Quantitative MRI mapping parameters to differentiate HER2-positive and HER2-low breast cancer

Abstract

Background

Breast cancer (BC) is the most common malignant tumor among women worldwide, and early detection and precise molecular subtyping are crucial for improving survival rates. Quantitative magnetic resonance imaging (MRI) mapping techniques have demonstrated potential in evaluating tumor tissue characteristics. The aim of this study was to evaluate the value of quantitative MRI mapping parameters in determining human epidermal growth factor receptor 2 (HER2) status in BC.

Methods

In this retrospective study, 178 women with histologically confirmed BC underwent preoperative MRI, including longitudinal relaxation time (T1), transverse relaxation time (T2), and effective transverse relaxation time (T2*) mapping sequences. Following triplicate region of interest (ROI) measurements by two independent radiologists per lesion, the median value for each radiologist was calculated and then averaged between them to derive the final quantitative parameter. Quantitative mapping parameters were compared across HER2 groups with Bonferroni-adjusted tests. Multivariable regression was performed to identify independent predictors for differentiating HER2 statuses.

Results

Significant differences in estrogen receptor (ER) and progesterone receptor (PR) status were observed between the HER2-low and HER2-positive groups (P<0.05). Pairwise comparisons showed that HER2-low BCs had significantly higher ΔT1 [ΔT1 = T1_pre (T1_pre is the native T1 relaxation time measured before contrast agent administration) − T1_post (T1_post is the T1 relaxation time measured after contrast injection)] and ΔT1_pre [ΔT1_pre = (T1_pre − T1_post)/T1_pre] values, lower T1_post and T2* values (all P<0.05) compared to HER2-positive BCs, while overall differences across all the three HER2 groups were not observed for all mapping parameters. Multivariate logistic regression analysis showed that T2* and T1_post were the most discriminative indexes [the area under the curve (AUC) =0.800 and 0.762, respectively] and ER was optimal pathological differentiation index (AUC =0.617). The multivariable model incorporating ER state, values of T2* and T1_post demonstrated favorable performance with AUC of 0.854.

Conclusions

A combined model incorporating ER state, T2* and T1_post value showed good discriminative ability in non-invasively differentiating HER2-low from HER2-positive BCs.

Introduction

Breast cancer (BC) is the most common malignancy among women worldwide (1), accounting for 11.7% of new cancers and 6.9% of cancer-related deaths annually (2). Its incidence continues to rise, particularly among younger women, making early detection and accurate molecular classification critical for improving survival (3).

BC is classified into four molecular subtypes based on the expression of estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2 (HER2), and Ki-67 proliferation index (Ki-67 using 14% as the cutoff value): LuminalA, LuminalB, HER2-positive, and triple-negative breast cancer (TNBC). HER2-positive tumors [immunohistochemistry (IHC) 3+ or IHC 2+/fluorescence in situ hybridization (FISH) 1+] are typically more aggressive and have benefited from HER2-targeted therapies (4,5). However, the recent recognition of HER2-low (IHC 1+ or IHC 2+/FISH −) as a distinct biological and therapeutic entity has reshaped treatment paradigms (6).

A phase III clinical trial showed that trastuzumab deruxtecan (T-DXd) significantly improved progression-free and overall survival in patients with HER2-low BC compared with standard chemotherapy (7) leading to guideline updates by the American Society of Clinical Oncology (ASCO) in 2023 (8). These developments underscore the urgent need for precise identification of HER2-low tumors.

Currently, HER2 status is determined by IHC and FISH, methods prone to sampling bias and tumor heterogeneity, especially in core needle biopsies where HER2-amplified and non-amplified cells may be intermingled (5). In this context, non-invasive imaging approaches capable of reflecting tumor microstructure and heterogeneity could complement conventional pathology.

