Correlation analysis of multiparametric magnetic resonance imaging features and molecular subtypes of breast cancer

INTRODUCTION

Breast cancer is the leading cause of cancer-related mortality among women worldwide and remains the most frequently diagnosed malignant tumor. Alarmingly, there has been a noticeable trend toward earlier ages of onset in recent years.^[1] Breast cancer is a heterogeneous disease characterized by diverse phenotypes, each requiring tailored therapeutic strategies and exhibiting distinct prognoses.^[2-4] According to the St. Gallen 2013 consensus,^[5] breast cancer is classified into four molecular subtypes: Luminal A, Luminal B, human epidermal growth factor receptor 2 (HER-2+) overexpression, and triple-negative breast cancer (TNBC).

Diverse molecular subtypes of breast cancer vary in disease mode, responses to treatment, prognosis, and survival rates. The Luminal A subtype is characterized by well-differentiated tumor cells and carries the most favorable prognosis, with the lowest rates of local recurrence and metastasis. The Luminal B tumors are less differentiated, exhibit lower levels of estrogen receptor (ER) and progesterone receptor (PR) expression, and are associated with a worse prognosis compared to Luminal A. Both luminal subtypes commonly metastasize to the bone and demonstrate sensitive to hormone therapy. HER-2 overexpression subtype typically has moderate to high nuclear grade and is most common associated with liver and brain metastases. This subtype is most effectively treated with HER-2-targeted therapies. TNBC, on the other hand, is distinguished by its aggressive nature, poor prognosis, high risk of early recurrence, and frequent distant organ metastasis, particularly to the lungs and brain. Chemotherapy is the main treatment for TNBC.^[6,7] While selective biopsy tissue samples have been evaluated for the histopathological characteristics of tumor tissues, this often represents only a portion of the heterogeneous tumor, and this limited sampling may not be sufficient to capture the tumor heterogeneity that impacts tumor progression and treatment processes.^[8]

Breast multiparametric magnetic resonance imaging (MRI) is the most accurate and sensitive diagnostic imaging technology for detecting breast cancer, widely used in breast cancer screening, differentiation between benign and malignant breast lesions, pre-operative staging of breast cancer, pre-operative or post-operative evaluation of breast-conserving surgery, and assessment after neoadjuvant chemotherapy. MRI has fine tissue resolution, allowing us to observe the morphological and functional characteristics of the entire tumor in vivo and predict tumor molecular subtypes through imaging examinations. This may provide prominent contributions to the development of early treatment plans and understanding of prognosis in clinical practice.^[3,9,10] Previous extensive research had investigated the role of T2-weighted imaging (T2WI) imaging features, the apparent diffusion coefficient (ADC), or dynamic contrast enhancement (DCE) in predicting breast cancer molecular subtypes.^[10-14] Nevertheless, there is currently no definitive consensus on the MRI imaging features of the various molecular subtypes of breast cancer.

The aim of our study is to evaluate the relationship between breast multiparametric MRI features (T2WI, ADC value, and DCE) imaging features and different molecular subtypes of breast cancer. The significance of this study is to predict the non-invasive molecular subtypes of pre-operative breast cancer, so as to reduce unnecessary needle biopsy, and guide clinicians to choose the most appropriate treatment.

MATERIAL AND METHODS

RESULTS

Comparison of MRI morphological and signal features of different subtypes of breast cancer

Table 3 summarizes the MRI features of breast cancer molecular subtypes, with 93 cases showing masses (69.4%) and 41 cases showing non-mass enhancement (30.6%). No statistically significant association was observed between molecular subtypes and lesion morphological types (P = 0.061); however, non-mass enhancement patterns with HER-2 (+) [Figure 1] were the most common, accounting for 61.5% of cases.

Table 3: MRI imaging features of 134 cases with different molecular subtype.

