Correlation Between Breast Parenchymal Stiffness Measured by 3T Magnetic Resonance Elastography and CT-Derived Hounsfield Units

Introduction

Breast tissue composition and its biomechanical characteristics have gained increasing attention because of their strong associations with imaging appearance and cancer risk. Breast density, reflecting the relative proportions of fibroglandular and fatty tissue, is a well-established independent risk factor for breast carcinoma and influences lesion conspicuity on mammography and magnetic resonance imaging (MRI) [1,2]. However, mammographic density provides only an indirect surrogate of tissue composition, prompting interest in more objective quantitative methods.

Hounsfield units (HU) provide a standardized quantitative measure of tissue X-ray attenuation on computed tomography (CT) and directly reflect tissue density. Fatty breast tissue typically shows negative values, whereas fibroglandular tissue demonstrates positive values.

On non-contrast CT, HU derived from localized parenchymal sampling serves as an objective indicator of breast tissue composition, with higher HU values corresponding to greater fibroglandular content. In recent years, HU has also been increasingly used as a complementary quantitative surrogate for mammographic density [1].

Magnetic resonance elastography (MRE), a phase-contrast MRI technique that quantifies tissue stiffness in kilopascals (kPa), offers direct insight into stromal and extracellular matrix biomechanics [3–5]. While prior breast MRE studies have largely focused on lesion characterization [6–9], the biomechanical properties of normal parenchyma and their relationship with radiologic density metrics such as HU remain insufficiently studied.

The COVID-19 pandemic inadvertently created a unique opportunity to study this correlation. During this period, chest CT scans were performed on a large portion of the population for screening and follow-up purposes, often within short intervals of other imaging modalities. Leveraging this circumstance allowed identification of patients who underwent both breast MRI with MRE and chest CT in close temporal proximity.

The primary aim of this study was to assess the correlation between breast MRE stiffness (kPa) and CT attenuation (HU) of the parenchyma, as an external validation of MRE measurements. The secondary objective was to establish reference stiffness ranges for fatty and fibroglandular tissue. We hypothesized that a higher parenchymal density (HU) corresponds to increased stiffness (kPa), reflecting the underlying fibroglandular content. The use of a 3T MR system with an optimized internal pneumatic driver setup allowed for high-resolution stiffness mapping of the breast, consistent with the contemporary literature [3,6,10].

Material and Methods

ETHICS APPROVAL:

Ethical approval for this study was obtained from the Ethics Committee of İstanbul Nişantaşı University (date: June 14, 2023, No: 2023/24). All procedures were conducted in accordance with the principles of the Declaration of Helsinki and its amendments. Owing to the retrospective nature of the study and the use of anonymized clinical data, the requirement for written informed consent was waived by the institutional ethics committee.

STUDY DESIGN AND POPULATION:

This single-center, retrospective methodological validation study was conducted at the Department of Radiology of the BHT CLINIC Istanbul Tema Hospital. The institutional imaging archive was reviewed to identify patients who underwent routine breast MRI including MRE, between March 2020 and September 2025. In October 2025, all eligible cases were retrospectively screened to determine those who had undergone a non-contrast chest computed tomography (CT) examination within ±1 month of their breast MRE. A total of 48 patients (age range, 29–56 years) fulfilled the inclusion criteria.

The exclusion criteria were: suboptimal image quality, previous mastectomy or breast surgery, presence of a parenchyma-distorting mass lesion, or the use of hormonal therapy known to influence breast tissue composition.

The primary objective of this study was to externally validate MRE-derived parenchymal stiffness (kPa) using CT-derived attenuation values (HU) as a quantitative radiologic indicator of breast tissue density. The secondary objective was to establish reference stiffness ranges for fatty and fibroglandular regions. All MRE stiffness measurements and CT HU assessments were performed retrospectively on a dedicated workstation using the hospital PACS (Picture Archiving and Communication System (PACS). Possible menstrual cycle-related variations in tissue elasticity were considered potential confounding factors; therefore, the ±1-month interval between modalities was acknowledged as a study limitation [11]. The use of CT-based HU as a reliable quantitative surrogate for breast density has been validated in the literature [1].