Quantitative magnetic resonance imaging (MRI) mapping, through T1 (longitudinal relaxation time), T2 (transverse relaxation time), and T2* relaxation time measurements, has shown potential in assessing tissue characteristics such as cellularity, fibrosis, and vascularity (9-13). Although initially developed for brain imaging (9), MRI mapping has expanded to applications in thoracic, abdominal, and musculoskeletal diseases (10-13), including BC diagnosis, molecular typing, and prognosis assessment (14-20). However, its potential role in differentiating HER2-low from HER2-positive BCs remains underexplored. Therefore, this study aimed to investigate whether quantitative MRI mapping parameters, particularly T2*Map and T1Map_post, could effectively distinguish HER2-low and HER2-positive subtypes, providing a non-invasive imaging biomarker to assist in HER2 classification and therapeutic stratification. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1449/rc).

Methods

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of The First Affiliated Hospital of Yangtze University prior to data collection (approval No. LLL202380). Written informed consent was obtained from all participants.

Patients

From July 2022 to May 2024, clinical data from 223 women with breast masses were collected at the Department of Breast Surgery of The First Affiliated Hospital of Yangtze University. Inclusion criteria were as follows: (I) female patients aged 30 to 70 years; (II) presence of a suspicious breast mass detected clinically or on imaging; (III) no prior treatment (surgery, chemotherapy, or radiotherapy) before MRI examination; (IV) preoperative multiparametric breast MRI (including mapping sequences); (V) pathologically confirmed diagnosis of BC after surgical resection or biopsy; (VI) complete clinical, pathological, and imaging data available for analysis. Exclusion criteria were: (I) patients who underwent biopsy prior to MRI examination (n=14); (II) patients with a history of chemotherapy or radiotherapy prior to imaging (n=7); (III) incomplete clinical, pathological, or imaging data (n=10); (IV) poor image quality due to artifacts (n=4); (V) benign breast lesions confirmed by histology (n=7); (VI) contraindications to MRI (n=3); (VII) patients with multifocal or multicentric BC (n=0), due to potential heterogeneity and inability to standardize ROI selection. A total of 178 BC patients were ultimately included in this study. Figure 1 shows the flowchart of patient enrollment and exclusion.

Figure 1.

Flowchart of patient inclusion and exclusion. DCE, dynamic contrast-enhanced; HER2, human epidermal growth factor receptor 2; MRI, magnetic resonance imaging.

Pathological assessment of immunohistochemical biomarkers

The ER, PR and HER2 statuses of all BC specimens were assessed with IHC testing according to the 2023 American Society of Clinical Oncology/College of American Pathologists (ASCO/CAP) guidelines. ER and PR positivity were defined as nuclear immunostaining ≥1%. Assessment for HER2 statuses were: (I) IHC 0: no staining observed or incomplete membrane staining that is faint/barely perceptible and in ≤10% of tumor cells; (II) IHC 1+: incomplete membrane staining that is faint/barely perceptible in >10% of tumor cells; (III) IHC 2+: weak to moderate complete membrane staining observed in >10% of tumor cells (equivocal, requiring FISH testing); (IV) IHC 3+: circumferential, intense membrane staining in >10% of tumor cells. Cases with an IHC score of 2+ underwent FISH testing. FISH was interpreted as positive if HER2/CEP17 ratio ≥2.0 with average HER2 copy number ≥4.0 signals per cell, or if HER2 copy number ≥6.0 regardless of ratio. Based on these results, patients were classified into three groups: (I) HER2-positive: IHC 3+, or IHC 2+ with FISH positive; (II) HER2-low: IHC 1+ or IHC 2+ with FISH negative; and (III) HER2-zero: IHC 0.

All pathological evaluations were conducted by two board-certified breast pathologists independently, with discrepancies resolved by consensus.

