Clinical value of radiomics and machine learning in breast ultrasound: a multicenter study for differential diagnosis of benign and malignant lesions

Combining Radiomics and Autoencoders to Distinguish Benign and Malignant Breast Tumors on US Images

Background US is clinically established for breast imaging, but its diagnostic performance depends on operator experience. Computer-assisted (real-time) image analysis may help in overcoming this limitation. Purpose To develop precise real-time-capable US-based breast tumor categorization by combining classic radiomics and autoencoder-based features from automatically localized lesions. Materials and Methods A total of 1619 B-mode US images of breast tumors were retrospectively analyzed between April 2018 and January 2024. nnU-Net was trained for lesion segmentation. Features were extracted from tumor segments, bounding boxes, and whole images using either classic radiomics, autoencoder, or both. Feature selection was performed to generate radiomics signatures, which were used to train machine learning algorithms for tumor categorization. Models were evaluated using the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity and were statistically compared with histopathologically or follow-up-confirmed diagnosis. Results The model was developed on 1191 (mean age, 61 years ± 14 [SD]) female patients and externally validated on 50 (mean age, 55 years ± 15]). The development data set was divided into two parts: testing and training lesion segmentation (419 and 179 examinations) and lesion categorization (503 and 90 examinations). nnU-Net demonstrated precision and reproducibility in lesion segmentation in test set of data set 1 (median Dice score [DS]: 0.90 [IQR, 0.84-0.93]; P = .01) and data set 2 (median DS: 0.89 [IQR, 0.80-0.92]; P = .001). The best model, trained with 23 mixed features from tumor bounding boxes, achieved an AUC of 0.90 (95% CI: 0.83, 0.97), sensitivity of 81% (46 of 57; 95% CI: 70, 91), and specificity of 87% (39 of 45; 95% CI: 77, 87). No evidence of difference was found between model and human readers (AUC = 0.90 [95% CI: 0.83, 0.97] vs 0.83 [95% CI: 0.76, 0.90]; P = .55 and 0.90 vs 0.82 [95% CI: 0.75, 0.90]; P = .45) in tumor classification or between model and histopathologically or follow-up-confirmed diagnosis (AUC = 0.90 [95% CI: 0.83, 0.97] vs 1.00 [95% CI: 1.00,1.00]; P = .10). Conclusion Precise real-time US-based breast tumor categorization was developed by mixing classic radiomics and autoencoder-based features from tumor bounding boxes. ClinicalTrials.gov identifier: NCT04976257 Published under a CC BY 4.0 license. Supplemental material is available for this article. See also the editorial by Bahl in this issue.

Research on the application of radiomics in breast cancer: A bibliometrics and visualization analysis

Breast cancer is the most prevalent form of cancer worldwide. Therefore, improved disease detection has emerged as a focal point in clinical studies. At the forefront of innovation, radiomics has the capability to extract comprehensive insights from medical images, ultimately enhancing the accuracy of diagnostic procedures. There has been rapid growth in the field of radiomics research on breast cancer in the past few years. We explored pertinent research articles in the Web of Science Core Collection database to gain a thorough understanding of breast cancer radiomics. We used CiteSpace to conduct a bibliometric analysis of the annual distribution of different nations, institutions, journals, authors, keywords, and references in the field of breast cancer radiomics. GraphPad Prism software was used to examine and graph yearly and country-specific trends and the proportions of publications. The tools utilized for the visualization of science mapping included CiteSpace and VOSviewer. Of the 891 publications, most were original articles (731, 91.09%) and a few were reviews (160, 8.91%). Most academic research has been published in China and the United States. The study centers predominantly consisted of major academic institutions, such as Fudan University and the Chinese Academy of Sciences, with some of their members being prominent figures in the field. Pinker, Katja has published the largest number of research papers. The majority of these studies have been published in medical journals focusing on radiology and oncology in recent years. In the realm of cutting-edge medical research, the top two keywords, magnetic resonance imaging and machine learning stand at the forefront as current areas of intense focus. Breast cancer radiomics is advancing rapidly, presenting numerous opportunities and obstacles. Our study of the literature in this academic area aimed to pinpoint the primary themes addressed in the studies and anticipate prospective avenues for research.

Adaptive Machine Learning Approach for Importance Evaluation of Multimodal Breast Cancer Radiomic Features

Breast cancer holds the highest diagnosis rate among female tumors and is the leading cause of death among women. Quantitative analysis of radiological images shows the potential to address several medical challenges, including the early detection and classification of breast tumors. In the P.I.N.K study, 66 women were enrolled. Their paired Automated Breast Volume Scanner (ABVS) and Digital Breast Tomosynthesis (DBT) images, annotated with cancerous lesions, populated the first ABVS+DBT dataset. This enabled not only a radiomic analysis for the malignant vs. benign breast cancer classification, but also the comparison of the two modalities. For this purpose, the models were trained using a leave-one-out nested cross-validation strategy combined with a proper threshold selection approach. This approach provides statistically significant results even with medium-sized data sets. Additionally it provides distributional variables of importance, thus identifying the most informative radiomic features. The analysis proved the predictive capacity of radiomic models even using a reduced number of features. Indeed, from tomography we achieved AUC-ROC 89.9 % using 19 features and 92.1 % using 7 of them; while from ABVS we attained an AUC-ROC of 72.3 % using 22 features and 85.8 % using only 3 features. Although the predictive power of DBT outperforms ABVS, when comparing the predictions at the patient level, only 8.7% of lesions are misclassified by both methods, suggesting a partial complementarity. Notably, promising results (AUC-ROC ABVS-DBT 71.8 % - 74.1 % ) were achieved using non-geometric features, thus opening the way to the integration of virtual biopsy in medical routine.

Optimizing breast cancer diagnosis with photoacoustic imaging: An analysis of intratumoral and peritumoral radiomics

Background

The differentiation between benign and malignant breast tumors extends beyond morphological structures to encompass functional alterations within the nodules. The combination of photoacoustic (PA) imaging and radiomics unveils functional insights and intricate details that are imperceptible to the naked eye.

Purpose

This study aims to assess the efficacy of PA imaging in breast cancer radiomics, focusing on the impact of peritumoral region size on radiomic model accuracy.

Materials and methods

From January 2022 to November 2023, data were collected from 358 patients with breast nodules, diagnosed via PA/US examination and classified as BI-RADS 3-5. The study used the largest lesion dimension in PA images to define the region of interest, expanded by 2 mm, 5 mm, and 8 mm, for extracting radiomic features. Techniques from statistics and machine learning were applied for feature selection, and logistic regression classifiers were used to build radiomic models. These models integrated both intratumoral and peritumoral data, with logistic regressions identifying key predictive features.

Results

The developed nomogram, combining 5 mm peritumoral data with intratumoral and clinical features, showed superior diagnostic performance, achieving an AUC of 0.950 in the training cohort and 0.899 in validation. This model outperformed those based solely on clinical features or other radiomic methods, with the 5 mm peritumoral region proving most effective in identifying malignant nodules.

Conclusion

This research demonstrates the significant potential of PA imaging in breast cancer radiomics, especially the advantage of integrating 5 mm peritumoral with intratumoral features. This approach not only surpasses models based on clinical data but also underscores the importance of comprehensive radiomic analysis in accurately characterizing breast nodules.

Predicting hormone receptor status in invasive breast cancer through radiomics analysis of long-axis and short-axis ultrasound planes

The hormone receptor (HR) status plays a significant role in breast cancer, serving as the primary guide for treatment decisions and closely correlating with prognosis. This study aims to investigate the predictive value of radiomics analysis in long-axis and short-axis ultrasound planes for distinguishing between HR-positive and HR-negative breast cancers. A cohort of 505 patients from two hospitals was stratified into discovery (Institute 1, 416 patients) and validation (Institute 2, 89 patients) cohorts. A comprehensive set of 788 ultrasound radiomics features was extracted from both long-axis and short-axis ultrasound planes, respectively. Utilizing least absolute shrinkage and selection operator (LASSO) regression analysis, distinct models were constructed for the long-axis and short-axis data. Subsequently, radiomics scores (Rad-scores) were computed for each patient. Additionally, a combined model was formulated by integrating data from long-axis and short-axis Rad-scores along with clinical factors. The diagnostic efficacy of all models was evaluated by the area under the receiver operating characteristic (ROC) curve (AUC). The long-axis and short-axis models, consisting of 11 features and 15 features, respectively, were established, yielding AUCs of 0.743 and 0.751 in the discovery cohort, and 0.795 and 0.744 in the validation cohort. The calculated long-axis and short-axis Rad-scores exhibited significant differences between HR-positive and HR-negative groups across all cohorts (all p < 0.001). Univariate analysis identified ultrasound-reported tumor size as an independent predictor. The combined model, incorporating long-axis and short-axis Rad-scores along with tumor size, achieved superior AUCs of 0.788 and 0.822 in the discovery and validation cohorts, respectively. The combined model effectively distinguishes between HR-positive and HR-negative breast cancers based on ultrasound radiomics features and tumor size, which may offer a valuable tool to facilitate treatment decision making and prognostic assessment.

[Preoperative Evaluation of Cervical Lymph Node Metastasis in Patients With Hashimoto's Thyroiditis Combined With Thyroid Papillary Carcinoma Using Machine Learning and Radiomics-Based Features: A Preliminary Study]

Objective

To analyze the radiomic and clinical features extracted from 2D ultrasound images of thyroid tumors in patients with Hashimoto's thyroiditis (HT) combined with papillary thyroid carcinoma (PTC) using machine learning (ML) models, and to explore the diagnostic performance of the method in making preoperative noninvasive identification of cervical lymph node metastasis (LNM).