MRE PROTOCOL AND STIFFNESS MEASUREMENTS:

All examinations were performed using a 3T MRI system (Signa Architect; GE Healthcare, Milwaukee, WI, USA) equipped with a multichannel torso coil. A 19-cm pneumatic internal driver was placed on the anterior chest wall to deliver the vibration.

A spin-echo echo-planar imaging (SE-EPI)-based MRE sequence was applied with a vibration frequency of 60 Hz, 4 phase offsets, and multidirectional sampling. The acquisition parameters included a FOV of 280–320 mm, matrix of approximately 128×128 (interpolated to 256), slice thickness of 3–4 mm, TR/TE of≈2000/50–70 ms, and NEX of 1–2. Parallel imaging was used when it was appropriate. All parameters were selected in accordance with published 3T breast MRE studies and experimental/clinical guidelines [3–6,10].

The wave-field quality was visually confirmed for phase wrapping and signal attenuation. Stiffness maps were generated using curl filtering, wave-field balancing, and Helmholtz-based local frequency estimation inversion. Stiffness-weighted magnitude maps were produced for reference visualization [10,12–15].

The 3T configuration with a sternal or thoracic driver provides high reproducibility and has been validated by Hawley et al [6]. The technical principles and clinical evolution of breast MRE have been described extensively in previous reports [3–5].

Axial sections encompassing the breast parenchyma were retrieved from the hospital’s Digital Imaging and Communications in Medicine (DICOM) – Picture Archiving and Communication Systems (PACS) archive for kPa measurement. For each breast, at least 3 square-shaped regions of interest (ROIs) (≥250–260 mm2) were manually placed within the parenchyma, encompassing both fatty and fibroglandular areas (Figures 1, 2). Vessels, ducts, lobular borders, and peripheral artifacts were carefully avoided. The wave-field quality was visually verified for each ROI [3,6,10].

CT PROTOCOL AND HU MEASUREMENTS:

All CT examinations were performed on a GE Revolution GSI 256-detector Multislice CT scanner (GE Healthcare, Waukesha, WI, USA).

Non-contrast axial acquisitions were used with a typical tube voltage of 120 kVp (range 80–120 kVp depending on body habitus), automatic mAs modulation, slice thickness 1–1.25 mm, and a soft-tissue reconstruction kernel (GE “Standard” or “Soft”). Because HU values may vary with tube voltage and reconstruction kernel [1], protocol matching and calibration checks were performed whenever feasible.

Corresponding parenchymal levels were used to place 3 square-shaped ROIs (approximately 2.50–2.60 cm2 each) within the fatty and fibroglandular regions for HU measurement, carefully avoiding calcifications and vascular structures [1] (Figures 1, 2). Axial sections encompassing the breast parenchyma were retrieved from the hospital’s DICOM–PACS archive for HU measurement.

DATA GATHERING:

Because breast MRI and chest CT are acquired in different patient positions, the resulting images exhibit notable morphological and radiological anatomical differences. To ensure spatial correspondence, all quantitative data were derived from anatomically matched imaging levels across both modalities.

For each patient, bilateral and multiple regions of interest (ROIs) were manually placed within the fibroglandular parenchyma of both breasts, following the same sampling strategy used during stiffness mapping. The median parenchymal stiffness (kPa) from MRE and the corresponding median attenuation (HU) from CT were then computed to yield a single representative value per subject, thereby minimizing intra-individual variability and enhancing cross-modality comparability. This strategy was adopted to minimize intra-subject variability and to provide a representative central stiffness–density value per individual.

To ensure methodological consistency, ROI selection was performed in consensus by 2 radiologists with 12 and 25 years of experience in breast imaging, blinded to each other’s measurements. All extracted values were reviewed for artifacts or sampling errors prior to inclusion in the final dataset.

