Breast cancer is one of the most prevalent cancers affecting women globally, with about 2.3 million women diagnosed with the disease and 685,000 deaths in 2020 (1). Patients with primary breast cancer are often treated with neoadjuvant chemotherapy (NACT) to downsize the tumor and increase the probability of breast-conserving surgery (2). A non-invasive imaging technique that can provide information on tumor cellularity and perfusion during treatment would be beneficial, as reduced cellularity and perfusion are promising indicators of patient response to treatment (3).

Patients with breast cancer undergoing NACT often undergo repeated dynamic contrast-enhanced (DCE) MRI scans for treatment monitoring (4). DCE-MRI is a widespread technique that can provide information on tumor perfusion and cellularity through serial MRI scans acquired before and after the injection of a gadolinium-based contrast agent (3). Furthermore, quantitative estimation of perfusion-related parameters of breast tumors, including tumor blood flow (F_b), blood volume fraction (v_b), along with hemodynamic and cellularity-related parameters: capillary permeability–surface area product (PS); interstitial volume fraction (v_e), and extracellular volume fraction (v_d; calculated from the combination of blood volume and interstitial volume fractions) can be achieved by employing a recently developed DCE-MRI technique (5). However, certain safety concerns exist regarding gadolinium administration, particularly in patients with cancer who undergo repeated contrast-enhanced scans (6). Therefore, alternative imaging techniques that can provide equivalent perfusion and cellularity-related measurements without administering a contrast agent are of interest.

Conventional diffusion-weighted imaging (DWI), which is not used generally in breast cancer imaging, can be employed in oncology treatment response monitoring through the apparent diffusion coefficient (ADC). The ADC measures the diffusivity of water molecules in the tissue and is assumed to serve as an indicator of cellular density. As such, as tumor cellularity decreases in response to treatment, the ADC value increases (7). The ADC is therefore expected to be directly proportional to the DCE-MRI measurements of the tissue’s interstitial and extracellular volume fractions. However, few studies have examine this relationship, and one study in breast tumors has challenged the expectation suggesting that the ADC is incompletely understood (8, 9). Also, blood in the microcirculation can contaminate the DWI signal decay, contributing to the ADC value (10). A technique that can potentially address the problems affecting both DCE-MRI and DWI is intravoxel incoherent motion (IVIM), an advanced form of DWI. It has been proposed that IVIM enables simultaneous assessment of tissue diffusion and perfusion by separating the effects of the microcirculation of blood in the capillary network (so-called pseudo-diffusion) from water diffusion in the rest of the tissue. This method requires DWI acquisitions with multiple b-values (low and high) and fits a bi-exponential model to the data to estimate the diffusion-related parameter D_t (tissue diffusion) and perfusion-related parameters, including D_p (the pseudo-diffusion coefficient), f (the perfused fraction), and their product f×D_p (microvascular blood flow) (10, 11).

In the past decade, growing interest in exploring the potential applications of IVIM in breast tumors has produced studies differentiating benign and malignant tumors (12, 13). IVIM perfusion-related parameters have also shown some promise for evaluating breast tumor response to NACT (14–16). This in turn has reopened the question of whether IVIM could be used as a contrast-agent-free alternative to DCE-MRI for measuring breast tumor perfusion. Few studies have investigated the correlations between IVIM and DCE-MRI perfusion-related parameters in breast tumors and have produced contradictory results (17–19). These studies examined correlations at a single visit; however, a correlation between perfusion parameter changes caused by treatment is meaningful and suggests that IVIM could be a contrast-agent-free surrogate to the DCE-MRI method in monitoring serial changes in tumor perfusion. Further, none of these studies provided an absolute estimation of tumor blood flow; they did not perform a direct comparison with the IVIM parameter purported to measure microvascular blood flow (f×D_p).