Methods

A total of 528 patients with HT combined with PTC were enrolled and divided into two groups based on their pathological results of the presence or absence of LNM. The groups were subsequently designated the With LNM Group and the Without LNM Group. Three ultrasound doctors independently delineated the regions of interest and extracted radiomic features. Two modes, radiomic features and radiomics-clinical features, were used to construct random forest (RF), support vector machine (SVM), LightGBM, K-nearest neighbor (KNN), and XGBoost models. The performance of these five ML models in the two modes was evaluated by the receiver operating characteristic (ROC) curves on the test dataset, and SHapley Additive exPlanations (SHAP) was used for model visualization.

Results

All five ML models showed good performance, with area under the ROC curve (AUC) ranging from 0.798 to 0.921. LightGBM and XGBoost demonstrated the best performance, outperforming the other models (P<0.05). The ML models constructed with radiomics-clinical features performed better than those constructed using only radiomic features (P<0.05). The SHAP visualization of the best-performing models indicated that the anteroposterior diameter, superoinferior diameter, original_shape_VoxelVolume, age, wavelet-LHL_firstorder_10Percentile, and left-to-right diameter had the most significant effect on the LightGBM model. On the other hand, the superoinferior diameter, anteroposterior diameter, left-to-right diameter, original_shape_VoxelVolume, original_firstorder_InterquartileRange, and age had the most significant effect on the XGBoost model.

Conclusion

ML models based on radiomics and clinical features can accurately evaluate the cervical lymph node status in patients with HT combined with PTC. Among the 5 ML models, LightGBM and XGBoost demonstrate the best evaluation performance.

Radiomics analysis combining gray-scale ultrasound and mammography for differentiating breast adenosis from invasive ductal carcinoma

Objectives

To explore the utility of gray-scale ultrasound (GSUS) and mammography (MG) for radiomic analysis in distinguishing between breast adenosis and invasive ductal carcinoma (IDC).

Methods

Data from 147 female patients with pathologically confirmed breast lesions (breast adenosis: 61 patients; IDC: 86 patients) between January 2018 and December 2022 were retrospectively collected. A training cohort of 113 patients (breast adenosis: 50 patients; IDC: 63 patients) diagnosed from January 2018 to December 2021 and a time-independent test cohort of 34 patients (breast adenosis: 11 patients; IDC: 23 patients) diagnosed from January 2022 to December 2022 were included. Radiomic features of lesions were extracted from MG and GSUS images. The least absolute shrinkage and selection operator (LASSO) regression was applied to select the most discriminant features, followed by logistic regression (LR) to construct clinical and radiomic models, as well as a combined model merging radiomic and clinical features. Model performance was assessed using receiver operating characteristic (ROC) analysis.

Results

In the training cohort, the area under the curve (AUC) for radiomic models based on MG features, GSUS features, and their combination were 0.974, 0.936, and 0.991, respectively. In the test cohort, the AUCs were 0.885, 0.876, and 0.949, respectively. The combined model, incorporating clinical and all radiomic features, and the MG plus GSUS radiomics model were found to exhibit significantly higher AUCs than the clinical model in both the training cohort and test cohort (p<0.05). No significant differences were observed between the combined model and the MG plus GSUS radiomics model in the training cohort and test cohort (p>0.05).

Conclusion

The effectiveness of radiomic features derived from GSUS and MG in distinguishing between breast adenosis and IDC is demonstrated. Superior discriminatory efficacy is shown by the combined model, integrating both modalities.

Artificial intelligence-based, semi-automated segmentation for the extraction of ultrasound-derived radiomics features in breast cancer: a prospective multicenter study

PURPOSE

To investigate the feasibility of an artificial intelligence (AI)-based semi-automated segmentation for the extraction of ultrasound (US)-derived radiomics features in the characterization of focal breast lesions (FBLs).

MATERIAL AND METHODS

Two expert radiologists classified according to US BI-RADS criteria 352 FBLs detected in 352 patients (237 at Center A and 115 at Center B). An AI-based semi-automated segmentation was used to build a machine learning (ML) model on the basis of B-mode US of 237 images (center A) and then validated on an external cohort of B-mode US images of 115 patients (Center B).

RESULTS

A total of 202 of 352 (57.4%) FBLs were benign, and 150 of 352 (42.6%) were malignant. The AI-based semi-automated segmentation achieved a success rate of 95.7% for one reviewer and 96% for the other, without significant difference (p = 0.839). A total of 15 (4.3%) and 14 (4%) of 352 semi-automated segmentations were not accepted due to posterior acoustic shadowing at B-Mode US and 13 and 10 of them corresponded to malignant lesions, respectively. In the validation cohort, the characterization made by the expert radiologist yielded values of sensitivity, specificity, PPV and NPV of 0.933, 0.9, 0.857, 0.955, respectively. The ML model obtained values of sensitivity, specificity, PPV and NPV of 0.544, 0.6, 0.416, 0.628, respectively. The combined assessment of radiologists and ML model yielded values of sensitivity, specificity, PPV and NPV of 0.756, 0.928, 0.872, 0.855, respectively.

CONCLUSION

AI-based semi-automated segmentation is feasible, allowing an instantaneous and reproducible extraction of US-derived radiomics features of FBLs. The combination of radiomics and US BI-RADS classification led to a potential decrease of unnecessary biopsy but at the expense of a not negligible increase of potentially missed cancers.

Ultrasound-based radiomics for the differential diagnosis of breast masses: A systematic review and meta-analysis

OBJECTIVES

Ultrasound-based radiomics has demonstrated excellent diagnostic performance in differentiating benign and malignant breast masses. Given a few clinical studies on their diagnostic role, we conducted a meta-analysis of the potential effects of ultrasound-based radiomics for the differential diagnosis of breast masses, aiming to provide evidence-based medical basis for clinical research.

MATERIALS AND METHODS

We searched Embase, Web of Science, Cochrane Library, and PubMed databases from inception through to February 2023. The methodological quality assessment of the included studies was performed according to Quality Assessment of Diagnostic Accuracy Studies checklist. A diagnostic test accuracy systematic review and meta-analysis was performed in accordance with PRISMA guidelines. Sensitivity, specificity, and area under curve delineating benign and malignant lesions were recorded. We also used sensitivity analysis and subgroup analysis to explore potential sources of heterogeneity. Deeks' funnel plots was used to examine the publication bias.

RESULTS

A total of 11 studies were included in this meta-analysis. For the diagnosis of malignant breast masses worldwide, the overall mean rates of sensitivity and specificity of ultrasound-based radiomics were 0.90 (95% confidence interval [CI], 0.83-0.95) and 0.89 (95% CI, 0.82-0.94), respectively. The summary diagnostic odds ratio was 76 (95% CI, 26-219), and the area under the curve for the summary receiver operating characteristic curve was 0.95 (95% CI, 0.93-0.97).

CONCLUSION

Ultrasound-based radiomics has the potential to improve diagnostic accuracy to discriminate between benign and malignant breast masses, and could reduce unnecessary biopsies.

Differentiation of invasive ductal and lobular carcinoma of the breast using MRI radiomic features: a pilot study

Background

Breast cancer (BC) is one of the main causes of cancer-related mortality among women. For clinical management to help patients survive longer and spend less time on treatment, early and precise cancer identification and differentiation of breast lesions are crucial. To investigate the accuracy of radiomic features (RF) extracted from dynamic contrast-enhanced Magnetic Resonance Imaging (DCE MRI) for differentiating invasive ductal carcinoma (IDC) from invasive lobular carcinoma (ILC).

Methods

This is a retrospective study. The IDC of 30 and ILC of 28 patients from Dukes breast cancer MRI data set of The Cancer Imaging Archive (TCIA), were included. The RF categories such as shape based, Gray level dependence matrix (GLDM), Gray level co-occurrence matrix (GLCM), First order, Gray level run length matrix (GLRLM), Gray level size zone matrix (GLSZM), NGTDM (Neighbouring gray tone difference matrix) were extracted from the DCE-MRI sequence using a 3D slicer. The maximum relevance and minimum redundancy (mRMR) was applied using Google Colab for identifying the top fifteen relevant radiomic features. The Mann-Whitney U test was performed to identify significant RF for differentiating IDC and ILC. Receiver Operating Characteristic (ROC) curve analysis was performed to ascertain the accuracy of RF in distinguishing between IDC and ILC.

Results

Ten DCE MRI-based RFs used in our study showed a significant difference (p <0.001) between IDC and ILC. We noticed that DCE RF, such as Gray level run length matrix (GLRLM) gray level variance (sensitivity (SN) 97.21%, specificity (SP) 96.2%, area under curve (AUC) 0.998), Gray level co-occurrence matrix (GLCM) difference average (SN 95.72%, SP 96.34%, AUC 0.983), GLCM interquartile range (SN 95.24%, SP 97.31%, AUC 0.968), had the strongest ability to differentiate IDC and ILC.

Conclusions

MRI-based RF derived from DCE sequences can be used in clinical settings to differentiate malignant lesions of the breast, such as IDC and ILC, without requiring intrusive procedures.