A subgroup sensitivity analysis was subsequently conducted to evaluate potential time-dependent effects between modalities, comparing patients imaged within ≤14 days and those with an interval of 15–30 days between MRE and CT examinations. This comparison allowed for assessment of measurement reproducibility and the potential influence of temporal separation on parenchymal stability [11].

STATISTICAL ANALYSIS:

Primary Analysis: The primary endpoint was to evaluate the association between MRE-derived breast stiffness (kPa) and CT-based attenuation (HU). Pearson’s and Spearman’s correlation coefficients were calculated to assess linear and monotonic relationships. A multiple linear regression model (kPa=β0+β1·HU+β2·Age) was constructed to adjust for age, and model assumptions (linearity, normality of residuals, multicollinearity) were verified [10,12–15].

Secondary Analysis: Quantitative parameters were compared between fibroglandular and fatty tissues. The Shapiro-Wilk test was used to assess normality, followed by either the independent-samples t test or the Mann-Whitney U test, as appropriate. Tissue-specific reference ranges (2.5th–97.5th percentiles) were calculated for each category [3–6].

Repeatability Analysis: Measurement repeatability was evaluated in a randomly selected subgroup of 15 cases with duplicate MRE acquisitions. The within-subject standard deviation (SD), coefficient of repeatability (CR=1.96×√2×SD_diff), and intra-class correlation coefficient (ICC) were computed to quantify intra-observer reliability [6].

Power and Software: A priori power analysis indicated that for an expected correlation coefficient (r) of 0.6–0.8, the sample size (N=48) provided at least 90% power at α=0.05. All statistical analyses were performed using IBM SPSS Statistics for Windows, Version 26.0 (IBM Corp., Armonk, NY, USA). Additional data visualization and validation analyses were conducted in Python 3.11 (NumPy, pandas, scikit-learn, matplotlib) [2].

Results

A total of 48 patients were included (mean±SD age, 44.3±8.3 years; range, 29–56 years). Tissue composition distribution: fibroglandular, 28 (58%); fatty, 20 (42%) (Table 1).

In the primary analysis, a strong and statistically significant correlation was observed between breast parenchymal stiffness measured by MRE and attenuation values derived from chest CT (Figure 3). The Pearson correlation coefficient was r=0.834 (P<0.001), while the Spearman rank correlation yielded ρ=0.772 (P<0.001), confirming a consistent positive association across both parametric and non-parametric approaches. Multiple linear regression analysis further demonstrated that breast stiffness increased in proportion to CT attenuation and, to a lesser extent, with patient age, following the equation kPa=2.402+0.009·HU+0.005·Age. The HU coefficient (βHU=0.009 kPa per HU; bootstrap 95% CI, 0.007–0.010) remained highly significant, whereas the age coefficient (βAge=0.005 kPa per year; 95% CI, −0.006 to 0.017) showed a minor, non-significant positive trend. The model explained approximately 70.1% (R2=0.701) of the variance in stiffness measurements (Table 2). These findings indicate that higher CT attenuation values correspond to increased MRE-derived stiffness, suggesting that both modalities capture complementary and biologically coherent aspects of breast parenchymal composition and density [1,2].

When analyzed according to tissue composition, distinct differences were observed between fatty and fibroglandular parenchyma. The mean stiffness of fatty tissue was 1.85±0.32 kPa, with corresponding CT attenuation values of −89.7±17.3 HU (Table 1), whereas fibroglandular tissue demonstrated a markedly higher mean stiffness of 2.95±0.43 kPa and attenuation of 34.6±10.8 HU (Figure 4). The distribution-based reference stiffness ranges (2.5th–97.5th percentiles) (Table 3) were 1.30–2.44 kPa for fatty and 2.17–3.77 kPa for fibroglandular parenchyma (Figure 5). The differences between the 2 tissue classes were statistically significant for both stiffness and attenuation parameters (P<0.001).