MRI imaging feature	Luminal A (n=22)		Luminal B (n=82)		HER-2 (+) (n=13)		TNBC (n=17)		P-value
Tumor location (%)
Right	12 (54.5)		46 (56.1)		8 61.5)		11 (64.7)		0.711
Left	10 (45.5)		35 (42.7)		5 (38.5)		5 (29.4)
Bilateral	0 (0)		1 (1.2)		0 (0)		1 (5.9)
Lesion Type (%)
Mass	16 (72.7)		58 (70.7)		5 (38.5)		14 (82.4)		0.061
NME	6 (27.3)		24 (29.3)		8 (61.5)		3 (17.6)
Tumor number (%)
Unifocal	19 (86.4)		58 (70.7)		12 (92.3)		14 (82.4)		0.181
Multifocal	3 (13.6)		24 (29.3)		1 (7.7)		3 (17.6)
Size	2.3±1.7		2.9±1.9		3.8±1.8		2.6±1.6		0.137
T2 intensity (%)
Hypointense	6 (27.3)		24 (29.3)		1 (7.7)		2 (11.8)		0.014
Isointense	15 (68.2)		38 (46.3)		9 (69.2)		6 (35.3)
Hyperintense	1 (4.5)		20 (24.4)		3 (23.1)		9 (52.9)
ADC value	0.72±0.18		0.70±0.13		0.98±0.23		0.78±0.19		<0.001
Early strengthening rate	206.7±75.6		165.6±51.5		157.5±48.0		188.5±45.8		0.009
Kinetic curve pattern (%)
I	0 (0)		7 (8.5)		3 (23.1)		0 (0)		0.020
II	2 (9.1)		22 (26.8)		5 (38.5)		3 (17.6)
III	20 (90.9)		53 (64.6)		5 (38.5)		14 (82.4)
Mass		(n=16)		(n=58)		(n=5)		(n=14)
Shape (%)
Regular	5 (31.3)		19 (32.8)		1 (20)		10 (71.4)		0.039
Irregular	11 (68.7)		39 (67.2)		4 (80)		4 (28.6)
Margin (%)
Circumscribed	3 (18.8)		15 (25.9)		1 (20)		12 (85.7)		<0.001
Uneven/spiculated	13 (81.2)		43 (74.1)		4 (80)		2 (14.3)
Enhancement characteristic (%)
Homogeneous	4 (25)		6 (10.3)		2 (40)		1 (7.1)		0.043
Heterogeneous	9 (56.3)		41 (70.7)		1 (20)		6 (42.9)
Rim	3 (18.8)		11 (19.0)		2 (40)		7 (50)
NME		(n=6)		(n=24)		(n=8)		(n=3)
Distribution (%)
Focal	0 (0)		4 (16.7)		0 (0)		2 (66.7)		0.080
Linear	2 (33.3)		1 (4.2)		0 (0)		0 (0)
Segmental	2 (33.3)		13 (54.2)		7 (87.5)		0 (0)
Regional	2 (33.3)		3 (12.5)		0 (0)		1 (33.3)
M regions	0 (0)		2 (8.3)		1 (12.5)		0 (0)
Diffuse	0 (0)		1 (4.2)		0 (0)		0 (0)
Enhancement Patterns (%)
Homogeneous	1 (16.7)		1 (4.2)		1 (12.5)		0 (0)		0.301
Heterogenous	1 (16.7)		3 (12.5)		0 (0)		2 (66.7)
Clumped	2 (33.3)		12 (50)		3 (37.5)		0 (0)
Cluster ring	2 (33.3)		8 (33.3)		4 (50)		1 (33.3)

MRI: Magnetic resonance imaging, ADC: Apparent diffusion coefficient, NME: Non-mass enhancement, HER-2: Human epidermal growth factor receptor-2, TNBC: Triple-negative breast cancer

Close

Figure 1:

A 40-year-old woman diagnosed with invasive ductal carcinoma of the human epidermal growth factor receptor-2 subtype. (a) The lesion presents isointense on T2-weighted imaging (white arrow), (b) uneven high signal on diffusion-weighted imaging (white arrow), (c) corresponding low signal on apparent diffusion coefficient (ADC) mapping (white arrow), the average ADC value is 1.31×10^-3mm²/s. (d) The initial phase of dynamic contrast enhancement shows a segmental clustered ring enhancement (white arrow), (e) the kinetic curve (white arrow) demonstrates a persistent appearance (type I curve). (T2WI: T2-weighted imaging, DWI: Diffusion-weighted imaging, DCE: Dynamic contrast enhancement)