Automated Breast Volume Scanner Is More Valuable Than Hand-Held Ultrasound in Diagnosis of Small Breast cancer: An Analysis of 725 Patients With 912 Lesions Evaluations

ABSTRACT

This study aimed to evaluate the clinical value of automated breast volume scanner (ABVS) compared with hand-held ultrasound (HHUS). From January 2015 to May 2019, a total of 912 breast lesions in 725 consecutive patients were included in this study. κ statistics were calculated to identify interobserver agreement of ABVS and HHUS. The diagnostic performance for ABVS and HHUS was expressed as the area under the receiver operating characteristic curve, as well as the corresponding 95% confidence interval, sensitivity, and specificity. The sensitivities of ABVS and HHUS were 95.95% and 93.69%, and the specificities were 85.47% and 81.20%, respectively. A difference that nearly reached statistical significance was observed in sensitivities between ABVS and HHUS (P = 0.0525). The specificity of ABVS was significantly higher than that of HHUS (P = 0.006). When lesions were classified according to their maximum diameter, the sensitivity and specificity of ABVS were significantly higher than HHUS for lesions ≤20 mm, while they made no statistical significance between ABVS and HHUS for lesions >20 mm. The interobserver agreement for ABVS was better than that of HHUS. Automated breast volume scanner was more valuable than HHUS in diagnosing breast cancer, especially for lesions ≤20 mm, and it could be a valuable diagnostic tool for breast cancer.

Fusion of transfer learning models with LSTM for detection of breast cancer using ultrasound images

Breast Cancer (BC) is one of the top reasons for fatality in women worldwide. As a result, timely identification is critical for successful therapy and excellent survival rates. Transfer Learning (TL) approaches have recently shown promise in aiding in the early recognition of BC. In this work, three TL models, MobileNetV2, ResNet50, and VGG16, were combined with LSTM to extract the features from Ultrasound Images (USIs). Furthermore, the Synthetic Minority Over-sampling Technique (SMOTE) with Tomek (SMOTETomek) was employed to balance the extracted features. The proposed method with VGG16 achieved an F1 score of 99.0 %, Matthews Correlation Coefficient (MCC) and Kappa Coefficient of 98.9 % with an Area Under Curve (AUC) of 1.0. The K-fold method was applied for cross-validation and achieved an average F1 score of 96 %. Moreover, the Gradient-weighted Class Activation Mapping (Grad-CAM) method was applied for visualization, and the Local Interpretable Model-agnostic Explanations (LIME) method was applied for interpretability. The Normal Approximation Interval (NAI) and bootstrapping methods were used to calculate Confidence Intervals (CIs). The proposed method achieved a Lower CI (LCI), Upper CI (UCI), and Mean CI (MCI) of 96.50 %, 99.75 %, and 98.13 %, respectively, with the NAI, while 95 % LCI of 93.81 %, an UCI of 96.00 %, and a bootstrap mean of 94.90 % with the bootstrap method. Furthermore, the performance of the six state-of-the-art (SOTA) TL models, such as Xception, NASNetMobile, InceptionResNetV2, MobileNetV2, ResNet50, and VGG16, were compared with the proposed method.

Potential added value of computed tomography radiomics to multimodal prediction models for benign and malignant breast tumors

Background

Early diagnosis is crucial to the treatment of breast cancer, but conventional imaging detection is challenging. Radiomics has the potential to improve early diagnostic efficacy in a noninvasive manner. This study examined whether integrating computed tomography (CT) radiomics information based on ultrasound (US) models can improve the efficacy of breast cancer prediction.

Methods

We retrospectively analyzed 420 patients with pathologically confirmed benign or malignant breast tumors. Clinical data and examination images were collected, and the population was divided into training (n=294) and validation (n=126) groups at a ratio of 7:3. The region of interest (ROI) was manually segmented along the tumor boundary using MaZda software, and the features of each ROI was extracted. After dimension reduction and screening, the best features were retained. Subsequently, random forest (RF), support vector machines, and K-nearest neighbor classifiers were used to establish prediction models in an US and combined-methods group.

Results

Finally, 8 of the 379 features were retained in the US group. Random forest was found to be the best model, and the area under the curve (AUC) of the training and validation groups was 0.90 [95% confidence interval (CI): 0.852-0.942] and 0.85 (95% CI: 0.775-0.930), respectively. Meanwhile, 12 of the 750 features were retained in the combined group. In this regard, random forest proved to be the best model, and the AUC of the training and validation group was 0.95 (95% CI: 0.918-0.981) and 0.92 (95% CI: 0.866-0.969), respectively. The calibration curve showed a good fit of the model. The decision curve showed that the clinical net benefit of the combined group was far greater than that of any single examination, and the prediction model of the combined group exhibited a degree of practical clinical value.

Conclusions

The combined model based on US and CT images has potential application value in the prognostic prediction of benign and malignant breast diseases.

Noninvasive Assessment of Tumor Histological Grade in Invasive Breast Carcinoma Based on Ultrasound Radiomics and Clinical Characteristics: A Multicenter Study

Rationale and Objectives: We aimed to develop and validate prediction models for histological grade of invasive breast carcinoma (BC) based on ultrasound radiomics features and clinical characteristics. Materials and Methods: A number of 383 patients with invasive BC were retrospectively enrolled and divided into a training set (207 patients), internal validation set (90 patients), and external validation set (86 patients). Ultrasound radiomics features were extracted from all the eligible patients. The Boruta method was used to identify the most useful features. Seven classifiers were adopted to developed prediction models. The output of the classifier with best performance was labeled as the radiomics score (Rad-score) and the classifier was selected as the Rad-score model. A combined model combining clinical factors and Rad-score was developed. The performance of the models was evaluated using receiver operating characteristic curve. Results: Seven radiomics features were selected from 788 candidate features. The logistic regression model performing best among the 7 classifiers in the internal and external validation sets was considered as Rad-score model, with areas under the receiver operating characteristic curve (AUC) values of 0.731 and 0.738. The tumor size was screened out as the risk factor and the combined model was developed, with AUC values of 0.721 and 0.737 in the internal and external validation sets. Furthermore, the 10-fold cross-validation demonstrated that the 2 models above were reliable and stable. Conclusion: The Rad-score model and combined model were able to predict histological grade of invasive BC, which may enable tailored therapeutic strategies for patients with BC in routine clinical use.

The value of radiomics model based on ultrasound image features in the differentiation between minimal breast cancer and small benign breast masses

BACKGROUND

Female breast cancer has surpassed lung cancer as the most common cancer, and is also the main cause of cancer death for women worldwide. Breast cancer <1 cm showed excellent survival rate. However, the diagnosis of minimal breast cancer (MBC) is challenging.

OBJECTIVE

The purpose of our research is to develop and validate an radiomics model based on ultrasound images for early recognition of MBC.

METHODS

302 breast masses with a diameter of <10 mm were retrospectively studied, including 159 benign and 143 malignant breast masses. The radiomics features were extracted from the gray-scale ultrasound image of the largest face of each breast mass. The maximum relevance minimum reduncancy and recursive feature elimination methods were used to screen. Finally, 10 features with the most discriminating value were selected for modeling. The random forest was used to establish the prediction model, and the rad-score of each mass was calculated. In order to evaluate the effectiveness of the model, we calculated and compared the area under the curve (AUC) value, sensitivity, specificity, accuracy, positive predictive value, and negative predictive value of the model and three groups with different experience in predicting small breast masses, and drew calibration curves and decision curves to test the stability and consistency of the model.

RESULTS

When we selected 10 radiomics features to calculate the rad-score, the prediction efficiency was the best, the AUC values for the training set and testing set were 0.840 and 0.793, which was significantly better than the insufficient experience group (AUC = 0.673), slightly better than the moderate experience group (AUC = 0.768), and was inferior to the experienced group (AUC = 0.877). The calibration curve and decision curve also showed that the radiomics model had satisfied stability and clinical application value.

CONCLUSION

The radiomics model based on ultrasound image features has a satisfied predictive ability for small breast masses, and is expected to become a potential tool for the diagnosis of MBC, and it is a zero cost (in terms of patient participation and imaging time).

Rapid Segmentation and Diagnosis of Breast Tumor Ultrasound Images at the Sonographer Level Using Deep Learning

BACKGROUND

Breast cancer is one of the most common malignant tumors in women. A noninvasive ultrasound examination can identify mammary-gland-related diseases and is well tolerated by dense breast, making it a preferred method for breast cancer screening and of significant clinical value. However, the diagnosis of breast nodules or masses via ultrasound is performed by a doctor in real time, which is time-consuming and subjective. Junior doctors are prone to missed diagnoses, especially in remote areas or grass-roots hospitals, due to limited medical resources and other factors, which bring great risks to a patient's health. Therefore, there is an urgent need to develop fast and accurate ultrasound image analysis algorithms to assist diagnoses.

METHODS

We propose a breast ultrasound image-based assisted-diagnosis method based on convolutional neural networks, which can effectively improve the diagnostic speed and the early screening rate of breast cancer. Our method consists of two stages: tumor recognition and tumor classification. (1) Attention-based semantic segmentation is used to identify the location and size of the tumor; (2) the identified nodules are cropped to construct a training dataset. Then, a convolutional neural network for the diagnosis of benign and malignant breast nodules is trained on this dataset. We collected 2057 images from 1131 patients as the training and validation dataset, and 100 images of the patients with accurate pathological criteria were used as the test dataset.