In the repeatability assessment performed in a random subgroup of 15 cases, the within-subject standard deviation (SD) was 0.055 kPa, with a corresponding coefficient of repeatability (CR) of 0.214 kPa, and an approximate intra-class correlation coefficient (ICC) of 0.35. These reproducibility indices are consistent in magnitude with previously published 3T breast MRE studies utilizing sternal or thoracic drivers, confirming the technical stability of the current setup [13].

Our study is the first to establish preliminary reference MRE stiffness ranges (in kPa) for fatty and fibroglandular breast tissue by directly comparing them with HU values, the radiologic gold standard for tissue density assessment. This provides a quantitative foundation for future diagnostic evaluation and risk-stratification research.

Discussion

In this study, breast parenchymal stiffness measured by 3T MRE showed a strong positive correlation with CT attenuation values (r=0.83, P<0.001), supporting the concept that parenchymal density and stiffness represent converging biomechanical properties. The observed relationship aligns closely with previous studies linking breast density to stiffness and risk [1,2]. Fibroglandular areas demonstrated significantly higher stiffness (mean≈3.0 kPa) compared with fatty tissue (mean≈1.9 kPa), consistent with reported values from Hawley et al [13] and Patel et al [3].

Reported normative MRE values for breast parenchyma vary depending on sequence design, driver configuration, and field strength. At 3 T, Hawley et al [6] found mean stiffness values around 2.5–3.5 kPa for fibroglandular regions, while Lorenzen et al [11] demonstrated menstrual-cycle-dependent variations up to 15%. The present findings fit within these established ranges, suggesting that the internal pneumatic driver setup used in our system achieved adequate wave propagation and signal-to-noise ratio [4–6].

Recent developments, such as gravitational or portable low-field transducers, may further broaden accessibility and comfort in breast MRE [16,17]. Our correlation between HU and stiffness adds to this technological evolution, indicating that routine non-contrast CT could provide a complementary radiologic biomarker for tissue mechanical state, particularly in patients already undergoing chest imaging.

From a physiologic perspective, fibroglandular tissue exhibits higher stiffness due to its greater collagen, ductal, and stromal content, while adipose tissue remains compliant. The positive age trend observed (β_age=0.005 kPa/year) reflects age-related stromal remodeling and relative glandular involution. These biomechanical features mirror the inverse association between breast density and fat fraction observed in CT and MRI studies [1,2,11].

Technically, the close HU–kPa relationship supports the accuracy and quantitative reliability of breast MRE when performed under controlled conditions. The use of 3T imaging and curl-filtered inversion algorithms provided a stable reconstruction pipeline [10,12–15]. Nevertheless, small measurement variances (within-subject SD≈0.055 kPa, CR≈0.21 kPa) show the need for standardized driver positioning and phase synchronization across institutions [13].

Compared with mammography and ultrasound, breast MRE provides a radiation-free, operator-independent, and fully quantitative assessment of parenchymal biomechanics that cannot be obtained with conventional modalities [3,4,6]. Mammography offers only a two-dimensional estimation of density and demonstrates reduced performance in women with dense breasts [1], whereas ultrasound elastography is inherently operator-dependent and limited in evaluating deep or heterogeneous parenchymal regions [8,9]. In our cohort, the strong correlation between MRE-derived stiffness and CT-based HU further supports the biological coherence of MRE as an objective biomarker of tissue composition [3,4,6]. Although not intended to replace routine breast imaging, MRE provides complementary value particularly in dense breast tissue and in clinical situations requiring reliable biomechanical characterization [2,3].