MRI examination

A Siemens Skyra 3.0T MRI (Siemens Healthineers AG, Munich, Germany) was used for scanning. The scan sequences included conventional breast MRI, T1Map_pre, T2Map, T2*Map, dynamic contrast-enhanced MRI (DCE-MRI), and T1Map_post. Pre-contrast T1 mapping (T1Map_pre) provided a baseline assessment of intrinsic tissue composition. DCE-MRI captured the subsequent hemodynamic dynamics and microvascular perfusion during the first pass of the contrast agent. Finally, post-contrast T1 mapping (T1Map_post), acquired after contrast equilibrium, allowed for the quantification of contrast agent distribution in the extracellular space. These three components constitute an integrated quantitative framework, encompassing static baseline characterization, dynamic process imaging, and delayed phase distribution analysis. Together, this multi-parametric approach enables a synergistic analysis of tissue structure, perfusion, and interstitial volume. Prior to scanning, thorough communication and reassurance were provided to patients to minimize anxiety-induced movement resulting from uncertainty or fear, coupled with concurrent respiratory coaching and training. Parallel imaging techniques were used to reduce scan time, and respiratory motion artifacts were mitigated by orienting the phase-encoding direction along the left-right axis and implementing BLADE (a radial k-space filling) acquisition strategy. The contrast agent dd-diethylenetriaminepentaacetic acid (Dd-DTPA) (gadoterate meglumine, Bayer Schering Pharma, Germany) was injected via an antecubital vein using a high-pressure syringe at a dosage of 0.2 mmol/kg and the rate of 2.5 mL/s. The T1Map_post scan was performed 9 min 40 s after enhancement.

T1 mapping was acquired using a three-dimensional spoiled gradient-echo sequence with a variable flip angle and short repetition time (TR). T2 mapping was performed using a two-dimensional multi-echo spin-echo (2D-MC-SE) sequence. T2* mapping was acquired with a two-dimensional fast gradient-echo (2D-FLASH) sequence. Scanning parameters for mapping and DCE-MRI are shown in Table 1. Figure 2 is intended as a visual comparison of multiparametric MRI sequences in a representative BC patient. It displays coregistered images from four quantitative techniques: the pre-contrast T1 map (T1Map_pre) reflecting native tissue composition; the post-contrast T1 map (T1Map_post) showing contrast agent distribution; the T2 map sensitive to tissue water content; and the T2 map* sensitive to magnetic field inhomogeneities.

Table 1. Sequences and acquisition parameters of mapping and DCE-MRI.

Quantitative maps imaging parameters	TR (ms)	TE (ms)	FOV (mm2)	Slice thickness (mm)	Gap (mm)	Slice
T1Map_pre	4.54	2.02	300×300	1.6	–	104
T2Map	2,410	13.8	300×300	4.0	1.2	30
T2*Map	724	3.54	300×300	4.0	1.2	30
DCE-MRI	4.95	2.39	300×300	1.6	–	104
T1Map_post	4.54	2.02	300×300	1.6	–	104

DCE-MRI, dynamic contrast-enhanced magnetic resonance imaging; FOV, field of view; TE, echo time; TR, repetition time.

Figure 2.

This figure shows the four core parametric MRI maps from a 53-year-old woman diagnosed with HER2-low breast cancer. (A) Pre-contrast T1 map (T1Map_pre): reflects native tissue composition. (B) Post-contrast T1 map (T1Map_post): shows the distribution of gadolinium contrast agent in the extracellular space. (C) T2 map: sensitive to tissue water content and edema. (D) T2 map*: sensitive to local magnetic field inhomogeneities caused by substances such as iron or blood products. Yellow arrows indicate the breast tumor. HER2, human epidermal growth factor receptor 2; MRI, magnetic resonance imaging.