RESULTS

The experimental results based on this dataset show that the MIoU of tumor location recognition is 0.89 and the average accuracy of benign and malignant diagnoses is 97%. The diagnosis performance of the developed diagnostic system is basically consistent with that of senior doctors and is superior to that of junior doctors. In addition, we can provide the doctor with a preliminary diagnosis so that it can be diagnosed quickly.

CONCLUSION

Our proposed method can effectively improve diagnostic speed and the early screening rate of breast cancer. The system provides a valuable aid for the ultrasonic diagnosis of breast cancer.

Quo vadis Radiomics? Bibliometric analysis of 10-year Radiomics journey

OBJECTIVES

Radiomics is the high-throughput extraction of mineable and-possibly-reproducible quantitative imaging features from medical imaging. The aim of this work is to perform an unbiased bibliometric analysis on Radiomics 10 years after the first work became available, to highlight its status, pitfalls, and growing interest.

METHODS

Scopus database was used to investigate all the available English manuscripts about Radiomics. R Bibliometrix package was used for data analysis: a cumulative analysis of document categories, authors affiliations, country scientific collaborations, institution collaboration networks, keyword analysis, comprehensive of co-occurrence network, thematic map analysis, and 2021 sub-analysis of trend topics was performed.

RESULTS

A total of 5623 articles and 16,833 authors from 908 different sources have been identified. The first available document was published in March 2012, while the most recent included was released on the 31st of December 2021. China and USA were the most productive countries. Co-occurrence network analysis identified five words clusters based on top 50 authors' keywords: Radiomics, computed tomography, radiogenomics, deep learning, tomography. Trend topics analysis for 2021 showed an increased interest in artificial intelligence (n = 286), nomogram (n = 166), hepatocellular carcinoma (n = 125), COVID-19 (n = 63), and X-ray computed (n = 60).

CONCLUSIONS

Our work demonstrates the importance of bibliometrics in aggregating information that otherwise would not be available in a granular analysis, detecting unknown patterns in Radiomics publications, while highlighting potential developments to ensure knowledge dissemination in the field and its future real-life applications in the clinical practice.

CLINICAL RELEVANCE STATEMENT

This work aims to shed light on the state of the art in radiomics, which offers numerous tangible and intangible benefits, and to encourage its integration in the contemporary clinical practice for more precise imaging analysis.

KEY POINTS

• ML-based bibliometric analysis is fundamental to detect unknown pattern of data in Radiomics publications. • A raising interest in the field, the most relevant collaborations, keywords co-occurrence network, and trending topics have been investigated. • Some pitfalls still exist, including the scarce standardization and the relative lack of homogeneity across studies.

Incremental learning for an evolving stream of medical ultrasound images via counterfactual thinking

Despite the fact that traditional deep learning (DL) approaches provide promising accuracy and efficiency in medical ultrasound image analysis, they cannot replace the physician in making a diagnosis since the DL model is only appropriate in static application scenarios. Currently, most DL-based models are incapable of learning new tasks in the dynamic clinical environments due to the catastrophic forgetting of old tasks. To address the above problem, we propose an incremental classifier that is sequentially trained on evolving tasks for medical ultrasound images by counterfactual thinking. Specifically, the proposed model consists of a feature extractor and a classifier that can add new classes at any time during training. Toward a more discriminative model in the continual learning setting, a contrastive strategy is designed to leverage fine-grained information by generating a series of counterfactual regions. For model optimization, we design a multi-task loss made up of a knowledge distillation loss, a cross-entropy loss, and a contrasting loss. This objective jointly enjoys the merits of less forgetting, better accuracy, and fine-grained information utilization. A newly collected dataset with 52 medical ultrasound classification tasks is used to demonstrate the effectiveness of our method. The proposed approach achieves 76.59%, 11.67%, and 7.93% in terms of the average incremental accuracy, forgetting rate, and feature retention, respectively.

Development of a nomogram-based model combining intra- and peritumoral ultrasound radiomics with clinical features for differentiating benign from malignant in Breast Imaging Reporting and Data System category 3-5 nodules

Background

The differences in benign and malignant breast tumors are not only within the nodules but also involve changes in the surrounding tissues. Radiomics can reveal many details that are not discernible to the naked eye. This study aimed to distinguish between benign and malignant breast nodules using an ultrasound-based intra- and peritumoral radiomics model.

Methods

This study retrospectively collected the information from 379 patients with Breast Imaging Reporting and Data System (BI-RADS) category 3-5 nodules and clear pathological diagnosis of breast nodules screened by routine ultrasound examination in the Sixth People's Hospital Affiliated to Medical College of Shanghai Jiao Tong University from January 2017 to December 2022. The largest dimension of the lesion on the 2D ultrasound image was selected to outline the area of interest which was conformally and outwardly expanded automatically by 5 mm to extract intra- and peritumor radiomics features. The included cases were randomly divided into training sets and test sets in a ratio of 7:3. The optimal features of the included models were retained by statistical and machine learning methods of dimensionality reduction, and logistic regression was used as the classifier to build an intratumoral model and a combined intratumoral-peritumoral radiomics model, respectively; through single-factor and multifactor logistic regression, the optimal features that could predict benign and malignant breast tumors were screened. The clinical and imaging models were established by selecting independent risk factors as clinical and imaging features through univariate and multifactorial logistic regression.

Results

Among 379 BI-RADS category 3-5 breast nodules, there were 124 malignant nodules and 255 benign nodules; patients were aged 14 to 88 (46.22±15.51) years, and the age differences, radiomics score, and mass diameter between the training and test sets were not statistically significant (P>0.05). The intra- and peritumor radiomics model had an area under the curve (AUC) of 0.840 [95% confidence interval (CI): 0.766-0.914] in the test set. The model with intra- and peritumoral ultrasound radiomics features combined with clinical features had an AUC value of 0.960 (95% CI: 0.920-0.999).

Conclusions

The nomogram, developed using intratumoral and peritumoral radiomics features combined with clinical risk features, demonstrated superior performance in distinguishing between benign and malignant BI-RADS 3-5 lesions.

General Multiscenario Ultrasound Image Tumor Diagnosis Method Based on Unsupervised Domain Adaptation

OBJECTIVE

The utilization of computer-aided diagnosis (CAD) in breast ultrasound image classification has been limited by small sample sizes and domain shift. Current ultrasound classification methods perform inadequately when exposed to cross-domain scenarios, as they struggle with data sets from unobserved domains. In the medical field, there are situations in which all images must share the same networks as they capture the same symptom of the same participant, implying that they share identical structural content. Nevertheless, most domain adaptation methods are not suitable for medical images as they overlook the common features among the images.

METHODS

To overcome these challenges, we propose a novel diverse-domain 2-D feature selection network (FSN), which uses the similarities among medical images and extracts features with a reconstruction network with shared weights. Additionally, it penalizes the feature domain distance through two adversarial learning modules that align the feature space and select common features. Our experiments illustrate that the proposed method is robust and can be applied to ultrasound images of various diseases.

RESULTS

Compared with the latest domain adaptive methods, 2-D FSN markedly enhances the accuracy of classification of breast, thyroid and endoscopic ultrasound images, achieving accuracies of 82.4%, 96.4% and 89.7%, respectively. Furthermore, the model was evaluated on an unsupervised domain adaptation task using ultrasound images from multiple sources and achieved an average accuracy of 77.3% across widely varying domains.

CONCLUSION

In general, 2-D FSN improves the classification ability of the model on multidomain ultrasound data sets through the learning of common features and the combination of multimodule intelligence. The algorithm has good clinical guidance value.

Diagnostic value of the dual-modal imaging radiomics model for subpleural pulmonary lesions

PURPOSE

To investigate the clinical value of the radiomics model of grayscale ultrasound (GUS) and contrast-enhanced ultrasound (CEUS) to diagnosis subpleural pulmonary tuberculosis and nonpulmonary tuberculosis based on GUS and CEUS images.

METHODS

This study included 221 patients with 228 lesions diagnosed using the composite reference standard. The patients were randomly divided into training (n = 183) and test (n = 45) cohorts in an 8:2 ratio. The regions of interest of the GUS and CEUS images were manually segmented to extract the radiomic features. The GUS, CEUS and GUS+CEUS radiomics models were constructed via the multistep selection of highly correlated features. Receiver operating characteristic curves of the different models were plotted, and the area under the curve (AUC), accuracy, sensitivity, specificity, positive predictive value and negative predictive value (NPV) of the different models were compared.

RESULTS

Following Least Absolute Shrinkage and Selection Operator dimension reduction we selected 4, 9, and 11 features to construct the GUS, CEUS, and GUS+CEUS radiomics models, respectively. The AUC values of the three groups in the test cohort were 0.689, 0.748 and 0.779, respectively, and they did not differ significantly. In the test cohort, the GUS+CEUS radiomics model exhibited the highest AUC (0.779), accuracy (75.56%), and NPV (68.7%) of the three models.

CONCLUSIONS

The GUS+CEUS radiomics model possesses good clinical value in diagnosing pulmonary tuberculosis.

Assessment of Renal Fibrosis in Patients With Chronic Kidney Disease Using Shear Wave Elastography and Clinical Features: A Random Forest Approach

OBJECTIVE

Renal fibrosis is the common pathological hallmark of chronic kidney disease (CKD) progression. In this study, a random forest (RF) classifier based on 2-D shear wave elastography (SWE) and clinical features for the differential severity of renal fibrosis in patients with CKD is proposed.

