A Radiomics Approach for the Classification of Fibroepithelial Lesions on Breast Ultrasonography

Differentiation Between Phyllodes Tumors and Fibroadenomas of Breast Using Mammography-based Machine Learning Methods: A Preliminary Study

OBJECTIVE

To investigate the diagnostic performance of a mammography-based radiomics model for distinguishing phyllodes tumors (PTs) from fibroadenomas (FAs) of the breast.

MATERIALS AND METHODS

A total of 156 patients were retrospectively included (75 with PTs, 81 with FAs) and divided into training and validation groups at a ratio of 7:3. Radiomics features were extracted from craniocaudal and mediolateral oblique images. The least absolute shrinkage and selection operator (LASSO) algorithm and principal component analysis (PCA) were performed to select features. Three machine learning classifiers, including logistic regression (LR), K-nearest neighbor classifier (KNN) and support vector machine (SVM), were implemented in the radiomics model, imaging model and combined model. Receiver operating characteristic curves, area under the curve (AUC), sensitivity and specificity were computed.

RESULTS

Among 1084 features, the LASSO algorithm selected 17 features, and PCA further selected 6 features. Three machine learning classifiers yielded the same AUC of 0.935 in the validation group for the radiomics model. In the imaging model, KNN yielded the highest accuracy rate of 89.4% and AUC of 0.947 in the validation set. For the combined model, the SVM classifier reached the highest AUC of 0.918 with an accuracy rate of 86.2%, sensitivity of 83.9%, and specificity of 89.4% in the training group. In the validation group, LR yielded the highest AUC of 0.973. The combined model had a relatively higher AUC than the radiomics model or imaging model, especially in the validation group.

CONCLUSIONS

Mammography-based radiomics features demonstrate good diagnostic performance for discriminating PTs from FAs.

Mammography-based radiomics analysis and imaging features for predicting the malignant risk of phyllodes tumours of the breast

AIM

To determine whether the mammography (MG)-based radiomics analysis and MG/ultrasound (US) imaging features could predict the malignant risk of phyllodes tumours (PTs) of the breast.

MATERIALS AND METHODS

Seventy-five patients with PTs were included retrospectively (39 with benign PTs, 36 with borderline/malignant PTs) and divided into thetraining (n=52) and validation groups (n=23). The clinical information, MG and US imaging characteristics, and histogram features were extracted from craniocaudal (CC) and mediolateral oblique (MLO) images. The lesion region of interest (ROI) and perilesional ROI were delineated. Multivariate logistic regression analysis was performed to determine the malignant factors of PTs. Receiver operating characteristic (ROC) curves were generated, and the area under the curve (AUC), sensitivity, and specificity were calculated.

RESULTS

There was no significant difference found in the clinical or MG/US features between benign and borderline/malignant PTs. In the lesion ROI, variance in the CC view and mean and variance in the MLO view were independent predictors. The AUC was 0.942, sensitivity and specificity were 96.3% and 92%, respectively, in the training group. In the validation group, the AUC was 0.879, the sensitivity was 91.7%, and the specificity was 81.8%. In the perilesional ROI, the AUCs were 0.904 and 0.939, sensitivities were 88.9% and 91.7%, and the specificities were 92% and 90.9% in the training and validation groups, respectively.

CONCLUSIONS

MG-based radiomic features could predict the risk of malignancy of patients with PTs and may be used as a potential tool to differentiate benign and borderline/malignant PTs.

Differential Diagnosis of DCIS and Fibroadenoma Based on Ultrasound Images: a Difference-Based Self-Supervised Approach

Differentiation of ductal carcinoma in situ (DCIS, a precancerous lesion of the breast) from fibroadenoma (FA) using ultrasonography is significant for the early prevention of malignant breast tumors. Radiomics-based artificial intelligence (AI) can provide additional diagnostic information but usually requires extensive labeling efforts by clinicians with specialized knowledge. This study aims to investigate the feasibility of differentially diagnosing DCIS and FA using ultrasound radiomics-based AI techniques and further explore a novel approach that can reduce labeling efforts without sacrificing diagnostic performance. We included 461 DCIS and 651 FA patients, of whom 139 DCIS and 181 FA patients constituted a prospective test cohort. First, various feature engineering-based machine learning (FEML) and deep learning (DL) approaches were developed. Then, we designed a difference-based self-supervised (DSS) learning approach that only required FA samples to participate in training. The DSS approach consists of three steps: (1) pretraining a Bootstrap Your Own Latent (BYOL) model using FA images, (2) reconstructing images using the encoder and decoder of the pretrained model, and (3) distinguishing DCIS from FA based on the differences between the original and reconstructed images. The experimental results showed that the trained FEML and DL models achieved the highest AUC of 0.7935 (95% confidence interval, 0.7900-0.7969) on the prospective test cohort, indicating that the developed models are effective for assisting in differentiating DCIS from FA based on ultrasound images. Furthermore, the DSS model achieved an AUC of 0.8172 (95% confidence interval, 0.8124-0.8219), indicating that our model outperforms the conventional radiomics-based AI models and is more competitive.

Preoperative prediction of axillary lymph node metastasis in patients with breast cancer based on radiomics of gray-scale ultrasonography

Background

To investigate the performance of a radiomics model based on gray-scale ultrasonography (US) for the preoperative non-invasive prediction of ipsilateral axillary lymph node (ALN) metastasis in patients with breast cancer (BC).