Image processing and analysis

All images were imported into the post-processing platform (Siemens Syngo MR E11). For T1 mapping, a Bloch-Siegert B1 mapping technique (acquisition time: 6 s) was applied to correct for B1 field inhomogeneity (21). The T1 values were calculated pixel-wise using the linearized Ernst equation (22). In our study, to correct for potential patient motion during the multi-echo acquisition for T2 mapping, all images at different echo times (TEs) were co-registered to the first echo image (shortest TE) using a standard rigid-body registration algorithm. This is a common preprocessing step in quantitative MRI to ensure voxel-wise correspondence for accurate fitting. The image processing was performed using tools from the FSL software library (specifically, the FLIRT tool) (23). Mono-exponential fitting was then applied to derive T2 values for each pixel. Multi-channel data were combined using the sum-of-squares method to generate composite images of T2* mapping. Similar to T2 mapping, mono-exponential fitting was employed to calculate T2* values. The pixel-wise T1 value was calculated by fitting a three-parameter mono-exponential recovery model to the signal from five source images. These five images were acquired at five different inversion times (TI =100, 200, 350, 500, 800 ms) using a modified Look-Locker inversion recovery (MOLLI) sequence, with a fixed short TE (2.02 ms) to minimize T2* effects. The pixel-wise T2 value was calculated by fitting a mono-exponential decay model to the signal from [13.8] source images. These 13.8 images were acquired at 13.8 different TE (13.8 ms). The pixel-wise T2* value was calculated by fitting a mono-exponential decay model to the signal from 3.54 source images. These 3.54 images were acquired at 3.54 different TE (3.54 ms), after multi-channel combination using the sum-of-squares method. Pixels with signal-to-noise ratio (SNR) below 5 in the final echo were excluded from fitting analysis. This quality control criterion affected fewer than 2% refers to the proportion of pixels within the manually delineated breast tissue region of interest (ROI), contributing to the reliability of the final T2* map [for each lesion, a circular ROI was placed over the central portion of the tumor to obtain the mean signal intensity (S) and its standard deviation (SD)]. To estimate the background noise, four rectangular ROIs were placed in the artifact-free background outside the patient’s body. The mean background signal (Sb) was calculated as the average of these four measurements: Sb = (Sb1 + Sb2 + Sb3 + Sb4)/4. Subsequently, the SNR was calculated as SNR = (S − Sb)/SD. The largest 2D-primary tumor cross-sectional area was used as an ROI for quantitative image analysis. ROI delineation was performed manually on the axial slice showing the maximum tumor diameter on the DCE-MRI sequence acquired 90 s after contrast administration. This slice was selected as it typically provides optimal tumor-to-background contrast and is most commonly used in clinical evaluation (24). The ROIs were then transferred to the corresponding T1Map_pre, T2Map, T2*Map, and T1Map_post images for quantitative measurement. ROI delineation was independently performed by two board-certified breast radiologists, each with more than 7 years of experience who were blinded to patients’ information and pathologic results. For each lesion, two radiologists performed three measurements of the ROI independently, and the median of the three measurements was taken. The average value of the two radiologists was then calculated. The final results for all imaging parameters were determined using the final average value as the criterion. The calculation methods for the derived T1 mapping parameters (ΔT1 and ΔT1_pre) were as follows: ΔT1 = T1_pre − T1_post; ΔT1_pre = (T1_pre − T1_post)/T1_pre.

Where T1_pre is the native T1 relaxation time (in milliseconds) measured before contrast agent administration. T1_post is the T1 relaxation time (in milliseconds) measured approximately 9 min and 40 s after contrast injection. ΔT1 represents the absolute reduction in T1 relaxation time due to contrast agent accumulation within the tumor tissue. ΔT1_pre represents the relative reduction (percentage change) in T1 relaxation time, standardized by the baseline value.

Statistical analysis

Statistical analyses were performed using SPSS 26.0, and graphs were created with GraphPad Prism 9.0.0. A P value <0.05 was considered a statistically significant difference. Normality of variances was assessed by the one-sample Kolmogorov-Smirnov test. ANOVA and Bonferroni correction or Tamhane’s T2 test were used for normally distributed data, counting data was compared using χ² test.

The final averaged ROI value served as the imaging parameter and clinical factors showing a significant difference between HER2-low and HER2-positive groups in univariate analysis were enrolled for multivariate logistic regression analysis (25) to select independent predictive factors of HER2 statuses, and a multivariate model incorporating all independent parameters was then constructed. Receiver operating characteristic (ROC) curves were plotted for parameters showing significant differences and the multivariate model with area under ROC curves (AUCs) to evaluate the efficacy of quantitative and qualitative parameters in distinguishing between HER2-low and HER2-positive groups, and compared using pairwise DeLong test (26). Youden index (27) (the sensitivity value plus the specificity value) was calculated for identifying optimal cut-off value of the joint model. A forest plot (28) was generated to visualize the effect sizes across both models.

