Study on the Differentiation of Infiltrating Breast Cancer Molecular Subtypes Based on Ultrasound Radiomics

To establish and validate a 2-dimensional ultrasound (US) radiomics model for the noninvasive preoperative differentiation of various molecular subtypes of infiltrating breast cancer (IBC).

A retrospective analysis of 210 female patients diagnosed with IBC through surgical operation or needle biopsy pathology at our hospital between May 2019 and February 2024 was conducted. Relevant data were collected to establish predictive models for different molecular subtypes of IBC.

Based on 5936 US radiomics features, 39, 25 and 19 optimal features were identified for the differentiation of luminal versus nonluminal types, luminal A versus luminal B types and human epidermal growth factor receptor 2 (HER2) overexpression versus triple-negative (TN) IBC subgroups, respectively. The corresponding areas under the curve for the training and validation sets were 0.901 and 0.752 (luminal vs. nonluminal), 0.931 and 0.773 (luminal A vs. luminal B) and 0.962 and 0.842 (HER2 overexpression vs. TN), respectively, indicating robust discriminatory performance of these models for different pathological molecular subtypes of IBC.

A radiomics model based on US images is capable of effectively differentiating between various molecular subtypes of IBC prior to surgery, holding promise in assisting medical professionals in crafting tailored diagnostic and therapeutic strategies.