Methods

A total of 192 pathologically confirmed BC patients were included in this study. The training set was comprised of 132 patients from hospital 1 and the test set was comprised of 60 patients from hospital 2. All patients underwent US before percutaneous core biopsy and the results of ALN status reported by a radiologist with 12 years of experience were recorded. Radiomic features were extracted from the gray-scale US images. Max-relevance and min-redundancy (MRMR) and least absolute shrinkage and selection operator (LASSO) were used for data dimension reduction and feature selection. A radiomics model was constructed using LASSO and was validated using the leave group out cross-validation (LGOCV) method. The performance of the model was validated with receiver operating characteristic (ROC), calibration curve, and decision curve analysis.

Results

A total of 860 features were extracted from the gray-scale US images of each breast lesion, and 9 radiomic features were selected for model construction. The area under the curve (AUC), sensitivity, and specificity of the model for predicting ALN metastasis were 0.85, 78.9%, and 77.3% in the training set and 0.65, 68.0%, and 79.4% in the test set, respectively. The prediction performance of the model was significantly higher than that of the radiologist (AUC: 0.85 vs. 0.59, P<0.01) in the training set and was slightly higher than that of the radiologist (AUC: 0.65 vs. 0.63, P>0.05) in the test set.

Conclusions

The non-invasive radiomics model has the ability to predict ALN metastasis for patients with breast cancer and may outperform US-reported ALN status performed by the radiologist.

Radiomics Score Combined with ACR TI-RADS in Discriminating Benign and Malignant Thyroid Nodules Based on Ultrasound Images: A Retrospective Study

This study aimed to explore the ability of combination model of ultrasound radiomics score (Rad-score) and the thyroid imaging, reporting and data system by the American College of Radiology (ACR TI-RADS) in predicting benign and malignant thyroid nodules (TNs). Up to 286 radiomics features were extracted from ultrasound images of TNs. By using the lowest probability of classification error and average correlation coefficients (POE + ACC) and the least absolute shrinkage and selection operator (LASSO), we finally selected four features to establish Rad-score (Vertl-RLNonUni, Vertl-GLevNonU, WavEnLH-s4 and WavEnHL-s5). DeLong’s test and decision curve analysis (DCA) showed that the method of combining Rad-score and ACR TI-RADS had the best performance (the area under the receiver operating characteristic curve (AUC = 0.913 (95% confidence interval (CI), 0.881–0.939) and 0.899 (95%CI, 0.840–0.942) in the training group and verification group, respectively), followed by ACR TI-RADS (AUC = 0.898 (95%CI, 0.863–0.926) and 0.870 (95%CI, 0.806–0.919) in the training group and verification group, respectively), and followed by Rad-score (AUC = 0.750 (95%CI, 0.704–0.792) and 0.750 (95%CI, 0.672–0.817) in the training group and verification group, respectively). We concluded that the ability of ultrasound Rad-score to distinguish benign and malignant TNs was not as good as that of ACR TI-RADS, and the ability of the combination model of Rad-score and ACR TI-RADS to discriminate benign and malignant TNs was better than ACR TI-RADS or Rad-score alone. Ultrasound Rad-score might play a potential role in improving the differentiation of malignant TNs from benign TNs.

Radiomics: A Primer for Breast Radiologists

Radiomics has a long-standing history in breast imaging with computer-aided detection (CAD) for screening mammography developed in the late 20th century. Although conventional CAD had widespread adoption, the clinical benefits for experienced breast radiologists were debatable due to high false-positive marks and subsequent increased recall rates. The dramatic growth in recent years of artificial intelligence-based analysis, including machine learning and deep learning, has provided numerous opportunities for improved modern radiomics work in breast imaging. There has been extensive radiomics work in mammography, digital breast tomosynthesis, MRI, ultrasound, PET-CT, and combined multimodality imaging. Specific radiomics outcomes of interest have been diverse, including CAD, prediction of response to neoadjuvant therapy, lesion classification, and survival, among other outcomes. Additionally, the radiogenomics subfield that correlates radiomics features with genetics has been very proliferative, in parallel with the clinical validation of breast cancer molecular subtypes and gene expression assays. Despite the promise of radiomics, there are important challenges related to image normalization, limited large unbiased data sets, and lack of external validation. Much of the radiomics work to date has been exploratory using single-institution retrospective series for analysis, but several promising lines of investigation have made the leap to clinical practice with commercially available products. As a result, breast radiologists will increasingly be incorporating radiomics-based tools into their daily practice in the near future. Therefore, breast radiologists must have a broad understanding of the scope, applications, and limitations of radiomics work.

Identifying ultrasound features of positive expression of Ki67 and P53 in breast cancer using radiomics

PURPOSE

To examine the relationship between ultrasonic findings and positive expression of Ki67 and P53 in breast cancer.

MATERIAL AND METHODS

Surgical resection specimens of 263 breast cancer lesions were examined. Ultrasound examination and pathological examination were performed on each lesion for retrospective analysis. We applied regression analysis to the ultrasonic features related to the positive expression of Ki67 and P53 and obtained the corresponding models. To analyze diagnostic efficiency, we calculated the area under the curve (AUC). Additionally, we created a heat map to show the results of the cluster analysis.

RESULTS

Lesions with higher Ki67 expression were associated with posterior acoustic enhancement, absence of an echo halo and a higher Breast Imaging Reporting and Data System (BI-RADS) category. P53-positive cancer were associated with an absence of an echo halo and a higher BI-RADS category. The AUC of the regression models of Ki67 and P53 was 0.78 and 0.71, respectively.

CONCLUSIONS

Our study revealed that breast cancer ultrasonic findings were closely related to expression of molecular indicators, suggesting that ultrasound can be used to provide useful information to clinicians.