Export to PPT

Statistical analysis revealed significant variations in morphological characteristics of mass lesions across molecular subtypes, including shape (P = 0.039), margin (P < 0.001), internal enhancement pattern (P = 0.043), and T2WI signal intensity (P = 0.014). Luminal A [Figure 2] and lumina B [Figure 3] subtypes typically demonstrated irregular shapes, unclear or spiculated margins, heterogeneous enhancement on DCE, and predominantly heterogeneous low or iso-signal on T2WI, while TNBC [Figure 4] frequently appeared as round or regular shapes (71.4%) with clear margins, predominantly ring enhancement on DCE (50%), and heterogeneous high signal on T2WI (52.9%). T2 signal homogeneity analysis showed significant disparities: Luminal A subtype (68.2%) and HER-2 (+) subtypes (69.2%) exhibited higher rates of homogeneous isointense signals compared to Luminal B (46.3%) and TNBC (35.3%) (P = 0.014). Non-mass enhancement features predominantly displayed lobular segmental distribution with clumped or clustered ring enhancement patterns, though these differences did not reach statistical significance across subtypes (P = 0.080; P = 0.301, respectively). Multifocal/multicentric or bilateral disease occurred in 23.1% of cases, with Luminal B representing the most frequent subtype (29.3%, P = 0.181). While no significant correlation emerged between tumor size and molecular subtypes (P = 0.137), dimensional analysis revealed that HER-2 (+) tumors had the largest mean diameter (3.8 ± 1.8 cm) compared to Luminal A (2.3 ± 1.7 cm).

Close

Figure 2:

A 45-year-old woman diagnosed with invasive ductal carcinoma of the Luminal A subtype. (a) The lesion presents uneven low signal on T2-weighted imaging (white arrow), (b) uneven high signal on diffusion-weighted imaging (white arrow), (c) corresponding low signal on apparent diffusion coefficient (ADC) mapping (white arrow), the average ADC value is 0.71×10^-3mm²/s. (d) The initial phase of dynamic contrast enhancement shows a heterogeneous internal enhanced mass with irregular shape and spiculated margin (white arrow), (e) the kinetic curve (white arrow) demonstrates a plateau appearance (type II curve). (T2WI: T2-weighted imaging, DWI: Diffusion-weighted imaging, DCE: Ddynamic contrast enhancement)

Export to PPT

Close

Figure 3:

A 52-year-old woman diagnosed with invasive ductal carcinoma of the Luminal B subtype. (a) The lesion presents uneven low signal on T2-weighted imaging (white arrow), (b) uneven high signal on diffusion-weighted imaging (white arrow), (c) corresponding low signal on apparent diffusion coefficient (ADC) mapping (white arrow), the average ADC value is 0.72×10^-3mm²/s. (d) The initial phase of dynamic contrast enhancement shows a heterogeneous internal enhanced mass with irregular shape and spiculated margin (white arrow), and a daughter lesion in front of the mass (red arrow), (e) the kinetic curve (white arrow) demonstrates a wash out appearance (type III curve). (T2WI: T2-weighted imaging, DWI: Diffusion-weighted imaging, DCE: Dynamic contrast enhancement)

Export to PPT

Close

Figure 4:

A 35-year-old woman diagnosed with invasive ductal carcinoma of the triple-negative breast cancer. (a) The lesion presents uneven high signal on T2-weighted imaging (white arrow), (b) circular high signal on diffusion-weighted imaging (white arrow), (c) corresponding circular low signal on apparent diffusion coefficient (ADC) mapping (white arrow), the average ADC value of the solid part is 0.84×10^-3mm²/s. (d) The initial phase of dynamic contrast enhancement shows a round regular ring enhanced mass with smooth margin (white arrow), (e) the kinetic curve (white arrow) demonstrates a plateau appearance (type II curve). (T2WI: T2-weighted imaging, DWI: Diffusion-weighted imaging, DCE: Dynamic contrast enhancement)

Export to PPT

Comparison ADC value of breast cancer in different molecular subtypes

Quantitative ADC analysis revealed statistically significant differences in mean ADC values across molecular subtypes (P < 0.001) [Figure 5]. The mean ADC values of HER-2 overexpressed breast cancer (0.98 ± 0.23 ×10⁻³ mm²/s) were higher than those of the other three subtypes, and those of luminal types were lower than non-luminal groups.

