Differentiation Between G1 and G2/G3 Phyllodes Tumors of Breast Using Mammography and Mammographic Texture Analysis

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.

The use of mammography-based radiomics nomograms for the preoperative prediction of the histological grade of invasive ductal carcinoma

BACKGROUND

Accurate prediction of the grade of invasive ductal carcinoma (IDC) before treatment is vital for individualized therapy and improving patient outcomes. This study aimed to develop and validate a mammography-based radiomics nomogram that would incorporate the radiomics signature and clinical risk factors in the preoperative prediction of the histological grade of IDC.

METHODS

The data of 534 patients from our hospital with pathologically confirmed IDC (374 in the training cohort and 160 in the validation cohort) were retrospectively analyzed. A total of 792 radiomics features were extracted from the patients' craniocaudal and mediolateral oblique view images. A radiomics signature was generated using the least absolute shrinkage and selection operator method. Multivariate logistic regression was adopted to establish a radiomics nomogram, the utility of which was evaluated using a receiver-operating characteristic curve, calibration curve, and decision curve analysis (DCA).

RESULTS

The radiomics signature was found to have a significant correlation with histological grade (P < 0.01), but the efficacy of the model is limited. The radiomics nomogram, which incorporated the radiomics signature and spicule sign into mammography, showed good consistency and discrimination in both the training cohort [area under the curve (AUC) = 0.75] and the validation cohort (AUC = 0.75). The calibration curves and DCA demonstrated the clinical usefulness of the proposed radiomics nomogram model.

CONCLUSIONS

A radiomics nomogram based on the radiomics signature and spicule sign can be used to predict the histological grade of IDC and assist in clinical decision-making for patients with IDC.

Global Radiomic Features from Mammography for Predicting Difficult-To-Interpret Normal Cases

This work aimed to investigate whether global radiomic features (GRFs) from mammograms can predict difficult-to-interpret normal cases (NCs). Assessments from 537 readers interpreting 239 normal mammograms were used to categorise cases as 120 difficult-to-interpret and 119 easy-to-interpret based on cases having the highest and lowest difficulty scores, respectively. Using lattice- and squared-based approaches, 34 handcrafted GRFs per image were extracted and normalised. Three classifiers were constructed: (i) CC and (ii) MLO using the GRFs from corresponding craniocaudal and mediolateral oblique images only, based on the random forest technique for distinguishing difficult- from easy-to-interpret NCs, and (iii) CC + MLO using the median predictive scores from both CC and MLO models. Useful GRFs for the CC and MLO models were recognised using a scree test. The CC and MLO models were trained and validated using the leave-one-out-cross-validation. The models' performances were assessed by the AUC and compared using the DeLong test. A Kruskal-Wallis test was used to examine if the 34 GRFs differed between difficult- and easy-to-interpret NCs and if difficulty level based on the traditional breast density (BD) categories differed among 115 low-BD and 124 high-BD NCs. The CC + MLO model achieved higher performance (0.71 AUC) than the individual CC and MLO model alone (0.66 each), but statistically non-significant difference was found (all p > 0.05). Six GRFs were identified to be valuable in describing difficult-to-interpret NCs. Twenty features, when compared between difficult- and easy-to-interpret NCs, differed significantly (p < 0.05). No statistically significant difference was observed in difficulty between low- and high-BD NCs (p = 0.709). GRF mammographic analysis can predict difficult-to-interpret NCs.

Machine Learning-Based Ultrasound Texture Analysis in Differentiation of Benign Phyllodes Tumors from Borderline-Malignant Phyllodes Tumors

PURPOSE

 Phyllodes tumors (PTs) are uncommon fibroepithelial breast lesions that are classified as three different forms as benign phyllodes tumor (BPT), borderline phyllodes tumor (BoPT), and malignant phyllodes tumor (MPT). Conventional radiologic methods make only a limited contribution to exact diagnosis, and texture analysis data increase the diagnostic performance. In this study, we aimed to evaluate the contribution of texture analysis of US images (TAUI) of PTs in order to discriminate between BPTs and BoPTs-MPTs.