Quantitative stiffness mapping of normal parenchyma could enhance baseline assessment before interventional or oncologic therapies, aid in understanding background parenchymal enhancement (BPE) variability, and potentially refine risk models that currently rely solely on mammographic or volumetric density [2,8,18]. In clinical practice, breast MRE does not replace mammography or ultrasound, which remain the first-line modalities for routine evaluation. Instead, MRE serves as a complementary technique, particularly in individuals with high breast density, in younger patients or those for whom radiation should be minimized, during pregnancy as a radiation-free alternative, and in scenarios where treatment-related stromal changes or fibrosis need to be monitored. Although not a screening tool, MRE can provide meaningful added value in selected clinical situations by quantifying the biomechanical properties of breast parenchyma and improving the interpretive context of conventional imaging [2,11,18]. Furthermore, integrating MRE with other quantitative MRI or multiparametric techniques (eg, diffusion or DCE-MRI) may offer synergistic diagnostic benefits [18,19].

Although breast MRE is not part of routine clinical imaging, its ability to quantify parenchymal biomechanics offers advantages over mammography and ultrasound, which assess tissue density only indirectly. As a non-invasive, radiation-free, and operator-independent method, MRE provides direct information about stromal composition, collagen content, and extracellular matrix features that relate to cancer risk and treatment response. Given its non-invasive nature and operator-independent biomechanical assessment, MRE holds potential as a future biomarker for breast tissue monitoring, even at a screening level. The strong correlation between HU and MRE-derived stiffness further supports the biological coherence between densitometric and biomechanical properties, suggesting potential usefulness in risk stratification, monitoring therapy-related fibrosis, assessing background parenchymal enhancement, and guiding individualized follow-up. Thus, techniques that directly measure elasticity and density may complement conventional breast imaging in selected clinical settings [2,11,18,20,21]. These findings may contribute to future refinements in breast cancer risk stratification strategies and potentially influence guidelines concerning breast tissue characterization

Compared with mammography and ultrasound, breast MRE offers a more comprehensive characterization of tissue properties. Mammography provides a two-dimensional density assessment and uses ionizing radiation, with reduced sensitivity in dense breasts, while ultrasound can measure elasticity but remains operator-dependent and anatomically limited. MRE, by contrast, is radiation-free, operator-independent, and generates three-dimensional quantitative stiffness maps that reflect stromal composition, collagen content, and extracellular matrix features. The strong correlation between HU and MRE-derived stiffness in our study further supports the biological concordance between tissue density and biomechanical behavior, reinforcing the clinical relevance of MRE [2,11,18,22].

The main limitations of this study are: (1) the small cohort size (n=48), constrained by the rare overlap of MRE and CT; (2) the possible influence of the menstrual cycle on parenchymal stiffness despite the ±1-month imaging window [11]; and (3) minor variations in CT acquisition parameters (kVp, kernel) that can affect HU reproducibility [1]. Additionally, breast MRI and chest CT are performed in different patient positions, and the resulting images may not appear morphologically or anatomically identical, which is another inherent limitation of cross-modality comparison. Future studies with larger, prospectively synchronized cohorts are warranted to validate and extend these findings. Future studies with larger, prospectively cycle-matched cohorts are needed to validate these preliminary findings, minimize hormonally driven variability, and integrate multimodal comparisons – including mammography, shear-wave elastography, and quantitative MRI biomarkers – to strengthen the biological relevance of breast tissue stiffness assessment.

Emerging reconstruction methods and machine-learning-based stiffness prediction models [20–22] may soon permit direct biomechanical quantification from standard MR or CT datasets. Moreover, low-field MRE prototypes [17] could make stiffness imaging more accessible, enabling screening-level applications in breast health monitoring.

Conclusions

Breast parenchymal stiffness measured by 3T MRE strongly correlates with CT-derived HU, confirming that both parameters reflect complementary aspects of tissue composition. The established normal stiffness ranges (1.3–2.4 kPa for fatty, 2.2–3.8 kPa for fibroglandular tissue) provide a quantitative baseline for future diagnostic and risk-stratification studies. These results reinforce the potential of MRE as a reliable, non-invasive biomarker of breast tissue biomechanics and underscore the value of integrating elastographic and densitometric information in comprehensive breast imaging.