Close

Figure 5:

(a and b) Box plots show the differences of apparent diffusion coefficient (ADC) value and dynamic contrast enhancement early enhancement rate for the four molecular subtypes (P < 0.001; P =0.009). The mean ADC values of human epidermal grouth factor receptor-2 (HER-2) overexpressed breast cancer (0.98 ± 0.23 ×10^-3 mm²/s) were higher than those of the other three subtypes, and those of Luminal types were lower than non-Luminal groups. Quantitative analysis revealed Luminal A tumors demonstrated the highest mean early-phase enhancement rate (206.7% ± 75.6), contrasting with HER-2(+) subtypes which showed the lowest enhancement (157.5% ± 48.0).

Export to PPT

DISCUSSION

Breast cancer exhibits diverse MRI features, with mass and non-mass enhancement being the two primary presentations. Our findings confirm that mass enhancement is more commonly observed, which is consistent with previous research.^[7,9,16-19] Notably, our study revealed that HER-2 overexpression was most frequently associated with non-mass enhancement, while TNBC predominated in mass type.

The MRI features identified in our study can be linked to underlying biological mechanisms and carry significant clinical implications. For instance, HER-2-overexpressing breast cancers typically exhibit a non-mass, comedo-like morphology with intraductal growth, which contains a higher proportion of DCIS compared to other subtypes. Moreover, the non-mass enhancement observed in HER-2-overexpressing tumors, often characterized by segmental or clustered ring enhancement, may reflect the tumor’s microinvasive growth pattern and ductal spread. This pattern is often associated with necrotic calcifications, which can be identified on mammography.^{[7,10,18,20-22]} These insights may inform targeted biopsy strategies by guiding attention to regions exhibiting these specific MRI features, ultimately enhancing diagnostic accuracy and personalizing patient management.

Similarly, the regular shape and well-defined margins of TNBC lesions observed on MRI may indicate a rapid growth pattern characterized by expansive pushing rather than infiltrative behavior. This distinctive feature could aid in differentiating TNBC from other subtypes, potentially influencing treatment decisions and prognostic assessments. Kazama et al.^[23] reported that TNBC types are more frequently presented as masses, tend to be solitary, and have a larger diameter. Researchers such as Megumi have proposed a negative correlation between tumor roundness and ER expression, while observing a positive correlation with the Ki-67 index.^[24] These findings align closely with the results obtained in our study.

Domestic and international studies have consistently shown that irregular shapes and irregular or speculated margins are more common in luminal types with better prognosis,^{[3,9,16,25-28]} Our study similarly corroborates these findings. The MRI imaging characteristics of luminal tumors are linked to slower tumor growth and earlier disease staging, suggesting a less aggressive and invasive biological behavior. This may be attributed to connective tissue hyperplasia, which acts as a physical barrier limiting tumor expansion.^[9,29] Furthermore, research by Lamb et al.^[30] has demonstrated a correlation between spiculated signs and the expression of ER/PR genes, underscoring the significant prognostic value of MRI features in predicting tumor behavior and clinical outcomes.

TNBC showed ring enhancement in DCE and uneven hyperintense on T2WI, in contrast, luminal subtypes predominantly demonstrated heterogeneous enhancement with predominantly hypointense or isointense signals on T2WI, findings that are consistent with prior research.^{[3,10,29,31,32]} The magnetic resonance features of TNBC are associated with its “basal-like” composition,^[10] characterized by high marginal microvascular density at the tumor periphery. This leads to hypoxia and necrosis during rapid tumor growth,^[33] which manifests as ring enhancement patterns and heterogeneous high signals on T2WI due to necrotic and cystic changes within the tumor. On the other hand, luminal lesions displayed heterogeneous enhancement with relatively low-signal intensity on T2WI, corresponding to fibrous scar formation within the tumor. This reflects hypoxia-driven stromal formation, compared to tumors with homogeneous signal characteristics, the presence of fibrotic foci in the lesion center may indicate a worse prognosis, suggesting that these imaging features could serve as potential biomarkers for assessing tumor behavior and clinical outcomes.^[10]

