AUTHOR=Wang Lanyun , Ding Yi , Yang Wenjun , Wang Hao , Shen Jinjiang , Liu Weiyan , Xu Jingjing , Wei Ran , Hu Wenjuan , Ge Yaqiong , Zhang Bei , Song Bin 

TITLE=A Radiomics Nomogram for Distinguishing Benign From Malignant Round-Like Breast Tumors

JOURNAL=Frontiers in Oncology

VOLUME=Volume 12 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2022.677803

DOI=10.3389/fonc.2022.677803

ISSN=2234-943X

ABSTRACT=Background: To develop a radiomics nomogram for presurgical distinction of cancerous from noncancerous round-like solid tumors.
Methods: This retrospective study enrolled patients with round-like tumors, who underwent digital mammography (DM) within 20 days preoperatively. Breast masses were segmented manually on DM images, extracting radiomic features. Four classification machine learning models were constructed, and respective areas under the receiver operating characteristic (ROC) curves (AUCs) for differential tumor diagnosis were obtained. The optimal classifier was selected (validation set). Then, predictive machine learning models employing radiomic features and/or patient features were applied for tumor assessment. The model’s AUC, accuracy, negative (NPV) and positive (PPV) predictive values, sensitivity and specificity were derived. 
Results: Totally 129 cases with pathologically confirmed benign and malignant tumors, including 91 and 38 in the training and testing cohorts, respectively, were examined. DM images yielded 1370 features/patient. The Least Absolute Shrinkage and Selection Operator for Gradient Boosting Classifier was the optimal classifier (AUC=0.87 [0.76-0.99]). ROC curves for the radiomics nomogram and the DM-only model were significantly different (P=0.000). The radiomics nomogram achieved an AUC of 0.90 (0.80-1.00) in the testing cohort, with a significant elevation versus the DM-based model (AUC=0.67[95%CI 0.51-0.84]). The radiomics nomogram was highly efficient in detecting malignancy, with accuracy, sensitivity, specificity, PPV and NPV of 0.868, 0.950, 0.778, 0.826, and 0.933, respectively (validation set). 
Conclusions: This radiomics nomogram that combines radiomics signature and clinical characteristics represents a noninvasive, cost-efficient presurgical prediction technique.