Table 2 Performance of intratumoral radiomics models in the validation group.

From: The value of intratumoral and peritumoral ultrasound radiomics model constructed using multiple machine learning algorithms for non-mass breast cancer

Radiomics modelsa

AUC (95%CI)

ACC (95%CI)

SEN (95%CI)

SPE (95%CI)

PPV (95%CI)

NPV (95%CI)

RS(in_RF)

0.793 (0.733–0.853)

0.736 (0.671–0.801

0.682 (0.613–0.751)

0.789 (0.728–0.850)

0.759 (0.696–0.823)

0.717 (0.650–0.784)

RS(in_DT)

0.619 (0.508–0.731)

0.645 (0.535–0.755)

0.976 (0.936–1.015)

0.257 (0.157–0.358)

0.606 (0.494–0.718)

0.900 (0.830–0.970)

RS(in_ET)

0.778 (0.717–0.840)

0.730 (0.664–0.796)

0.727 (0.661–0.793)

0.733 (0.668–0.799)

0.727 (0.661–0.793)

0.733 (0.668–0.799)

RS(in_SVM)

0.791 (0.730–0.851)

0.753 (0.689–0.817)

0.670 (0.601–0.740)

0.833 (0.778–0.889)

0.797 (0.738–0.857)

0.722 (0.655–0.788)

RS(in_LR)

0.631 (0.559–0.702)

0.626 (0.554–0.697)

0.748 (0.683–0.812)

0.461 (0.387–0.534)

0.653 (0.582–0.723)

0.574 (0.501–0.647)

RS(in_SGD)

0.716 (0.649–0.783)

0.682 (0.613–0.750)

0.659 (0.589–0.729)

0.703 (0.636–0.771)

0.682 (0.614–0.751)

0.681 (0.612–0.750)

RS(in_KNN)

0.733 (0.668–0.799)

0.682 (0.613–0.750)

0.652 (0.582–0.723)

0.713 (0.646–0.780)

0.706 (0.638–0.773)

0.660 (0.590–0.760)

RS(in_XGBoost)

0.805 (0.746–0.863)

0.764 (0.701–0.827)

0.716 (0.649–0.783)

0.811 (0.753–0.869)

0.786 (0.727–0.848)

0.745 (0.680–0.810)

RS(in_Adaboost)

0.628 (0.557–0.700)

0.615 (0.543–0.686)

0.871 (0.821–0.921)

0.337 (0.267–0.407)

0.587 (0.514–0.660)

0.707 (0.640–0.775)

RS(in_GBDT)

0.778 (0.717–0.840)

0.758 (0.695–0.822)

0.682 (0.613–0.751)

0.833 (0.778–0.889)

0.800 (0.741–0.859)

0.728 (0.662–0.794)

RS(in_CatBoost)

0.749 (0.685–0.814)

0.682 (0.613–0.750)

0.522 (0.448–0.596)

0.851 (0.798–0.904)

0.787 (0.726–0.848)

0.627 (0.556–0.699)

RS(in_LightGBM)

0.716 (0.649–0.783)

0.659 (0.589–0.729)

0.556 (0.482–0.629)

0.817 (0.760–0.87)

0.822 (0.765–0.879)

0.547 (0.474–0.621)

RS(in_Bayes)

0.708 (0.640–0.775)

0.721 (0.654–0.787)

0.912 (0.869–0.954)

0.468 (0.394–0.541)

0.694 (0.626–0.762)

0.800 (0.741–0.859)

  1. aThe intratumoral radiomics model was built using 13 algorithms: Random Forest (RF), Decision Tree (DT), Extra Trees (ET), Support Vector Machine (SVM), Logistic Regression (LR), Stochastic Gradient Descent (SGD), K-Nearest Neighbors (KNN), XGBoost, AdaBoost, Gradient Boosting Decision Tree (GBDT), CatBoost, LightGBM, and Bayes. AUC: area under curve, ACC: accuracy, SEN: sensitivity ,SPE: specificity, PPV: positive predictive Value, NPV: negative predictive value, CI: confidence interval.
Back to article page