The correlation between molecular subtypes of breast cancer and TIC remains a controversial in both domestic and international research. Some scholars argue that there is no significant association between TIC curves and molecular subtypes.^[7,26,29] However, our study reveals statistically significant differences in the early enhancement rate and TIC types among different subtypes. Luminal A subtype exhibited the highest average early enhancement rate, while HER-2 (+) type demonstrated the lowest. The washout pattern observed in Luminal A and TNBC may result from shorter mean transit times and higher outflow rates, indicating faster contrast agent clearance due to more permeable vasculature or enhanced lymphatic drainage.^[24] In contrast, the HER-2 (+) subtype predominantly exhibits non-mass enhancement, influenced by the surrounding normal breast tissue, leading to Type I or Type II TIC curves. This suggests slower contrast uptake and retention, reflecting the unique vascular architecture and microenvironment of HER-2-overexpressing tumors. These MRI-based insights into TIC patterns and molecular subtypes hold significant potential for guiding clinical decision-making. For instance, the distinct TIC profiles of different subtypes can inform targeted biopsy strategies, allowing clinicians to focus on regions with specific enhancement patterns that correlate with aggressive tumor behavior.

Our study confirmed that breast cancers with HER-2 overexpression had higher mean ADC values compared to the other three subtypes, consistent with prior findings.^{[11,12,26,34]} This may result from HER-2 gene amplification and high expression, which drive active cell proliferation and increased tumor angiogenesis. However, the incomplete vascular walls in these tumors lead to high permeability and enhanced water molecule diffusion. In addition, HER-2 overexpression often presents as non-mass enhancement, influenced by surrounding normal breast tissue, further contributing to higher ADC values. In contrast, our results showed lower ADC values in luminal types compared to HER-2(+) and TNBC, aligning with Moradi et al.^[26] This difference likely reflects distinct biological behaviors, such as the fibrous stromal components and slower water diffusion seen in luminal tumors due to their less aggressive growth patterns. These MRI insights into ADC values have important clinical implications. Higher ADC values in HER-2 overexpression suggest less invasive behavior, potentially guiding less extensive surgery or targeted therapies. Lower ADC values in luminal types indicate denser tumor structures, influencing biopsy strategies by highlighting areas with restricted diffusion that may harbor more aggressive regions. Furthermore, understanding these differences can aid in treatment planning, such as selecting appropriate neoadjuvant chemotherapy or predicting endocrine therapy responses in luminal subtypes.

Literature has shown that multifocality/multicentricity correlates with Luminal B and HER-2(+) subtypes.^[3,7,35] While our study did not find statistical difference, multifocal/multicentric breast cancer was mostly associated with Luminal B subtype.

This study has several limitations that warrant consideration. First, the retrospective design introduces potential biases, including inter-reader variability in MRI interpretation. The reliance on subjective radiologist assessments of MRI characteristics may compromise diagnostic reproducibility, potentially impacting the reliability of our observations. Second, the modest cohort size coupled with disproportionate representation across molecular subtypes raises concerns regarding statistical robustness and broader clinical generalizability. We propose that subsequent research employ prospective multicenter studies with expanded sample sizes, thereby enhancing statistical power through balanced subgroup stratification. Of particular note, the current findings lack external validation through independent cohorts, which restrict comprehensive evaluation of both technical reproducibility and translational potential. Our research group intends to pursue collaborative validation initiatives across multiple institutions to strengthen the clinical relevance of these preliminary results.

CONCLUSION

Multiparametric MRI features, particularly ADC values, DCE kinetics, and T2WI signals, demonstrate significant associations with breast cancer molecular subtypes. These imaging biomarkers offer potential for non-invasive subtype prediction, supporting more tailored diagnostic and treatment strategies.