METHODS

 The number of patients was 63 (41 BPTs, 12 BoPTs, and 10 MPTs). Patients were divided into two groups (Group 1-BPT, Group 2-BoPT/MPT). TAUI with LIFEx software was performed retrospectively. An independent machine learning approach, MATLAB R2020a (Math- Works, Natick, Massachusetts) was used with the dataset with p < 0.004. Two machine learning approaches were used to build prediction models for differentiating between Group 1 and Group 2. Receiver operating characteristics (ROC) curve analyses were performed to evaluate the diagnostic performance of statistically significant texture data between phyllodes subgroups.

RESULTS

 In TAUI, 10 statistically significant second order texture values were identified as significant factors capable of differentiating among the two groups (p < 0.05). Both of the models of our dataset make a diagnostic contribution to the discrimination between BopTs-MPTs and BPTs.

CONCLUSION

 In PTs, US is the main diagnostic method. Adding machine learning-based TAUI to conventional US findings can provide optimal diagnosis, thereby helping to choose the correct surgical method. Consequently, decreased local recurrence rates can be achieved.

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.

Pretreatment Multiparametric MRI-Based Radiomics Analysis for the Diagnosis of Breast Phyllodes Tumors

BACKGROUND

Preoperative pathological grading assessment is important for patients with breast phyllodes tumors (PTs).

PURPOSE

To develop and validate a clinical-radiomics model based on multiparametric MRI and clinical information for the pretreatment differential diagnosis of PTs.

STUDY TYPE

Retrospective.

POPULATION

A total of 216 patients with PTs, 133 in the training cohort (55 benign PTs [BPTs] and 78 borderline/malignant PTs [BMPTs]) and 83 in the validation cohort (28 BPTs and 55 BMPTs).

FIELD STRENGTH/SEQUENCE

1.5 T and 3 T; T2-weighted imaging (T2WI), precontrast T1-weighted imaging (T1WI) and dynamic contrast-enhanced T1-weighted imaging (DCE-T1WI).

ASSESSMENT

A total of 3138 radiomics features were computed to decode the imaging phenotypes of PTs. To build the classification models, the following workflow was followed: minimum-maximum scaling normalization method, recursive feature elimination based on ridge regression (Ridge-RFE), synthetic minority oversampling technique, and support vector machine classifier. We established several models based on the statistically significant features (Ridge-RFE selected) of each sequence to distinguish BPTs from BMPTs, including precontrast T1WI model, DCE-T1WI phase 1 model, T1WI feature fusion model, T2WI model, T1WI + T2WI model, clinical feature model, conventional MRI characteristics model, and combined clinical-radiomics model.

STATISTICAL TESTS

Univariate analysis was utilized to compare variables between the BPT and BMPT groups. The receiver operating characteristic curve (ROC) analysis was used to evaluate the diagnostic performance of these models.

RESULTS

In the training cohort, the clinical-radiomics model had excellent diagnostic efficiency, with an area under ROC (AUC) of 0.91 ± 0.02 (95% CI: 0.87-0.94). In the validation cohort, the AUCs were 0.79 ± 0.05 (95% CI: 0.70-0.87) for the combined model and 0.77 ± 0.05 (95% CI: 0.67-0.85) for the radiomics model.

DATA CONCLUSION

Compared with conventional MRI characteristics, radiomics features extracted from multiparametric MRI are helpful for improving the accuracy of differentiating the pathological grades of PTs preoperatively. The model based on radiomics and clinical information is expected to become a potential noninvasive tool for the assessment of PTs grades.

EVIDENCE LEVEL

4 TECHNICAL EFFICACY: Stage 2.

The Predictive Value of Magnetic Resonance Imaging-based Texture Analysis in Evaluating Histopathological Grades of Breast Phyllodes Tumor

PURPOSE

Knowing the distinction between benign and borderline/malignant phyllodes tumors (PTs) can help in the surgical treatment course. Herein, we investigated the value of magnetic resonance imaging-based texture analysis (MRI-TA) in differentiating between benign and borderline/malignant PTs.

