Objective:

To develop and validate a radiomics nomogram for preoperative prediction of lymphovascular invasion (LVI) in patients with invasive breast cancer (IBC).

Methods:

In this study

258 patients with histological

ly confirmed IBC with or without LVI

were randomly assigned to training and validation dataset at 7∶3. The maximum relevance minimum redundancy (mRMR) and the least absolute shrinkage and selection operator (LASSO) algorithms were used to screen the radiomics features obtained from two-dimensional ultrasound images

and the ultrasound-based radiomics score was established by these features. A final nomogram

combining the radiomics features and clinical features

for predicting LVI of IBC was established. Nomogram performance was assessed via both calibration and discrimination statistics.

Results:

Obtained from 10 features

the radiomics score indicated a favorable discriminatory capability in the training set with an area under curve (AUC) of 0.801

whereas a value of 0.786 was observed in the validation set. In the clinical model

ultrasound-reported axillary lymph node (ALN)

ultrasound Breast Imaging Reporting and Data System (BI-RADS) classification were effective for LVI prediction in IBC. The final nomogram integrated the ultrosound radiomics score and two clinical features. Good calibration was achieved for the nomogram in both the training and validated datasets (AUC of training set: 0.865; AUC of validation set: 0.857). DCA demonstrated that the combined model was superior to the others in terms of clinical practicability.

Conclusion:

The above described radiomics nomogram can preoperativelypredict LVI in patients with IBC and may constitute a usefully clinical tool to guide subsequent personalized treatment.