METHODS

A total of 162 patients diagnosed with CKD who underwent 2-D SWE and renal biopsy were prospectively enrolled from April 2019 to December 2021 and then randomized into training (n = 114) and validation (n = 48) cohorts at a ratio of 7:3. The least absolute shrinkage and selection operator (LASSO) regression and recursive feature elimination for support vector machines (SVM-RFE) algorithm were employed to select renal fibrosis-related features from clinical information and elastosonographic findings. An RF model was subsequently constructed using the aforementioned informative parameters in the training cohort and evaluated in terms of discrimination, calibration and clinical utility in both cohorts.

RESULTS

The LASSO and SVM-RFE analyses revealed that age, sex, blood urea nitrogen, renal resistive index, hypertension and the 2D-SWE value were independent risk variables associated with renal fibrosis severity. The established RF model incorporating these six variables exhibited fine discrimination in both the derivation (area under the curve [AUC]: 0.84, 95% confidence interval [CI]: 0.76-0.91) and validation (AUC: 0.88, 95% CI: 0.77-0.98) cohorts. Moreover, the calibration curve revealed satisfactory predictive accuracy, and the decision curve analysis revealed a significant clinical net benefit.

CONCLUSION

The developed RF model, via a combination of the 2-D SWE value and clinical information, indicated satisfactory diagnostic performance and clinical practicality toward differentiating moderate-severe from mild renal fibrosis, which may provide critical insight into risk stratification for patients with CKD.

Prediction of Early Perihematomal Edema Expansion Based on Noncontrast Computed Tomography Radiomics and Machine Learning in Intracerebral Hemorrhage

OBJECTIVES

To investigate the predictive value of noncontrast computed tomography (NCCT) models based on radiomics features and machine learning for early perihematomal edema (PHE) expansion in patients with spontaneous intracerebral hemorrhage (ICH).

METHODS

We retrospectively reviewed NCCT data from 214 patients with spontaneous ICH. All radiomics features were extracted from volume of interest of hematomas on admission scans. A total of 8 machine learning methods were applied for constructing models in the training and the test set. Receiver operating characteristic analysis and the areas under the curve were used to evaluate the predictive value.

RESULTS

A total of 23 features were finally selected to establish models of early PHE expansion after feature screening. Patients were randomly assigned into training (n = 171) and test (n = 43) sets. The accuracy, sensitivity, and specificity in the test set were 72.1%, 90.0%, and 66.7% for the support vector machine model; 79.1%, 70.0%, and 84.4% for the k-nearest neighbor model; 88.4%, 90.0%, and 87.9% for the logistic regression model; 74.4%, 90.0%, and 69.7% for the extra tree model; 74.4%, 90.0%, and 69.7% for the extreme gradient boosting model; 83.7%, 100%, and 78.8% for the multilayer perceptron (MLP) model; 72.1%, 100%, and 65.6% for the light gradient boosting machine model; and 60.5%, 90.0%, and 53.1% for the random forest model, respectively.

CONCLUSIONS

The MLP model seemed to be the best model for prediction of PHE expansion in patients with ICH. NCCT models based on radiomics features and machine learning could predict early PHE expansion and improve the discrimination of identify spontaneous intracerebral hemorrhage patients at risk of early PHE expansion.

What benefit can be obtained from magnetic resonance imaging diagnosis with artificial intelligence in prostate cancer compared with clinical assessments?

The present study aimed to explore the potential of artificial intelligence (AI) methodology based on magnetic resonance (MR) images to aid in the management of prostate cancer (PCa). To this end, we reviewed and summarized the studies comparing the diagnostic and predictive performance for PCa between AI and common clinical assessment methods based on MR images and/or clinical characteristics, thereby investigating whether AI methods are generally superior to common clinical assessment methods for the diagnosis and prediction fields of PCa. First, we found that, in the included studies of the present study, AI methods were generally equal to or better than the clinical assessment methods for the risk assessment of PCa, such as risk stratification of prostate lesions and the prediction of therapeutic outcomes or PCa progression. In particular, for the diagnosis of clinically significant PCa, the AI methods achieved a higher summary receiver operator characteristic curve (SROC-AUC) than that of the clinical assessment methods (0.87 vs. 0.82). For the prediction of adverse pathology, the AI methods also achieved a higher SROC-AUC than that of the clinical assessment methods (0.86 vs. 0.75). Second, as revealed by the radiomics quality score (RQS), the studies included in the present study presented a relatively high total average RQS of 15.2 (11.0-20.0). Further, the scores of the individual RQS elements implied that the AI models in these studies were constructed with relatively perfect and standard radiomics processes, but the exact generalizability and clinical practicality of the AI models should be further validated using higher levels of evidence, such as prospective studies and open-testing datasets.

Radiomics Analysis of Breast Lesions in Combination with Coronal Plane of ABVS and Strain Elastography

Background

Breast cancer is the most common tumor globally. Automated Breast Volume Scanner (ABVS) and strain elastography (SE) can provide more useful breast information. The use of radiomics combined with ABVS and SE images to predict breast cancer has become a new focus. Therefore, this study developed and validated a radiomics analysis of breast lesions in combination with coronal plane of ABVS and SE to improve the differential diagnosis of benign and malignant breast diseases.

Patients and Methods

620 pathologically confirmed breast lesions from January 2017 to August 2021 were retrospectively analyzed and randomly divided into a training set (n=434) and a validation set (n=186). Radiomic features of the lesions were extracted from ABVS, B-ultrasound, and strain elastography (SE) images, respectively. These were then filtered by Gradient Boosted Decision Tree (GBDT) and multiple logistic regression. The ABVS model is based on coronal plane features for the breast, B+SE model is based on features of B-ultrasound and SE, and the multimodal model is based on features of three examinations. The evaluation of the predicted performance of the three models used the receiver operating characteristic (ROC) and decision curve analysis (DCA).

Results

The area under the curve, accuracy, specificity, and sensitivity of the multimodal model in the training set are 0.975 (95% CI:0.959-0.991),93.78%, 92.02%, and 96.49%, respectively, and 0.946 (95% CI:0.913 -0.978), 87.63%, 83.93%, and 93.24% in the validation set, respectively. The multimodal model outperformed the ABVS model and B+SE model in both the training (P < 0.001, P = 0.002, respectively) and validation sets (P < 0.001, P = 0.034, respectively).

Conclusion

Radiomics from the coronal plane of the breast lesion provide valuable information for identification. A multimodal model combination with radiomics from ABVS, B-ultrasound, and SE could improve the diagnostic efficacy of breast masses.

Predicting Extrathyroidal Extension in Papillary Thyroid Carcinoma Using a Clinical-Radiomics Nomogram Based on B-Mode and Contrast-Enhanced Ultrasound

Papillary thyroid carcinoma (PTC) is the most common pathological type of thyroid cancer. PTC patients with extrathyroidal extension (ETE) are associated with poor prognoses. The preoperative accurate prediction of ETE is crucial for helping the surgeon decide on the surgical plan. This study aimed to establish a novel clinical-radiomics nomogram based on B-mode ultrasound (BMUS) and contrast-enhanced ultrasound (CEUS) for the prediction of ETE in PTC. A total of 216 patients with PTC between January 2018 and June 2020 were collected and divided into the training set (n = 152) and the validation set (n = 64). The least absolute shrinkage and selection operator (LASSO) algorithm was applied for radiomics feature selection. Univariate analysis was performed to find clinical risk factors for predicting ETE. The BMUS Radscore, CEUS Radscore, clinical model, and clinical-radiomics model were established using multivariate backward stepwise logistic regression (LR) based on BMUS radiomics features, CEUS radiomics features, clinical risk factors, and the combination of those features, respectively. The diagnostic efficacy of the models was assessed using receiver operating characteristic (ROC) curves and the DeLong test. The model with the best performance was then selected to develop a nomogram. The results show that the clinical-radiomics model, which is constructed by age, CEUS-reported ETE, BMUS Radscore, and CEUS Radscore, showed the best diagnostic efficiency in both the training set (AUC = 0.843) and validation set (AUC = 0.792). Moreover, a clinical-radiomics nomogram was established for easier clinical practices. The Hosmer-Lemeshow test and the calibration curves demonstrated satisfactory calibration. The decision curve analysis (DCA) showed that the clinical-radiomics nomogram had substantial clinical benefits. The clinical-radiomics nomogram constructed from the dual-modal ultrasound can be exploited as a promising tool for the pre-operative prediction of ETE in PTC.

Multiregional radiomic model for breast cancer diagnosis: value of ultrasound-based peritumoral and parenchymal radiomics

Background

Breast cancer consists not only of neoplastic cells but also of significant changes in the surrounding and parenchymal stroma, which can be reflected in radiomics. This study aimed to perform breast lesion classification through an ultrasound-based multiregional (intratumoral, peritumoral, and parenchymal) radiomic model.

Methods

We retrospectively reviewed ultrasound images of breast lesions from institution #1 (n=485) and institution #2 (n=106). Radiomic features were extracted from different regions (intratumoral, peritumoral, and ipsilateral breast parenchymal) and selected to train the random forest classifier with the training cohort (n=339, a subset of the institution #1 dataset). Then, the intratumoral, peritumoral, and parenchymal, intratumoral & peritumoral (In&Peri), intratumoral & parenchymal (In&P), and intratumoral & peritumoral & parenchymal (In&Peri&P) models were developed and validated on the internal (n=146, another subset of institution 1) and external (n=106, institution #2 dataset) test cohorts. Discrimination was evaluated using the area under the curve (AUC). Calibration curve and Hosmer-Lemeshow test assessed calibration. Integrated discrimination improvement (IDI) was used to assess performance improvement.