Results

Patient characteristics

A total of 178 patients were included, with a mean age of 48.9±9.9 years (range, 30–70 years). Histopathological subtypes comprised 81.5% (145/178) invasive ductal carcinoma (IDC) and 18.5% (33/178) invasive lobular carcinoma (ILC). Patients were stratified into three groups based on HER2 status: HER2-zero (n=20, mean age: 43.0±5.3 years), HER2-low (n=72, mean age: 49.4±9.3 years), and HER2-positive (n=86, mean age: 49.8±10.4 years), as shown in Table 2. Significant differences were observed in ER and PR status between the HER2-low and HER2-positive groups. Age differed significantly between the HER2-zero group and both HER2-low and HER2-positive groups, but no significant age difference was noted between HER2-low and HER2-positive groups. Menopausal status and Ki-67 index were comparable across all groups.

Table 2. Comparison of clinical and histopathological data between HER2-zero groups, HER2-low groups and HER2-positive groups.

Variable group	HER2-zero (n=20)	HER2-low (n=72)	HER2-positive (n=86)	P value
				All	HER2-zero vs. HER2-low	HER2-zero vs. HER2-positive	HER2-low vs. HER2-positive
Age (years)	43.0±5.3	49.4±9.3	49.8±10.4	0.15†	0.027†*	0.013†*	1†
Menopause (no/yes)	8/12	34/38	30/56	0.290‡	NA	NA	NA
ER (negative/positive)	4/16	20/52	44/42	0.02‡*	0.483‡	0.012‡*	0.003‡*
PR (negative/positive)	6/14	32/40	56/30	0.003‡*	0.246‡	0.004‡*	0.004‡*
Ki-67 (high/low)	18/2	60/12	80/6	0.155‡	NA	NA	NA

Data are presented as n or mean ± standard deviation. *, P<0.05; ^†, ANOVA test with Bonferroni correction; ^‡, Chi-squared test. Ki-67, marker of proliferation 67 (Ki-67 index ≥14% was used to define high proliferative activity). ANOVA, analysis of variance; ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; NA, not applicable; PR, progesterone receptor.

Comparison of imaging parameters

Quantitative MRI parameters differed significantly among the HER2-zero, HER2-low, and HER2-positive groups (Table 3 and Figure 3). The T1_pre values were highest in the HER2-zero group (2,226.6±265.6 ms), followed by the HER2-low group (1,903.4±373.6 ms) and the HER2-positive group (1,788.6±296.9 ms), with significant differences across groups.

Table 3. Comparison of mapping technology parameters between HER2-zero groups, HER2-low groups and HER2-positive groups.

Mapping technology parameters	HER2-zero (n=20)	HER2-low (n=72)	HER2-positive (n=86)	P value
				All	HER2-zero vs. HER2-low	HER2-zero vs. HER2-positive	HER2-low vs. HER2-positive
T1_pre, ms	2,226.640±258.622	1,903.449±373.553	1,788.628±296.89	<0.001†*	<0.001†*	<0.001†*	0.087†
T1_post, ms	468.490±134.681	462.264±126.798	580.930±158.158	<0.001†*	0.997†	0.008†*	<0.001†*
∆T1, ms	1,758.150±304.276	1,441.185±400.444	1,207.698±294.465	<0.001†*	0.001†*	<0.001†*	<0.001†*
∆T1_pre, ms	0.786±0.068	0.747±0.085	0.669±0.093	<0.001†*	0.226†	<0.001†*	<0.001†*
T2, ms	113.843±47.638	120.369±73.968	111.995±40.584	0.650†	1.000†	1.000†	1.000†
T2*, ms	43.847±17.187	35.920±11.257	53.233±16.949	<0.001†*	0.111†	0.037†*	<0.001†*

Data are presented as mean ± standard deviation. *, P<0.05; ^†, ANOVA test with Bonferroni correction. ANOVA, analysis of variance; HER2, human epidermal growth factor receptor 2.

Figure 3.

Violin plots showing the differences of quantitative mapping parameters among the HER2-zero, HER2-low and HER2-positive groups. (A) The T1_pre values were highest in the HER2-zero group, with statistically significant differences among the groups. (B) T1_post values were significantly lower in the HER2-low group compared with the HER2-positive group, whereas no significant difference was observed between the HER2-zero and HER2-low groups. (C) The ΔT1 values were highest in the HER2-zero group, with statistically significant differences observed among all three groups. (D) The ΔT1_pre values were higher in both the HER2-zero and HER2-low groups compared to the HER2-positive group, with statistically significant differences. (E) The T2 values were relatively low across all three groups, with no statistically significant differences observed among them. (F) The T2* values were higher in the HER2-positive group compared to the other two groups, showing statistically significant differences. *, P<0.05. HER2, human epidermal growth factor receptor 2; ns, not significant.

