Objective: To develop and validate machine learning (ML) models for diagnosing salivary gland adenoid cystic carcinoma (ACC) in the salivary glands based on clinical and ultrasound features. Methods: A total of 365 patients with ACC or non-ACC of the salivary glands treated at two centers were enrolled in training cohort, internal and external validation cohorts. Synthetic minority oversampling technique was used to address the class imbalance. The least absolute shrinkage and selection operator (LASSO) regression identified optimal features, which were subsequently utilized to construct predictive models employing five ML algorithms. The performance of the models was evaluated across a comprehensive array of learning metrics, prominently the area under the receiver operating characteristic curve (AUC). Results: Through LASSO regression analysis, six key features—sex, pain symptoms, number, cystic areas, rat tail sign, and polar vessel—were identified and subsequently utilized to develop five ML models. Among these models, the support vector machine (SVM) model demonstrated superior performance, achieving the highest AUCs of 0.899 and 0.913, accuracy of 90.54% and 91.53%, and F1 scores of 0.774 and 0.783 in both the internal and external validation cohorts, respectively. Decision curve analysis further revealed that the SVM model offered enhanced clinical utility compared to the other models. Conclusions: The ML model based on clinical and US features provide an accurate and noninvasive method for distinguishing ACC from non-ACC. Critical relevance statement: This machine learning model, constructed based on clinical and ultrasound characteristics, serves as a valuable tool for the identification of salivary gland adenoid cystic carcinoma. Key Points: Rat tail sign and polar vessel on US predict adenoid cystic carcinoma (ACC). Machine learning models based on clinical and US features can identify ACC. The support vector machine model performed robustly and accurately. Graphical
Preoperative identification of different stromal subtypes of pleomorphic adenoma (PA) of the salivary gland is crucial for making treatment decisions. We aimed to developed and validated a model based on histogram analysis (HA) of ultrasound (US) images for predicting tumor stroma ratio (TSR) in SPA. A total of 219 PA patients were divided into low-TSR (stroma-low) and high-TSR (stroma-high) groups, and enrolled in a training cohort (n =151) and a validation cohort (n =68). The least absolute shrinkage and selection operator regression algorithm was used to screen the most optimal clinical, US, and HA features. The selected features were entered into multivariable logistic regression analyses for further selection of independent predictors. Different models, including the nomogram model, the clinic-US (Clin + US) model, and the HA model, were built based on independent predictors using logistic regression. The performance levels of the models were evaluated and validated on the training and validation cohorts. Lesion size, shape, cystic areas, vascularity, HA_mean, and HA_skewness were identified as independent predictors for constructing the nomogram model. The nomogram model incorporating the clinical, US, and HA features achieved AUCs of 0.839 and 0.852 in the training and validation cohorts, respectively, demonstrating good predictive performance and calibration. Decision curve analysis and clinical impact curves further confirmed its clinical usefulness. The nomogram model we developed offers a practical tool for preoperative TSR prediction in PA, potentially enhancing clinical decision-making.
Accurate preoperative discrimination of salivary gland pleomorphic adenoma (SPA) stromal subtypes is essential for therapeutic plannings. We aimed to establish and test machine learning (ML) models for classification of stromal subtypes in SPA based on ultrasound histogram analysis. A total of 256 SPA patients were enrolled in the study and categorized into two groups: stroma-low and stroma-high. The dataset was split into a training cohort with 177 patients and a validation cohort with 79 patients. The least absolute shrinkage and selection operator (LASSO) regression identified optimal features, which were then utilized to build predictive models using logistic regression (LR) and eight ML algorithms. The effectiveness of the models was evaluated using a range of performance metrics, with a particular focus on the area under the receiver operating characteristic curve (AUC). After LASSO regression, six key features (lesion size, shape, cystic areas, vascularity, mean, and skewness) were selected to develop predictive models. The AUCs ranged from 0.575 to 0.827 for the nine models. The support vector machine (SVM) algorithm achieved the highest performance with an AUC of 0.827, accompanied by an accuracy of 0.798, precision of 0.792, recall of 0.862, and an F1 score of 0.826. The LR algorithm also exhibited robust performance, achieving an AUC of 0.818, slightly trailing behind the SVM algorithm. Decision curve analysis indicated that the SVM-based model provided superior clinical utility compared to other models. The ML model based on ultrasound histogram analysis offers a precise and non-invasive approach for preoperative categorization of stromal subtypes in SPA.
We aimed to develop and validate a nomogram based on conventional ultrasound (CUS) radiomics model to differentiate radial scar (RS) from invasive ductal carcinoma (IDC) of the breast. In total, 208 patients with histopathologically diagnosed RS or IDC of the breast were enrolled. They were randomly divided in a 7:3 ratio into a training cohort (n = 145) and a validation cohort (n = 63). Overall, 1316 radiomics features were extracted from CUS images. Then a radiomics score was constructed by filtering unstable features and using the maximum relevance minimum redundancy algorithm and the least absolute shrinkage and selection operator logistic regression algorithm. Two models were developed using data from the training cohort: one using clinical and CUS characteristics (Clin + CUS model) and one using clinical information, CUS characteristics, and the radiomics score (radiomics model). The usefulness of nomogram was assessed based on their differentiating ability and clinical utility. Nine features from CUS images were used to build the radiomics score. The radiomics nomogram showed a favorable predictive value for differentiating RS from IDC, with areas under the curve of 0.953 and 0.922 for the training and validation cohorts, respectively. Decision curve analysis indicated that this model outperformed the Clin + CUS model and the radiomics score in terms of clinical usefulness. The results of this study may provide a novel method for noninvasively distinguish RS from IDC.