Breast cancer is the most frequent malignant tumor in women with the highest mortality. It remains a worldwide public health dilemma, leading to 450,000 deaths each year [
1‐
3]. Although it is curable in ~ 70–80% of patients with early-stage, non-metastatic disease, advanced breast cancer with distant organ metastases is considered incurable with currently available therapies. During the last 20 years, five intrinsic molecular subtypes of breast cancer (Luminal A, Luminal B, human epidermal growth factor receptor 2 (HER2)-enriched, Basal-like and Claudin-low) have been identified as a result of activation of these genes [
4]. Studies have shown that for patients with breast cancer at early stage, the subtype is more important to define treatment strategies, determine the therapeutic outcome and prognosis than histopathologic type [
5,
6]. As a heterogeneous disease, the biological characteristics and clinical behaviors of breast cancer are different among the several distinct entities [
7]. Triple negative breast cancer (TNBC), a special clinical pathological subtype of breast cancer with negative expressions of estrogen receptor (ER), progesterone receptor (PR), and HER-2 [
8], accounts for about 10 to 20% of breast cancer. It is the most aggressive subtype of breast cancer [
7]. No effective targeted molecular therapy is available for TNBC and its prognosis is generally poor [
9]. Clinically, chemotherapy is the only effective treatment for TNBC [
10,
11]. Therefore, early diagnosis of TNBC from non-TNBC (NTNBC) patients is important for better planning of therapy strategies and for predicting the response to neoadjuvant chemotherapy (NCT) administered before surgery for breast cancer [
12]. At present, molecular subtyping of breast cancer depends mainly on immunohistochemistry analysis, in which biopsy is used to collect tumor tissue sample. This is an invasive method and there are limitations for sampling and analysis [
13]. In contrast, imaging is noninvasive and can reflect the overall characteristics of tumor, allowing analysis of difference among subtypes at molecular level and dynamical evaluation of therapeutic outcomes [
14,
15].
As a result of implementation of recommendations by the National Comprehensive Cancer Network (NCCN) guidelines, contrast-enhanced chest CT scan has become a part of routine preoperative examination for breast cancer patients in some hospitals. Different from X-ray mammography, ultrasonography and magnetic resonance imaging of breast lesions, the primary purpose of contrast-enhanced chest CT is to assist clinically staging the disease [
16]. It is generally believed that CT is not as good as X-ray mammography to show microcalcification nor as accurate as ultrasound to diagnose breast cystic lesions, although it can incidentally detect breast cancer [
17]. It is generally not capable to differentiate benign and malignant breast lesions [
18]. However, with the development of radiomics [
19‐
21], which can provide a comprehensive quantification of the tumor phenotype by analyzing robustly [
10‐
12] a large set of quantitative data with characterization algorithms [
22], it has been become possible to use imaging features for treatment monitoring and outcome prediction in various cancers [
23,
24]. Therefore, we speculate that this method can extract information that is inviable to naked eye from routine preoperative contrast-enhanced chest CT scans for molecular subtyping and characterizing biological features of breast cancer. This would provide new tools and additional information from the routine preoperative CT to characterize the lesions, in addition to assist clinical staging of tumors.
In this study, we attempted to extract radiomic features on the images of preoperative contract-enhanced chest CT for diagnosing TNBC without additional radiation exposure and cost. The findings would optimize planning of treatment scheme for patients with breast cancer.