Results

The performance of the In&Peri (AUC values 0.892 and 0.866), In&P (0.866 and 0.863), and In&Peri&P (0.929 and 0.911) models was significantly better than that of the intratumoral model (0.849 and 0.838) in the internal and external test cohorts (IDI test, all P<0.05). The intratumoral, In&Peri and In&Peri&P models showed good calibration (Hosmer-Lemeshow test, all P>0.05). The multiregional (In&Peri&P) model had the highest discrimination among the 6 radiomic models in the test cohorts, respectively.

Conclusions

The multiregional model combining radiomic information of intratumoral, peritumoral, and ipsilateral parenchymal regions yielded better performance than the intratumoral model in distinguishing malignant breast lesions from benign lesions.

Ultrasound-based deep learning in the establishment of a breast lesion risk stratification system: a multicenter study

OBJECTIVES

To establish a breast lesion risk stratification system using ultrasound images to predict breast malignancy and assess Breast Imaging Reporting and Data System (BI-RADS) categories simultaneously.

METHODS

This multicenter study prospectively collected a dataset of ultrasound images for 5012 patients at thirty-two hospitals from December 2018 to December 2020. A deep learning (DL) model was developed to conduct binary categorization (benign and malignant) and BI-RADS categories (2, 3, 4a, 4b, 4c, and 5) simultaneously. The training set of 4212 patients and the internal test set of 416 patients were from thirty hospitals. The remaining two hospitals with 384 patients were used as an external test set. Three experienced radiologists performed a reader study on 324 patients randomly selected from the test sets. We compared the performance of the DL model with that of three radiologists and the consensus of the three radiologists.

RESULTS

In the external test set, the DL model achieved areas under the receiver operating characteristic curve (AUCs) of 0.980 and 0.945 for the binary categorization and six-way categorizations, respectively. In the reader study set, the DL BI-RADS categories achieved a similar AUC (0.901 vs. 0.933, p = 0.0632), sensitivity (90.98% vs. 95.90%, p = 0.1094), and accuracy (83.33% vs. 79.01%, p = 0.0541), but higher specificity (78.71% vs. 68.81%, p = 0.0012) than those of the consensus of the three radiologists.

CONCLUSIONS

The DL model performed well in distinguishing benign from malignant breast lesions and yielded outcomes similar to experienced radiologists. This indicates the potential applicability of the DL model in clinical diagnosis.

KEY POINTS

• The DL model can achieve binary categorization for benign and malignant breast lesions and six-way BI-RADS categorizations for categories 2, 3, 4a, 4b, 4c, and 5, simultaneously. • The DL model showed acceptable agreement with radiologists for the classification of breast lesions. • The DL model performed well in distinguishing benign from malignant breast lesions and had promise in helping reduce unnecessary biopsies of BI-RADS 4a lesions.

Deep learn-based computer-assisted transthoracic echocardiography: approach to the diagnosis of cardiac amyloidosis

Myocardial amyloidosis (CA) differs from other etiological pathologies of left ventricular hypertrophy in that transthoracic echocardiography is challenging to assess the texture features based on human visual observation. There are few studies on myocardial texture based on echocardiography. Therefore, this paper proposes an adaptive machine learning method based on ultrasonic image texture features to identify CA. In this retrospective study, a total of 289 participants (50 cases of myocardial amyloidosis; Hypertrophic cardiomyopathy: 70 cases; Uremic cardiomyopathy: 92 cases; Hypertensive heart disease: 77 cases). We extracted the myocardial ultrasonic imaging features of these patients and screened the features, and four models of random forest (RF), support vector machine (SVM), logistic regression (LR) and gradient decision-making lifting tree (GBDT) were established to distinguish myocardial amyloidosis from other diseases. Finally, the diagnostic efficiency of the model was evaluated and compared with the traditional ultrasonic diagnostic methods. In the overall population, the four machine learning models we established could effectively distinguish CA from nonCA diseases, AUC (RF 0.77, SVM 0.81, LR 0.81, GBDT 0.71). The LR model had the best diagnostic efficiency with recall, F1-score, sensitivity and specificity of 0.21, 0.34, 0.21 and 1.0, respectively. Slightly better than the traditional ultrasonic diagnosis model. In further subgroup analysis, the myocardial amyloidosis group was compared one-by-one with the patients with hypertrophic cardiomyopathy, uremic cardiomyopathy, and hypertensive heart disease groups, and the same method was used for feature extraction and data modeling. The diagnostic efficiency of the model was further improved. Notably, in identifying of the CA group and HHD group, AUC values reached more than 0.92, accuracy reached more than 0.87, sensitivity equal to or greater than 0.81, specificity 0.91, and F1 score higher than 0.84. This novel method based on echocardiography combined with machine learning may have the potential to be used in the diagnosis of CA.

Prediction of the Malignancy of a Breast Lesion Detected on Breast Ultrasound: Radiomics Applied to Clinical Practice

The study aimed to evaluate the performance of radiomics features and one ultrasound CAD (computer-aided diagnosis) in the prediction of the malignancy of a breast lesion detected with ultrasound and to develop a nomogram incorporating radiomic score and available information on CAD performance, conventional Breast Imaging Reporting and Data System evaluation (BI-RADS), and clinical information. Data on 365 breast lesions referred for breast US with subsequent histologic analysis between January 2020 and March 2022 were retrospectively collected. Patients were randomly divided into a training group (n = 255) and a validation test group (n = 110). A radiomics score was generated from the US image. The CAD was performed in a subgroup of 209 cases. The radiomics score included seven radiomics features selected with the LASSO logistic regression model. The multivariable logistic model incorporating CAD performance, BI-RADS evaluation, clinical information, and radiomic score as covariates showed promising results in the prediction of the malignancy of breast lesions: Area under the receiver operating characteristic curve, [AUC]: 0.914; 95% Confidence Interval, [CI]: 0.876-0.951. A nomogram was developed based on these results for possible future applications in clinical practice.

Radiogenomic analysis of prediction HER2 status in breast cancer by linking ultrasound radiomic feature module with biological functions

BACKGROUND

Human epidermal growth factor receptor 2 (HER2) overexpressed associated with poor prognosis in breast cancer and HER2 has been defined as a therapeutic target for breast cancer treatment. We aimed to explore the molecular biological information in ultrasound radiomic features (URFs) of HER2-positive breast cancer using radiogenomic analysis. Moreover, a radiomics model was developed to predict the status of HER2 in breast cancer.

METHODS

This retrospective study included 489 patients who were diagnosed with breast cancer. URFs were extracted from a radiomics analysis set using PyRadiomics. The correlations between differential URFs and HER2-related genes were calculated using Pearson correlation analysis. Functional enrichment of the identified URFs-correlated HER2 positive-specific genes was performed. Lastly, the radiomics model was developed based on the URF-module mined from auxiliary differential URFs to assess the HER2 status of breast cancer.

RESULTS

Eight differential URFs (p < 0.05) were identified among the 86 URFs extracted by Pyradiomics. 25 genes that were found to be the most closely associated with URFs. Then, the relevant biological functions of each differential URF were obtained through functional enrichment analysis. Among them, Zone Entropy is related to immune cell activity, which regulate the generation of calcification in breast cancer. The radiomics model based on the Logistic classifier and URF-module showed good discriminative ability (AUC = 0.80, 95% CI).

CONCLUSION

We searched for the URFs of HER2-positive breast cancer, and explored the underlying genes and biological functions of these URFs. Furthermore, the radiomics model based on the Logistic classifier and URF-module relatively accurately predicted the HER2 status in breast cancer.

Artificial Intelligence in Breast Ultrasound: From Diagnosis to Prognosis-A Rapid Review

BACKGROUND

Ultrasound (US) is a fundamental diagnostic tool in breast imaging. However, US remains an operator-dependent examination. Research into and the application of artificial intelligence (AI) in breast US are increasing. The aim of this rapid review was to assess the current development of US-based artificial intelligence in the field of breast cancer.

METHODS

Two investigators with experience in medical research performed literature searching and data extraction on PubMed. The studies included in this rapid review evaluated the role of artificial intelligence concerning BC diagnosis, prognosis, molecular subtypes of breast cancer, axillary lymph node status, and the response to neoadjuvant chemotherapy. The mean values of sensitivity, specificity, and AUC were calculated for the main study categories with a meta-analytical approach.

RESULTS

A total of 58 main studies, all published after 2017, were included. Only 9/58 studies were prospective (15.5%); 13/58 studies (22.4%) used an ML approach. The vast majority (77.6%) used DL systems. Most studies were conducted for the diagnosis or classification of BC (55.1%). At present, all the included studies showed that AI has excellent performance in breast cancer diagnosis, prognosis, and treatment strategy.

CONCLUSIONS

US-based AI has great potential and research value in the field of breast cancer diagnosis, treatment, and prognosis. More prospective and multicenter studies are needed to assess the potential impact of AI in breast ultrasound.

