Machine learning models predict histological grading of breast ductal carcinoma in situ:based on mammography signs and clinical informations
YANG Lingqiao
YANG Jun
MA Mengwei
CHEN Weiguo
XU Zeyuan
Abstract：Objective To explore the feasibility of constructing a machine learning model based on mammography signs and clinical informations to predict the histological grade in the ductal carcinoma in situ.Methods A retrospective analysis were conducted on the mammography signs and clinical informations of 239 patients who had histologically confirmed breast ductal carcinoma in situ(DCIS).Based on pathological results,these patients were categorized into:non-high-grade group(n=109)and high-grade group(n=130).The collected 10 clinical informations and 15 mammography signs were statistically analyzed,and the features with statistical differences were selected to construct three machine learning models,namely eXtreme Gradient Boosting,logistic regression and multinomial naive bayes,with the area under the ROC curve(AUC)was used as the main index to select the optimal mode.Results The AUC values for the training sets of eXtreme Gradient Boosting,logistic regression and multinomial Naive Bayes were 0.790,0.794,0.802,and the AUC values of text sets were 0.760,0.758,0.774,and the accuracies were 0.760,0.759,0.774,the sensitivities were 0.725,0.825,0.800,the specificities were 0.625,0.434,0.625.Conclusion The histological grade models of ductal carcinoma in situ based on machine learning have better prediction efficiency,and the multinomial naive Bayes has the best prediction efficiency.
Keywords：breast cancerductal carcinoma in situhistology grademammographymachine learning
Publication Date：2025-01-27
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：6( 31-36 )