METHODS

Forty-three women with 44 histologically proven PTs underwent breast MRI before surgery and were classified into benign (n = 26) and borderline/malignant groups (n = 18 [15 borderline, 3 malignant]). Clinical and routine MRI parameters (CRMP) and MRI-TA were used to distinguish benign from borderline/malignant PT. In total, 298 texture parameters were extracted from fat-suppression (FS) T2-weighted, FS unenhanced T1-weighted, and FS first-enhanced T1-weighted sequences. To evaluate the diagnostic performance, receiver operating characteristic curve analysis was performed for the K-nearest neighbor classifier trained with significantly different parameters of CRMP, MRI sequence-based TA, and the combination strategy.

RESULTS

Compared with benign PTs, borderline/malignant ones presented a higher local recurrence (p = 0.045); larger size (p < 0.001); different time-intensity curve pattern (p = 0.010); and higher frequency of strong lobulation (p = 0.024), septation enhancement (p = 0.048), cystic component (p = 0.023), and irregular cystic wall (p = 0.045). TA of FS T2-weighted images (0.86) showed a significantly higher area under the curve (AUC) than that of FS unenhanced T1-weighted (0.65, p = 0.010) or first-enhanced phase (0.72, p = 0.049) images. The texture parameters of FS T2-weighted sequences tended to have a higher AUC than CRMP (0.79, p = 0.404). Additionally, the combination strategy exhibited a similar AUC (0.89, p = 0.622) in comparison with the texture parameters of FS T2-weighted sequences.

CONCLUSION

MRI-TA demonstrated good predictive performance for breast PT pathological grading and could provide surgical planning guidance. Clinical data and routine MRI features were also valuable for grading PTs.

The Potential Value of Texture Analysis Based on Dynamic Contrast-Enhanced MR Images in the Grading of Breast Phyllode Tumors

Purpose

To explore the value of texture analysis (TA) based on dynamic contrast-enhanced MR (DCE-MR) images in the differential diagnosis of benign phyllode tumors (BPTs) and borderline/malignant phyllode tumors (BMPTs).

Methods

A total of 47 patients with histologically proven phyllode tumors (PTs) from November 2012 to March 2020, including 26 benign BPTs and 21 BMPTs, were enrolled in this retrospective study. The whole-tumor texture features based on DCE-MR images were calculated, and conventional imaging findings were evaluated according to the Breast Imaging Reporting and Data System (BI-RADS). The differences in the texture features and imaging findings between BPTs and BMPTs were compared; the variates with statistical significance were entered into logistic regression analysis. The receiver operating characteristic (ROC) curve was used to assess the diagnostic performance of models from image-based analysis, TA, and the combination of these two approaches.

Results

Regarding texture features, three features of the histogram, two features of the gray-level co-occurrence matrix (GLCM), and three features of the run-length matrix (RLM) showed significant differences between the two groups (all p < 0.05). Regarding imaging findings, however, only cystic wall morphology showed significant differences between the two groups (p = 0.014). The areas under the ROC curve (AUCs) of image-based analysis, TA, and the combination of these two approaches were 0.687 (95% CI, 0.518–0.825, p = 0.014), 0.886 (95% CI, 0.760–0.960, p < 0.0001), and 0.894 (95% CI, 0.754–0.970, p < 0.0001), respectively.

Conclusion

TA based on DCE-MR images has potential in differentiating BPTs and BMPTs.

Predicting the pathological grade of breast phyllodes tumors: a nomogram based on clinical and magnetic resonance imaging features

OBJECTIVE

To explore the potential factors related to the pathological grade of breast phyllodes tumors (PTs) and to establish a nomogram to improve their differentiation ability.

METHODS

Patients with PTs diagnosed by post-operative pathology who underwent pretreatment magnetic resonance imaging (MRI) from January 2015 to June 2020 were retrospectively reviewed. Traditional clinical features and MRI features evaluated according to the fifth BI-RADS were analyzed by statistical methods and introduced to a stepwise multivariate logistic regression analysis to develop a prediction model. Then, a nomogram was developed to graphically predict the probability of non-benign (borderline/malignant) PTs.