Development and Validation of an Ultrasound-Based Radiomics Nomogram for Identifying HER2 Status in Patients with Breast Carcinoma

(1) Objective: To evaluate the performance of ultrasound-based radiomics in the preoperative prediction of human epidermal growth factor receptor 2-positive (HER2+) and HER2- breast carcinoma. (2) Methods: Ultrasound images from 309 patients (86 HER2+ cases and 223 HER2- cases) were retrospectively analyzed, of which 216 patients belonged to the training set and 93 patients assigned to the time-independent validation set. The region of interest of the tumors was delineated, and the radiomics features were extracted. Radiomics features underwent dimensionality reduction analyses using the intra-class correlation coefficient (ICC), Mann-Whitney U test, and the least absolute shrinkage and selection operator (LASSO) algorithm. The radiomics score (Rad-score) for each patient was calculated through a linear combination of the nonzero coefficient features. The support vector machine (SVM), K nearest neighbors (KNN), logistic regression (LR), decision tree (DT), random forest (RF), naive Bayes (NB) and XGBoost (XGB) machine learning classifiers were trained to establish prediction models based on the Rad-score. A clinical model based on significant clinical features was also established. In addition, the logistic regression method was used to integrate Rad-score and clinical features to generate the nomogram model. The leave-one-out cross validation (LOOCV) method was used to validate the reliability and stability of the model. (3) Results: Among the seven classifier models, the LR achieved the best performance in the validation set, with an area under the receiver operating characteristic curve (AUC) of 0.786, and was obtained as the Rad-score model, while the RF performed the worst. Tumor size showed a statistical difference between the HER2+ and HER2- groups (p = 0.028). The nomogram model had a slightly higher AUC than the Rad-score model (AUC, 0.788 vs. 0.786), but no statistical difference (Delong test, p = 0.919). The LOOCV method yielded a high median AUC of 0.790 in the validation set. (4) Conclusion: The Rad-score model performs best among the seven classifiers. The nomogram model based on Rad-score and tumor size has slightly better predictive performance than the Rad-score model, and it has the potential to be utilized as a routine modality for preoperatively determining HER2 status in BC patients non-invasively.

Multi-modality radiomics nomogram based on DCE-MRI and ultrasound images for benign and malignant breast lesion classification

OBJECTIVE

To develop a multi-modality radiomics nomogram based on DCE-MRI, B-mode ultrasound (BMUS) and strain elastography (SE) images for classifying benign and malignant breast lesions.

MATERIAL AND METHODS

In this retrospective study, 345 breast lesions from 305 patients who underwent DCE-MRI, BMUS and SE examinations were randomly divided into training (n = 241) and testing (n = 104) datasets. Radiomics features were extracted from manually contoured images. The inter-class correlation coefficient (ICC), Mann-Whitney U test and the least absolute shrinkage and selection operator (LASSO) regression were applied for feature selection and radiomics signature building. Multivariable logistic regression was used to develop a radiomics nomogram incorporating radiomics signature and clinical factors. The performance of the radiomics nomogram was evaluated by its discrimination, calibration, and clinical usefulness and was compared with BI-RADS classification evaluated by a senior breast radiologist.

RESULTS

The All-Combination radiomics signature derived from the combination of DCE-MRI, BMUS and SE images showed better diagnostic performance than signatures derived from single modality alone, with area under the curves (AUCs) of 0.953 and 0.941 in training and testing datasets, respectively. The multi-modality radiomics nomogram incorporating the All-Combination radiomics signature and age showed excellent discrimination with the highest AUCs of 0.964 and 0.951 in two datasets, respectively, which outperformed all single modality radiomics signatures and BI-RADS classification. Furthermore, the specificity of radiomics nomogram was significantly higher than BI-RADS classification (both p < 0.04) with the same sensitivity in both datasets.

CONCLUSION

The proposed multi-modality radiomics nomogram based on DCE-MRI and ultrasound images has the potential to serve as a non-invasive tool for classifying benign and malignant breast lesions and reduce unnecessary biopsy.

Development and validation of an ultrasound-based radiomics nomogram for predicting the luminal from non-luminal type in patients with breast carcinoma

INTRODUCTION

The molecular subtype plays a significant role in breast carcinoma (BC), which is the main indicator to guide treatment and is closely associated with prognosis. The aim of this study was to investigate the feasibility and efficacy of an ultrasound-based radiomics nomogram in preoperatively discriminating the luminal from non-luminal type in patients with BC.

METHODS

A total of 264 BC patients who underwent routine ultrasound examination were enrolled in this study, of which 184 patients belonged to the training set and 80 patients to the test set. Breast tumors were delineated manually on the ultrasound images and then radiomics features were extracted. In the training set, the T test and least absolute shrinkage and selection operator (LASSO) were used for selecting features, and the radiomics score (Rad-score) for each patient was calculated. Based on the clinical risk features, Rad-score, and combined clinical risk features and Rad-score, three models were established, respectively. The performances of the models were validated with receiver operator characteristic (ROC) curve and decision curve analysis.

RESULTS

In all, 788 radiomics features per case were obtained from the ultrasound images. Through radiomics feature selection, 11 features were selected to constitute the Rad-score. The area under the ROC curve (AUC) of the Rad-score for predicting the luminal type was 0.828 in the training set and 0.786 in the test set. The nomogram comprising the Rad-score and US-reported tumor size showed AUCs of the training and test sets were 0.832 and 0.767, respectively, which were significantly higher than the AUCs of the clinical model in the training and test sets (0.691 and 0.526, respectively). However, there was no significant difference in predictive performance between the Rad-score and nomogram.

CONCLUSION

Both the Rad-score and nomogram can be applied as useful, noninvasive tools for preoperatively discriminating the luminal from non-luminal type in patients with BC. Furthermore, this study might provide a novel technique to evaluate molecular subtypes of BC.

Extending artificial intelligence research in the clinical domain: a theoretical perspective

Academic research to the utilization of artificial intelligence (AI) has been proliferated over the past few years. While AI and its subsets are continuously evolving in the fields of marketing, social media and finance, its application in the daily practice of clinical care is insufficiently explored. In this systematic review, we aim to landscape various application areas of clinical care in terms of the utilization of machine learning to improve patient care. Through designing a specific smart literature review approach, we give a new insight into existing literature identified with AI technologies in the clinical domain. Our review approach focuses on strategies, algorithms, applications, results, qualities, and implications using the Latent Dirichlet Allocation topic modeling. A total of 305 unique articles were reviewed, with 115 articles selected using Latent Dirichlet Allocation topic modeling, meeting our inclusion criteria. The primary result of this approach incorporates a proposition for future research direction, abilities, and influence of AI technologies and displays the areas of disease management in clinics. This research concludes with disease administrative ramifications, limitations, and directions for future research.

Integration of ultrasound radiomics features and clinical factors: A nomogram model for identifying the Ki-67 status in patients with breast carcinoma

Objective

The aim of this study was to develop and validate an ultrasound-based radiomics nomogram model by integrating the clinical risk factors and radiomics score (Rad-Score) to predict the Ki-67 status in patients with breast carcinoma.

Methods

Ultrasound images of 284 patients (196 high Ki-67 expression and 88 low Ki-67 expression) were retrospectively analyzed, of which 198 patients belonged to the training set and 86 patients to the test set. The region of interest of tumor was delineated, and the radiomics features were extracted. Radiomics features underwent dimensionality reduction analysis by using the independent sample t test and least absolute shrinkage and selection operator (LASSO) algorithm. The support vector machine (SVM), logistic regression (LR), decision tree (DT), random forest (RF), naive Bayes (NB) and XGBoost (XGB) machine learning classifiers were trained to establish prediction model based on the selected features. The classifier with the highest AUC value was selected to convert the output of the results into the Rad-Score and was regarded as Rad-Score model. In addition, the logistic regression method was used to integrate Rad-Score and clinical risk factors to generate the nomogram model. The leave group out cross-validation (LGOCV) method was performed 200 times to verify the reliability and stability of the nomogram model.

Results

Six classifier models were established based on the 15 non-zero coefficient features. Among them, the LR classifier achieved the best performance in the test set, with the area under the receiver operating characteristic curve (AUC) value of 0.786, and was obtained as the Rad-Score model, while the XGB performed the worst (AUC, 0.615). In multivariate analysis, independent risk factor for high Ki-67 status was age (odds ratio [OR] = 0.97, p = 0.04). The nomogram model based on the age and Rad-Score had a slightly higher AUC than that of Rad-Score model (AUC, 0.808 vs. 0.798) in the test set, but no statistical difference (p = 0.144, DeLong test). The LGOCV yielded a median AUC of 0.793 in the test set.

Conclusions

This study proposed a convenient, clinically useful ultrasound radiomics nomogram model that can be used for the preoperative individualized prediction of the Ki-67 status in patients with BC.

A deep learning-based diagnostic pattern for ultrasound breast imaging: can it reduce unnecessary biopsy?

Background

Early studies have demonstrated the potential of deep learning in bringing revolutionary changes in medical analysis. However, it is unknown which deep learning based diagnostic pattern is more effective for differentiating malignant and benign breast lesions (BLs) and can assist radiologists to reduce unnecessary biopsies.

Methods

A total of 506 malignant BLs and 557 benign BLs were enrolled in this study after excluding incomplete ultrasound images. 396 malignant BLs and 447 benign BLs were included in the training cohort while 110 malignant and 110 benign BLs were included in the validation cohort. All BLs in the training and validation cohort were biopsy-proven. The most common convolutional neural networks (VGG-16 and VGG-19) were applied to identify malignant and benign BLs using grey-scale ultrasound images. Two radiologists determined the malignant (suggestion for biopsy) and benign (suggestion for follow-up) BLs with a 2-step reading session. The first step was based on conventional ultrasound (US) images alone to make a biopsy or follow-up decision. The second step was to take deep learning results into account for the decision adjustment. If a deep learning result of a first-classified benign BL was above the cut-off value, then it was re-classified as malignant.

