Objective:

To explore the additional value of preoperative CT-based radiomics signature for non-invasive prediction of Ki-67 proliferation index in breast cancer.

Methods:

We retrospectively collected the data from 245 patients who were pathologically diagnosed with invasive breast cancer from May. 2016 to Dec. 2017. All patients were performed routine staging enhanced chest CT before surgery. The pat

ients were chronologically divided into training group (145 patients) and validation group (100 patients). Regions of interest (ROI) were manually delineated around the tumor profile. Using the interclass correlation coefficients

mRMR algorithm and Boruta algorithm

we performed feature selection and construction of the radiomics signature. The predictive performances of the radiomics signatures for Ki-67 proliferation index were evaluated with receiver operating characteristic (ROC) curve in training group

then validated in validation group with the obtained predictive threshold.

Results:

The radiomics signatures were constituted by eight selective features

showed good discrimination for Ki-67 proliferation index

with area under curve (AUC) of 0.782 (95% CI: 0.691-0.874) in training group and 0.781 (95% CI: 0.686-0.876) in validation group.

Conclusion:

The radiomics signatures based on preoperative staging CT have certain value for preoperative prediction of Ki-67 proliferation index in breast cancer

which can provide additional value of preoperative staging enhanced chest CT for clinical treatment decisions

and it may serve as a noninvasive approach to facilitate the preoperative prediction of Ki-67 proliferation index in clinical practice.