RESULTS

Finally, 61 benign, 73 borderline and 48 malignant PTs were identified in 182 patients. Family history of tumor, diameter, lobulation, cystic component, signal on fat saturated T₂ weighted imaging (FS T₂WI), BI-RADS category and time-signal intensity curve (TIC) patterns were found to be significantly different between benign and non-benign PTs. The nomogram was finally developed based on five risk factors: family history of tumor, lobulation, cystic component, signal on FS T₂WI and internal enhancement. The AUC of the nomogram was 0.795 (95% CI: 0.639, 0.835).

CONCLUSION

Family history of tumor, lobulation, cystic components, signals on FS T₂WI and internal enhancement are independent predictors of non-benign PTs. The prediction nomogram developed based on these features can be used as a supplemental tool to pre-operatively differentiate PTs grades.

ADVANCES IN KNOWLEDGE

More sample size and characteristics were used to explore the factors related to the pathological grade of PTs and establish a predictive nomogram for the first time.

Radiomics Based on CECT in Differentiating Kimura Disease From Lymph Node Metastases in Head and Neck: A Non-Invasive and Reliable Method

Background: Kimura disease may be easily misdiagnosed as malignant tumors such as lymph node metastases based on imaging and clinical symptoms. The aim of this article is to investigate whether the radiomic features and the model based on the features on venous-phase contrast-enhanced CT (CECT) images can distinguish Kimura disease from lymph node metastases in the head and neck.

Methods: A retrospective analysis of 14 patients of head and neck Kimura disease (a total of 38 enlarged lymph nodes) and 39 patients with head and neck lymph node metastases (a total of 39 enlarged lymph nodes), confirmed by biopsy or surgery resection, was conducted. All patients accepted CECT within 10 days before biopsy or surgery resection. Radiomic features based on venous-phase CECT were generated automatically from Artificial-Intelligence Kit (AK) software. All lymph nodes were randomly divided into the training set (n = 54) and testing set (n = 23) in a ratio of 7:3. ANOVA + Mann–Whitney, Spearman correlation, least absolute shrinkage and selection operator, and Gradient Descent were introduced for the reduction of the highly redundant features. Binary logistic regression model was constructed based on the selected features. Receiver operating characteristic was used to evaluate the diagnostic performance of the features and the model. Finally, a nomogram was established for model application.

Results: Seven features were screened out at the end. Significant difference was found between the two groups for all the features with area under the curves (AUCs) ranging from 0.759 to 0.915. The AUC of the model's identification performance was 0.970 in the training group and 0.977 in the testing group. The disease discrimination efficiency of the model was better than that of any single feature.

Conclusions: The radiomic features and the model based on these features on venous-phase CECT images had very good performance for the discrimination between Kimura disease and lymph node metastases in the head and neck.

Magnetic resonance imaging semantic and quantitative features analyses: an additional diagnostic tool for breast phyllodes tumors

OBJECTIVE

This study aimed to differentiate benign and non-benign (borderline/malignant) phyllodes tumors of the breast by the semantic and quantitative features in magnetic resonance imaging (MRI).

METHODS

The female patients, diagnosed with phyllodes tumors by MRI and pathological test, were retrospectively selected from December, 2006 to April, 2019. The MRI of benign, borderline and malignant phyllodes tumors was analyzed using 8 semantic features and 20 computed quantitative features from diffuse contrast-enhanced magnetic resonance imaging (DCE-MRI). The semantic features were analyzed by univariate analysis. The least absolute shrinkage and selection operator (LASSO) method was used to identify the optimal subset of MRI quantitative features. According to the results from multivariate logistic regression for the semantic and quantitative features, the model was constructed to differentiate benign and non-benign (borderline/malignant) phyllodes tumors.

RESULTS

Thirty-two benign (58.18%), 13 borderline (23.64%) and 10 malignant (18.18%) phyllodes tumors were identified in 54 patients. Five semantic features were proved to be significantly correlated with pathologic grade, including size, the T1 weighted image signal intensity, fat-saturated T2-weighted image signal intensity, enhanced signal intensity, and kinetic curve pattern. With the analysis of LASSO method, three quantitative texture features with significant predictive ability were selected. The model combining both the semantic and quantitative features was proved to have good performance in differentiation on phyllodes tumors, yielding an area under receiver operating characteristic curve, accuracy, sensitivity and specificity of 0.893, 0.933, 1.000, and 0.818, respectively.