T1_post values were significantly lower in the HER2-low group (462.3±126.8 ms) compared with the HER2-positive group (580.9±158.2 ms, P=0.008), whereas no significant difference was observed between the HER2-zero and HER2-low groups (P=0.997). Similarly, the ΔT1 values were higher in HER2-low tumors (1,441.2±400.4 ms) than in HER2-positive tumors (1,207.7±294.5 ms, P<0.001). The ΔT1_pre values were also higher in HER2-low tumors (0.747±0.085) compared with HER2-positive tumors (0.669±0.093, P<0.001). In contrast, T2* values were significantly lower in HER2-low tumors than in HER2-positive tumors (P<0.001). No statistically significant difference in T2 values was found among the three groups (P=0.650).

Multivariable logistic analysis and features performance

The diagnostic performance of quantitative MRI parameters in distinguishing HER2-low from HER2-positive tumors is summarized in Table 4 and Figure 4. In univariate analysis, values of T1_post, ΔT1, ΔT1_pre, and T2* were significantly associated with HER2 status. The AUC values for individual imaging parameters ranged from 0.672 to 0.800. Specifically, T2* value demonstrated the highest AUC of 0.800, followed by T1_post with an AUC of 0.762. Sensitivity across parameters ranged from 66.3% to 89.5%, specificity from 55.6% to 86.1%, and overall accuracy from 58.3% to 87.0%. Tumor grade and Ki-67 status also showed modest but non-significant effects. These results suggest that the association between HER2 status and T2* value is not solely attributable to confounding by tumor differentiation or other biological characteristics. The detailed regression coefficients and confidence intervals are shown in Figure 5.

Table 4. ROC analysis of clinicopathological features and mapping parameters for HER2-low and HER2-positive groups.

Features	Cut-off	AUC	Sensitivity	Specificity
ER	0.5	0.617	0.722	0.512
PR	0.5	0.603	0.556	0.651
T1_post	461.55	0.762	0.872	0.639
∆T1	1447.839	0.672	0.884	0.583
∆T1_pre	0.749	0.750	0.895	0.556
T2*	47.1	0.800	0.663	0.861
Combined model	0.702	0.854	0.651	0.944

AUC, area under the curve; ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; PR, progesterone receptor; ROC, receiver operating characteristic.

Figure 4.

ROC curves of single diffusion metric and the combined model in discriminating HER2-low and HER2-positive breast cancers. The combined model includes T1 Map_post, T2* Map, and ER status. AUC, area under the curve; ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; ROC, receiver operating characteristic.

Figure 5.

Multivariable linear regression analysis of factors associated with T1Map_post and T2*Map values. Forest plot illustrating the regression coefficients and 95% CI from multivariable linear regression models assessing the associations between clinicopathological variables and two quantitative MRI parameters: T1Map_post (left) and T2Map (right). CI, confidence interval; HER2, human epidermal growth factor receptor 2; MRI, magnetic resonance imaging.

T2*, T1_post, and ER status were identified as independent predictors in differentiating HER2 statuses, the multivariable model combining these features achieved an AUC of 0.854, which was significantly higher than that of any single imaging metric (AUC range, 0.603–0.800). Comparisons between features and models are shown in Table 5.

Table 5. Comparisons of performance of significant features for differentiating HER2 status.

