Elastography Technique in Conventional Ultrasound for the Diagnosis of Breast Masses: A Prospective Multicenter Study

ObjectiveThis study aimed to explore the value of elasticity score (ES) and strain ratio (SR) combined with conventional ultrasound in distinguishing benign and malignant breast masses and reducing biopsy of BI-RADS (Breast Imaging Reporting and Data System) 4a lesions.MethodsThis prospective, multicenter study included 910 patients from nine different hospitals. The acquisition and analysis of conventional ultrasound and strain elastography (SE) were obtained by radiologists with more than 5 years of experience in breast ultrasound imaging. The diagnostic sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under curve (AUC) of conventional ultrasound alone and combined tests with ES and/or SR were calculated and compared.ResultsThe optimal cutoff value of SR for differentiating benign from malignant masses was 2.27, with a sensitivity of 60.2% and a specificity of 84.8%. When combined with ES and SR, the AUC of the new BI-RADS classification increased from 0.733 to 0.824 (p < 0.001); the specificity increased from 48.1% to 68.5% (p < 0.001) without a decrease in the sensitivity (98.5% vs. 96.4%, p = 0.065); and the PPV increased from 52.2% to 63.7% (p < 0.001) without a loss in the NPV (98.2% vs. 97.1%, p = 0.327). All three combinations of conventional ultrasound, ES, and SR could reduce the biopsy rate of category 4a lesions without reducing the malignant rate of biopsy (from 100% to 68.3%, 34.9%, and 50.4%, respectively, all p < 0.001).ConclusionsSE can be used as a useful and non-invasive additional method to improve the diagnostic performance of conventional ultrasound by increasing AUC and specificity and reducing the unnecessary biopsy of BI-RADS 4a lesions.

ITVT-06. Application of artificial intelligence and radiomics for the analysis of intraoperative ultrasound images of brain tumors

BACKGROUND Intraoperative ultrasound (ioUS) images of brain tumors contain information that has not yet been exploited. The present work aims to analyze images in both B-mode and strain-elastography using techniques based on artificial intelligence and radiomics. We pretend to assess the capacity for differentiating glioblastomas (GBM) from solitary brain metastases (SBM) and also to assess the ability to predict the overall survival (OS) in GBM. METHODS We performed a retrospective analysis of patients who underwent craniotomy between March 2018 to June 2020 with GBM and SBM diagnoses. Cases with an ioUS study were included. In the first group of patients, an analysis based on deep learning was performed. An existing neural network (Inception V3) was used to classify tumors into GBM and SBM. The models were evaluated using the area under the curve (AUC), classification accuracy, and precision. In the second group, radiomic features from the tumor region were extracted. Radiomic features associated with OS were selected employing univariate correlations. Then, a survival analysis was conducted using Cox regression. RESULTS For the classification task, a total of 36 patients were included. 26 GBM and 10 SBM. Models were built using a total of 812 ultrasound images. For B-mode, AUC and accuracy values ranged from 0.790 to 0.943 and from 72 to 89 % respectively. For elastography, AUC and accuracy values ranged from 0.847 to 0.985 and from 79 to 95 % respectively. Sixteen patients were available for the survival analysis. A total of 52 radiomic features were extracted. Two texture features from B-mode (Conventional mean and GLZLM_SZLGE) and one texture feature from strain-elastography (GLZLM_LZHGE) were significantly associated with OS. CONCLUSIONS Automated processing of ioUS images through deep learning can generate high-precision classification algorithms. Radiomic tumor region features in B-mode and elastography appear to be significantly associated with OS in GBM.

The Role of Elastography in Fibroadenoma of the Breast and Differentiating it from its Mimics

Background Fibroadenomas are one of the most common benign diseases of the breast varying in number and size in all quadrants of the breast. Less than 2% of lesions with the typical ultrasound features of a fibroadenoma, are found to be malignant on biopsy. Aim of the Work To evaluate the diagnostic performance of combined ultrasonography and strain elastography and to compare the imaging characteristics of fibroadenomas with other benign and malignant mimics and assess the results by histopathological characteristics. Patients and Methods Type of Study: Cross-sectional study. Study Setting: The study was conducted at Ain Shams University Hospitals, Radiodiagnosis department. Study Period: 6 months (From March 2020 till September 2020). Study Population: Female Patients with breast lumps. Results Ultrasound elastography combined with B-mode ultrasonography had shown variable degrees of sensitivity and specificity. However, in our study it showed high sensitivity and specificity as compared to other studies. The most reliable tool in our study was the strain ratio having the highest sensitivity and specificity. Conclusion Strain elastography combined to B-mode sonography improves the diagnostic performance of differentiating benign from malignant well circumscribed breast lesions with an additional benefit of preventing unnecessary biopsies and speeding up the diagnosis of malignant lesions instead of follow up of the patient for early diagnosis of malignancies.