CONCLUSION

The constructed model based on the semantic and quantitative features of DCE-MRI can significantly improve the differential diagnosis of phyllodes tumors in breast.

Evaluation of Texture Analysis for the Differential Diagnosis of Mass-Forming Pancreatitis From Pancreatic Ductal Adenocarcinoma on Contrast-Enhanced CT Images

Purpose: To investigate the potential of computed tomography (CT) imaging features and texture analysis to differentiate between mass-forming pancreatitis (MFP) and pancreatic ductal adenocarcinoma (PDAC).

Materials and Methods: Thirty patients with pathologically proved MFP and 79 patients with PDAC were included in this study. Clinical data and CT imaging features of the two lesions were evaluated. Texture features were extracted from arterial and portal phase CT images using commercially available software (AnalysisKit). Multivariate logistic regression analyses were used to identify relevant CT imaging and texture parameters to discriminate MFP from PDAC. Receiver operating characteristic curves were performed to determine the diagnostic performance of predictions.

Results: MFP showed a larger size compared to PDAC (p = 0.009). Cystic degeneration, pancreatic ductal dilatation, vascular invasion, and pancreatic sinistral portal hypertension were more frequent and duct penetrating sign was less frequent in PDAC compared to MFP. Arterial CT attenuation, arterial, and portal enhancement ratios of MFP were higher than PDAC (p < 0.05). In multivariate analysis, arterial CT attenuation and pancreatic duct penetrating sign were independent predictors. Texture features in arterial phase including SurfaceArea, Percentile40, InverseDifferenceMoment_angle90_offset4, LongRunEmphasis_angle45_offset4, and uniformity were independent predictors. Texture features in portal phase including LongRunEmphasis_angle135_offset7, VoxelValueSum, LongRunEmphasis_angle135_offset4, and GLCMEntropy_angle45_offset1 were independent predictors. Areas under the curve of imaging feature-based, texture feature-based in arterial and portal phases, and the combined models were 0.84, 0.96, 0.93, and 0.98, respectively.

Conclusions: CT texture analysis demonstrates great potential to differentiate MFP from PDAC.

Differentiation Between G1 and G2/G3 Phyllodes Tumors of Breast Using Mammography and Mammographic Texture Analysis

Purpose: To determine the potential of mammography (MG) and mammographic texture analysis in differentiation between Grade 1 (G1) and Grade 2/ Grade 3 (G2/G3) phyllodes tumors (PTs) of breast.

Materials and methods: A total of 80 female patients with histologically proven PTs were included in this study. 45 subjects who underwent pretreatment MG from 2010 to 2017 were retrospectively analyzed, including 14 PTs G1 and 31 PTs G2/G3. Tumor size, shape, margin, density, homogeneity, presence of fat, or calcifications, a halo-sign as well as some indirect manifestations were evaluated. Texture analysis features were performed using commercial software. Receiver operating characteristic curve (ROC) was used to determine the sensitivity and specificity of prediction.

Results: G2/G3 PTs showed a larger size (>4.0 cm) compared to PTs G1 (64.52 vs. 28.57%, p = 0.025). A strong lobulation or multinodular confluent was more common in G2/G3 PTs compared to PTs G1 (64.52 vs. 14.29%, p = 0.004). Significant differences were also observed in tumors' growth speed and clinical manifestations (p = 0.007, 0.022, respectively). Ten texture features showed significant differences between the two groups (p < 0.05), Correlation_AllDirection_offset7_SD and ClusterProminence_AllDirection_offset7_SD were independent risk factors. The area under the curve (AUC) of imaging-based diagnosis, texture analysis-based diagnosis and the combination of the two approaches were 0.805, 0.730, and 0.843 (90.3% sensitivity and 85.7% specificity).

Conclusions: Texture analysis has great potential to improve the diagnostic efficacy of MG in differentiating PTs G1 from PTs G2/G3.