Comparison	ΔAUC	95% CI of ΔAUC	Z value	P value
T2* vs. T1_post	0.038	0.002–0.074	2.05	0.041
T2* vs. ΔT1	0.128	0.082–0.174	5.47	<0.001
T2* vs. ΔT1_pre	0.111	0.065–0.157	4.73	<0.001
T1_post vs. ΔT1	0.090	0.044–0.136	3.84	<0.001
T1_post vs. ΔT1_pre	0.073	0.027–0.119	3.11	0.002
Combined model vs. T2*	0.054	0.018–0.090	2.93	0.003
Combined model vs. T1_post	0.092	0.056–0.128	4.99	<0.001
T2* vs. ER status	0.183	0.127–0.239	6.42	<0.001
T2* vs. PR status	0.156	0.098–0.214	5.28	<0.001
T1_post vs. ER status	0.145	0.089–0.201	5.07	<0.001
T1_post vs. PR status	0.118	0.060–0.176	4.01	<0.001
Combined model vs. ER status	0.237	0.179–0.295	7.91	<0.001

AUC, area under the curve; CI, confidence interval; ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; PR, progesterone receptor.

Discussion

This study evaluated the role of quantitative MRI mapping parameters combined with clinicopathological variables in differentiating HER2-low from HER2-positive BCs. T2* and T1_post values demonstrated statistically significant differences between the two groups, achieving AUCs of 0.800 and 0.762, respectively. Moreover, a combined model incorporating T2*, T1_post, and ER status yielded an AUC of 0.854, outperforming any single indicator (AUC range, 0.603–0.800).

HER2 is a critical biomarker for BC diagnosis, prognostic assessment, and therapeutic targeting (29). The emergence of novel HER2 antibody-drug conjugates (ADCs) has highlighted the clinical importance of stratifying HER2 expression into high, low, and zero categories (7). Approximately 40% to 50% of BCs exhibit HER2-low expression (30), and classification is essential for personalized therapy, given the variable response rates observed with HER2-targeted agents in heterogeneous tumors (31). Currently, HER2 status determination relies on invasive tissue sampling followed by IHC and/or FISH, both of which are dependent on biopsy quality, tumor heterogeneity, and pathological interpretation. A study by Mao et al. (32) found that integrating high-spatial-resolution ultrafast (UF) DCE-MRI kinetic curves with intratumoral heterogeneity (ITH) improved the non-invasive differentiation of HER2-low BCs. This method may guide targeted biopsy strategies, aid in selecting candidates for anti-HER2 ADC therapy, and optimize precision medicine for HER2-targeted treatment. However, as the study was a small-scale retrospective analysis, further validation through large-scale prospective studies is still needed. Our findings suggest that quantitative MRI mapping parameters, particularly T2* and T1-derived metrics, may offer a non-invasive and spatially comprehensive method to infer HER2 expression patterns within the entire tumor mass. Such imaging biomarkers could be particularly valuable in scenarios where biopsy specimens are limited, fragmented or non-representative, tumor heterogeneity raises concerns about sampling bias and longitudinal monitoring of HER2 level during neoadjuvant therapy which needs lesion status at different times.

In the present study, the T2* values of HER2-low BCs were significantly lower than those of HER2-positive tumors. This suggests that HER2-positive tumors may exhibit greater disruption of fibrocollagenous architecture, resulting in reduced magnetic field heterogeneity at tissue interfaces and thus a relative prolongation of T2*. In contrast, HER2-low tumors with more preserved glandular and stromal structures generate stronger field inhomogeneity, leading to shorter T2* relaxation time. Although hypoxia is commonly observed in high-grade and rapidly proliferating tumors, increased local concentrations of deoxyhemoglobin can shorten T2*. The present findings suggest that the effect of static structural disruption outweighs the hypoxia-induced susceptibility effects in HER2-positive tumors at the macroscopic imaging level. Previous studies have reported that T2* relaxation time tends to increase with higher histological grades, supporting the notion that architectural changes associated with tumor progression may dominate the observed T2* signal alterations. Furthermore, calcifications, which can also cause localized magnetic field inhomogeneity and T2* shortening, were carefully avoided during ROI placement to minimize their confounding effects. Regarding T1 mapping, we observed that T1_post values were lower and ΔT1/ΔT1Map_pre values were higher in HER2-low compared with HER2-positive tumors. Post-contrast T1 shortening reflects vascular permeability and contrast agent retention. HER2-low tumors may have more newly formed capillaries with slower washout, resulting in higher enhancement ratios and shorter T1 relaxation times. To mitigate variability due to contrast kinetics, we acquired T1Map_post at 9 min 40 s after contrast injection, targeting the equilibrium phase, and calculated ΔT1Map and ΔT1Map_pre to reflect true enhancement dynamics.