Results

In terms of area under the curve (AUC), the VGG-19 model yielded the best diagnostic performance in both training [0.939, 95% confidence interval (CI): 0.924-0.954] and testing dataset (0.959, 95% CI: 0.937-0.982). With the aid of deep learning models, the AUC of radiologists improved from 0.805 (95% CI: 0.744-0.865) to 0.827 (95% CI: 0.771-0.875, VGG-16) and 0.914 (95% CI: 0.871-0.957, VGG-19). The unnecessary biopsies decreased from 10.0% (11/110) to 8.2% (9/110) (assisted by VGG-16) and 0.9% (1/110) (assisted by VGG-19).

Conclusions

The application of deep learning patterns in breast US may improve the diagnostic performance of radiologists by offering a second opinion. And thus, the assist of deep learning algorithm can considerably reduce the unnecessary biopsy rate in the clinical management of breast lesions.

Machine learning based on automated breast volume scanner (ABVS) radiomics for differential diagnosis of benign and malignant BI‐RADS 4 lesions

BI‐RADS category 4 represents possibly malignant lesions and biopsy is recommended to distinguish benign and malignant. However, studies revealed that up to 67%–78% of BI‐RADS 4 lesions proved to be benign, but received unnecessary biopsies, which may cause unnecessary anxiety and discomfort to patients and increase the burden on the healthcare system. In this prospective study, machine learning (ML) based on the emerging breast ultrasound technology‐automated breast volume scanner (ABVS) was constructed to distinguish benign and malignant BI‐RADS 4 lesions and compared with different experienced radiologists. A total of 223 pathologically confirmed BI‐RADS 4 lesions were recruited and divided into training and testing cohorts. Radiomics features were extracted from axial, sagittal, and coronal ABVS images for each lesion. Seven feature selection methods and 13 ML algorithms were used to construct different ML pipelines, of which the DNN‐RFE (combination of recursive feature elimination and deep neural networks) had the best performance in both training and testing cohorts. The AUC value of the DNN‐RFE was significantly higher than less experienced radiologist at Delong's test (0.954 vs. 0.776, p = 0.004). Additionally, the accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the DNN‐RFE were 88.9%, 83.3%, 95.2%, 83.3%, and 95.2%, which also significantly better than less experienced radiologist at McNemar's test (p = 0.043). Therefore, ML based on ABVS radiomics may be a potential method to non‐invasively distinguish benign and malignant BI‐RADS 4 lesions.

Ultrasound radiomics in personalized breast management: Current status and future prospects

Breast cancer is the most common cancer in women worldwide. Providing accurate and efficient diagnosis, risk stratification and timely adjustment of treatment strategies are essential steps in achieving precision medicine before, during and after treatment. Radiomics provides image information that cannot be recognized by the naked eye through deep mining of medical images. Several studies have shown that radiomics, as a second reader of medical images, can assist physicians not only in the detection and diagnosis of breast lesions but also in the assessment of risk stratification and prediction of treatment response. Recently, more and more studies have focused on the application of ultrasound radiomics in breast management. We summarized recent research advances in ultrasound radiomics for the diagnosis of benign and malignant breast lesions, prediction of molecular subtype, assessment of lymph node status, prediction of neoadjuvant chemotherapy response, and prediction of survival. In addition, we discuss the current challenges and future prospects of ultrasound radiomics.

Efficient Prediction of Ki-67 Proliferation Index in Meningiomas on MRI: From Traditional Radiological Findings to a Machine Learning Approach

Background/aim This study aimed to explore the value of radiological and radiomic features retrieved from magnetic resonance imaging in the prediction of a Ki-67 proliferative index in meningioma patients using a machine learning model. Methods This multicenter, retrospective study included 371 patients collected from two centers. The Ki-67 expression was classified into low-expressed and high-expressed groups with a threshold of 5%. Clinical features and radiological features were collected and analyzed by using univariate and multivariate statistical analyses. Radiomic features were extracted from contrast-enhanced images, followed by three independent feature selections. Six predictive models were constructed with different combinations of features by using linear discriminant analysis (LDA) classifier. Results The multivariate analysis suggested that the presence of intratumoral necrosis (p = 0.032) and maximum diameter (p < 0.001) were independently correlated with a high Ki-67 status. The predictive models showed good performance with AUC of 0.837, accuracy of 0.810, sensitivity of 0.857, and specificity of 0.771 in the internal test and with AUC of 0.700, accuracy of 0.557, sensitivity of 0.314, and specificity of 0.885 in the external test. Conclusion The results of this study suggest that the predictive model can efficiently predict the Ki-67 index of meningioma patients to facilitate the therapeutic management.

Recent Advances in Machine Learning Applied to Ultrasound Imaging

Machine learning (ML) methods are pervading an increasing number of fields of application because of their capacity to effectively solve a wide variety of challenging problems. The employment of ML techniques in ultrasound imaging applications started several years ago but the scientific interest in this issue has increased exponentially in the last few years. The present work reviews the most recent (2019 onwards) implementations of machine learning techniques for two of the most popular ultrasound imaging fields, medical diagnostics and non-destructive evaluation. The former, which covers the major part of the review, was analyzed by classifying studies according to the human organ investigated and the methodology (e.g., detection, segmentation, and/or classification) adopted, while for the latter, some solutions to the detection/classification of material defects or particular patterns are reported. Finally, the main merits of machine learning that emerged from the study analysis are summarized and discussed.

An MRI-Based Clinical-Perfusion Model Predicts Pathological Subtypes of Prevascular Mediastinal Tumors

This study aimed to build machine learning prediction models for predicting pathological subtypes of prevascular mediastinal tumors (PMTs). The candidate predictors were clinical variables and dynamic contrast–enhanced MRI (DCE-MRI)–derived perfusion parameters. The clinical data and preoperative DCE–MRI images of 62 PMT patients, including 17 patients with lymphoma, 31 with thymoma, and 14 with thymic carcinoma, were retrospectively analyzed. Six perfusion parameters were calculated as candidate predictors. Univariate receiver-operating-characteristic curve analysis was performed to evaluate the performance of the prediction models. A predictive model was built based on multi-class classification, which detected lymphoma, thymoma, and thymic carcinoma with sensitivity of 52.9%, 74.2%, and 92.8%, respectively. In addition, two predictive models were built based on binary classification for distinguishing Hodgkin from non-Hodgkin lymphoma and for distinguishing invasive from noninvasive thymoma, with sensitivity of 75% and 71.4%, respectively. In addition to two perfusion parameters (efflux rate constant from tissue extravascular extracellular space into the blood plasma, and extravascular extracellular space volume per unit volume of tissue), age and tumor volume were also essential parameters for predicting PMT subtypes. In conclusion, our machine learning–based predictive model, constructed with clinical data and perfusion parameters, may represent a useful tool for differential diagnosis of PMT subtypes.

Association of machine learning ultrasound radiomics and disease outcome in triple negative breast cancer

Triple negative breast cancer (TNBC) is a breast cancer subtype with unfavorable prognosis. We aimed to establish a machine learning-based ultrasound radiomics model to predict disease-free survival (DFS) in TNBC. Invasive TNBC>T1b between January 2009 and June 2018 with preoperative ultrasound were enrolled and assigned to training and independent test cohort. Radiomics and clinicopathological features related with DFS were selected by univariate and multivariate regression analysis. Training cohort of combined features was resampled with SMOTEENN to balance distribution and put into classifiers. Areas Under Curves (AUCs) of models were compared by DeLong's test. 562 women were included with 68 DFS events observed. Twenty prognostic radiomics features were extracted. Machine learning model by Naïve Bayes combining radiomics, clinicopathological features, and SMOTEENN had an AUC of 0.86 (95% CI 0.84-0.88), with sensitivity of 74.7% and specificity of 80.1% in training cohort. In independent test cohort, this three-combination model delivered an AUC of 0.90 (95% CI 0.83-0.95), higher than models based on radiomics (AUC=0.69, P=0.016) or radiomics + SMOTEENN (AUC=0.73, P=0.019). Integrating machine learning radiomics model based on ultrasound and clinicopathological features can predict DFS events for TNBC patients.

OUP accepted manuscript

The recent, rapid advances in immuno-oncology have revolutionized cancer treatment and spurred further research into tumor biology. Yet, cancer patients respond variably to immunotherapy despite mounting evidence to support its efficacy. Current methods for predicting immunotherapy response are unreliable, as these tests cannot fully account for tumor heterogeneity and microenvironment. An improved method for predicting response to immunotherapy is needed. Recent studies have proposed radiomics—the process of converting medical images into quantitative data (features) that can be processed using machine learning algorithms to identify complex patterns and trends—for predicting response to immunotherapy. Because patients undergo numerous imaging procedures throughout the course of the disease, there exists a wealth of radiological imaging data available for training radiomics models. And because radiomic features reflect cancer biology, such as tumor heterogeneity and microenvironment, these models have enormous potential to predict immunotherapy response more accurately than current methods. Models trained on preexisting biomarkers and/or clinical outcomes have demonstrated potential to improve patient stratification and treatment outcomes. In this review, we discuss current applications of radiomics in oncology, followed by a discussion on recent studies that use radiomics to predict immunotherapy response and toxicity.