Development of a prediction model for HER2 low breast cancer using quantitative intra- and peri-tumoral heterogeneity and MRI features on high-spatial resolution ultrafast DCE-MRI

Abstract

Background: Accurate preoperative human epidermal growth factor receptor 2 (HER2) status assessment is crucial for guiding treatment selection, particularly with the emergence of anti-HER2 antibody-drug conjugates (ADCs) for HER2-low breast cancer. However, current immunohistochemistry (IHC)-based classification is limited by spatial heterogeneity and sampling bias. Quantitative analysis of intra- and peri-tumoral heterogeneity (ITH) on imaging may offer a non-invasive, objective, and reproducible approach to distinguish HER2-low breast cancer from other subtypes. This study aimed to investigate quantitative ITH from high-spatial resolution ultrafast dynamic contrast-enhanced magnetic resonance imaging (UF DCE-MRI) based kinetic curves in distinguishing HER2 low from HER2 zero or positive breast cancer.

Methods: Consecutive breast cancer patients who underwent preoperative high-spatial-resolution UF DCE-MRI were retrospectively enrolled. They were stratified into HER2 zero, HER2 low, or HER2 positive groups based on IHC and in situ hybridization results. Traditional MRI findings and clinicopathological characteristics were evaluated, and personalized ITH scores were constructed using semi-quantitative parameters derived from kinetic curves. Models incorporating ITH, MRI, and clinicopathological distinctions were developed for dichotomized HER2 statuses prediction using multivariable logistic regression. The added value of ITH in the Final Combined Model was evaluated.

Results: This study enrolled 368 patients, with 45.9% (169/368) having HER2-low breast cancer. The ITH score was higher in HER2 low than that in HER2 zero (P<0.001), but lower than that in HER2 positive (P<0.001). The ITH score was higher in HER2 positive compared to HER2 zero (P<0.001). The Final Combined Model integrating ITH, MRI, and clinicopathological variables achieved good predictive performance, achieving area under the curve (AUC) values of 0.80 [95% confidence interval (CI): 0.75-0.86] for HER2 low vs. zero, 0.85 (95% CI: 0.80-0.89) for HER2 low vs. positive, and 0.83 (95% CI: 0.77-0.88) for HER2 zero vs. positive. The corresponding sensitivity/specificity values were 77%/72%, 77%/81%, and 94%/58%, respectively. The ITH score significantly enhanced HER2 status prediction, supported by AUC improvement (DeLong test, P<0.05), along with statistical significance in net reclassification improvement (NRI) (P<0.001) and integrated discrimination improvement (IDI) (P<0.001) across all tasks.

Conclusions: Integrating ITH from high-spatial resolution UF DCE-MRI-based kinetic curves improved the non-invasive differentiation of HER2-low breast cancer. This approach may guide targeted biopsy strategies and aid in selecting candidates for anti-HER2 ADC therapy, optimizing HER2-targeted precision medicine.

Keywords: Breast neoplasms; human epidermal growth factor receptor 2 (HER2); magnetic resonance imaging (MRI).

Figure 1

Flowchart showing criteria for patient inclusion and exclusion. HER2 categorization follows American Society of Clinical Oncology guidelines (2). DCE-MRI, dynamic contrast-enhanced magnetic resonance imaging; DCIS, ductal carcinoma in situ; HER2, human epidermal growth factor receptor 2.

Figure 2

Workflow of ITH analysis. (A) Representative axial image from the final phase of contrast-enhanced T1-weighted DCE-MRI. (B) Three-dimensional segmentation of the index tumor. (C) Subregional mapping of intra-tumoral heterogeneity. (D) Temporal signal intensity curve of the intra-tumoral region. (E) Subregional mapping of peri-tumoral heterogeneity. (F) Temporal signal intensity curve of the peri-tumoral region. DCE-MRI, dynamic contrast-enhanced magnetic resonance imaging; ITH, intra- and peri-tumoral heterogeneity; SI, signal intensity; TTP, time to peak intensity; WIS, wash-in slope; WR, washout rate.

Figure 3

Box plots show ITH scores stratified by true dichotomized HER2 status. The mean ITH score of HER2 low tumors was higher than that of HER2 zero tumors (P<0.001), but lower than that of HER2-positive tumors (P<0.001). Additionally, the mean ITH score of HER2 zero tumors was lower than that of HER2-positive tumors (P<0.001). P values were calculated using the Mann-Whitney test. HER2, human epidermal growth factor receptor 2; ITH, intra- and peri-tumoral heterogeneity.

Figure 4

ITH color mapping and T2WI of BC patients with HER2 low (A-C), HER2 zero (D-F), and HER2 positive (G-I) expression. (A-C) A 48-year-old woman with hormone receptor-positive no special type carcinoma in the left breast had HER2 low expression. The intra-tumor (A) and peri-tumor (B) regions were segmented and characterized using color mapping in axial fat-suppressed DCE-MRI. Her personalized ITH score was 0.59. (C) Axial T2-weighted image of the same woman showed no peri-tumoral edema. (D-F) Another 54-year-old woman with hormone receptor-negative metaplastic carcinoma in the right breast had HER2 zero expression. The intra-tumor (D) and peri-tumor (E) regions were segmented and characterized using color mapping in the axial fat-suppressed DCE-MRI (personalized ITH score =0.49). (F) An axial T2-weighted image in this woman showed pre-pectoral edema (white arrow). (G-I) A 52-year-old woman with hormone receptor-negative no special type carcinoma in the left breast had HER2 positive expression. The intra-tumor (G) and peri-tumor (H) was segmented and characterized using color mapping in the axial fat-suppressed DCE-MRI (personalized ITH score =0.84). (I) Axial T2-weighted image show pre-pectoral edema (white arrow). BC, breast cancer; DCE-MRI, dynamic contrast-enhanced magnetic resonance imaging; HER2, human epidermal growth factor receptor 2; ITH, intra- and peri-tumoral heterogeneity.

Figure 5

Boxplots show the predictive performance of Final Combined Models in predicting true dichotomized HER2 status. The predictive performance of Final Combined Models for all dichotomized HER2 tasks yielded P values less than 0.001, calculated using the Mann-Whitney test. HER2, human epidermal growth factor receptor 2.

Figure 6

Performance of the Final Combined Models for distinguishing dichotomized HER2 status. AUC, area under the curve; CI, confidence interval; HER2, human epidermal growth factor receptor 2; ROC, receiver operating characteristic.