A Few prior studies have addressed imaging differentiation between HER2-low and HER2-positive BC. Ramtohul et al. (33) utilized diffusion spectrum imaging to predict HER2 status, which achieved an AUC of 0.721, but did not specifically separate HER2-low tumors. Guo et al. (34) successfully differentiated HER2-low, HER2-positive BC, and HER2-zero tumors using radiomic features from T2-weighted imaging (T2WI) and DCE-MRI (AUC =0.79–0.80). Guo et al. (34) also applied DCE radiomics to distinguish HER2 subtypes with moderate accuracy (AUC =0.66–0.81). Recently, Zheng et al. (35) found that HER2-positive tumors exhibited larger volumes and greater peritumoral blood supply compared to HER2-low tumors. Mao et al. (31) further showed that αCTRW parameters from advanced diffusion models could differentiate HER2-low from HER2-positive BCs (AUC =0.877). A recent multi-center, large-sample study (35) found that diffusion-weighted imaging (DWI), ADC, and DCE features could distinguish low HER2 with AUC up to 0.782. Compared with these studies, our investigation is novel in employing direct quantitative mapping metrics (T2* and T1) rather than indirect radiomic features, thus providing a more biophysically grounded approach to characterize HER2 heterogeneity.

However, it is important to emphasize that MRI-based assessment currently cannot replace pathological confirmation of HER2 status. Rather, it may serve as a complementary tool to guide biopsy targeting, monitor treatment response, or stratify patients for further diagnostic evaluation. Compared with IHC/FISH, MRI offers advantages of being non-invasive, repeatable, and capable of capturing whole-tumor heterogeneity, but lacks the molecular specificity and gold-standard validation of direct tissue-based assays. Future large-scale, prospective, and multi-center studies are needed to validate MRI mapping signatures, standardize imaging protocols, and integrate imaging biomarkers into clinical decision-making pathways.

There are certain limitations in this study. Firstly, given the limitations of data availability and sample size, certain variables may not be available in all patients, especially molecular subtypes and certain clinical parameters, it needs to be further explored in future studies. And the sample size for the HER2-zero group was small (only 20 patients), which may have resulted in bias in the results involving this group. Consequently, a detailed discussion of the results pertaining to HER2-zero BCs was deemed unfeasible. In future studies, we will endeavour to include a larger sample size for HER2-zero BC. Secondly, this was a retrospective, single-center study, which may be subject to selection bias and limits the generalizability of the findings. Future research should incorporate a prospective, multicenter design to validate the diagnostic value of MRI mapping across broader populations. Thirdly, the ROI delineation in this study employed multiple 2D measurements averaged together. The use of 3D volumetric imaging to measure the volume of interest (VOI) may provide a more accurate and comprehensive view of the tumour, thereby compensating for the influence of internal inhomogeneity. Fourth, although we mentioned the potential influence of calcification on T1 shortening and magnetic field inhomogeneity that may affect T2*Map accuracy, other sources of ITH may also compromise the reliability of mapping parameters. These include cystic degeneration, necrotic tissue, intratumoral hemorrhage, or fibrosis, all of which can lead to local deviations in signal intensity and relaxation time. Although efforts were made to manually avoid such areas during ROI placement—guided by DCE-MRI and mapping morphology—subtle or microscopic heterogeneity may not be fully excluded. This could contribute to measurement variability and warrants further investigation, possibly with volumetric or texture-based analysis in future studies.

Conclusions

In conclusion, our results indicate that the clinical indicators ER, PR, and multiple mapping parameters (T1Map_post, ΔT1Map, ΔT1Map_pre, and T2*Map) have statistical significance in comparison between HER2-low and HER2-positive BCs. The multivariable model consisting of ER, T2* and T1_post exhibited promising potential to non-invasively differentiate BC HER2 status.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of The First Affiliated Hospital of Yangtze University prior to data collection (approval No. LLL202380). Written informed consent was obtained from all participants.

Data Sharing Statement

Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1449/dss

qims-16-04-288-dss.pdf ^{(67.3KB, pdf)}

DOI: 10.21037/qims-2025-1449