The primary aim of this study was to investigate whether IVIM and DCE-MRI perfusion-related parameters correlate and whether IVIM can offer a contrast-agent-free alternative to DCE-MRI for monitoring serial changes in tumor perfusion. The DCE-MRI data were analyzed to estimate absolute tumor blood flow, blood volume fraction, capillary permeability–surface area product, interstitial volume fraction, and extracellular volume fraction (5). This study assesses both between-subject and within-subject repeated measures correlations between the perfusion parameters estimated by IVIM and DCE-MRI (specifically perfusion fraction versus blood volume fraction and microvascular blood flow versus blood flow) in a cohort of patients with breast cancer imaged before treatment and after one and three cycles of NACT. Analyzing both correlations is valuable; between-subject correlation reveals the potential for estimating DCE-MRI perfusion parameters using IVIM at a given time, whereas within-subject repeated measures correlations indicate the potential for estimating change in DCE-MRI perfusion parameters using IVIM when assessing longitudinal changes in the same patient. The secondary aim of this study was to examine the correlation between tissue diffusion measures from DWI and DCE-MRI measures of the tissue’s interstitial and extracellular volume fractions. This would improve the understanding of tissue diffusion measures and their changes in response to treatment further, which are of interest for translation into breast cancer imaging as markers of treatment response (20).

In this study, 40 female patients were eligible and enrolled between August 2015 and April 2018 (median age 44.5 (39, 53) years). MRI data were obtained for all patients at baseline and 37 patients after one and three cycles of NACT (three withdrew following baseline MRI). However, the MRI data acquired after three NACT cycles from two patients were excluded from the analysis because no tumor was apparent on their MRI scans. This exclusion resulted in 112 MRI studies with DWI and DCE-MRI acquisitions. Table 1 presents the clinical characteristics of the patients.

Four DCE-MRI scans—two at baseline and two after three NACT cycles—were excluded because one patient could not tolerate the whole imaging protocol, two had technical issues (the back coil was switched on and off sporadically), and one moved during the DCE-MRI acquisition, leaving 108 studies with paired DWI and DCE-MRI data acquisitions ( Figure 2 ). Based on the cAIC results, 75 DCE-MRI data sets were analyzed using the two–compartment exchange model, 20 using the tissue uptake model, and 13 using the one–compartment model.

Smaller regions of interest (cold-spot and hot-spot regions) were generated from 91 out of 108 studies: 34 at the baseline, 33 after one NACT cycle, and 24 after three NACT cycles (some tumors shrank below the 5x5 pixels threshold during NACT). DCE data sets were fitted using the two-compartment exchange/tissue-uptake/one-compartment models for 68/13/10 cold-spot regions and for 72/9/10 hot-spot regions.

For DWI data analysis, there were a number of cases where estimates of the IVIM parameters pseudo-diffusion and perfusion fraction reached one of their limiting values, and these parameters were excluded from the statistical analyses (2 cases from the whole-tumor region, 8 from the cold-spot region, and 4 from the hot-spot region).

Despite the examination of 108 paired DWI and DCE-MRI datasets, no statistically significant between-subject or within-subject repeated measures correlations were found between the IVIM and DCE-MRI perfusion parameters of the study’s primary interest (perfusion fraction versus blood volume fraction and microvascular blood flow versus blood flow). These findings align with previous breast cancer studies, which also found no correlation between any IVIM perfusion parameters and DCE-MRI parameter related to perfusion, K^trans (transfer constant) (18, 19). K^trans may not solely reflect tumor blood flow but also vessel permeability (33). The present study went further by estimating tumor blood flow and blood volume fraction from DCE-MRI during NACT, but still found no correlations. One possible explanation for the lack of correlation might be significant tissue heterogeneity in tumors; the parameters were estimated from the whole-tumor regions. Where possible, two smaller regions of interest (5×5 pixels) in each whole-tumor region were generated to reduce the likelihood of heterogeneity. It was assumed that these smaller regions would be more homogenous. However, no clear correlations were found in these smaller regions, and the data were observed to be more variable than the whole-tumor region, as reflected by the number of outliers and a wider range in the box plot scale ( Figure 4 , Supplementary Figure A1 ). An alternative method for future studies that might aid in selecting homogeneous tumor regions could be histogram analysis of pixel-wise IVIM and DCE-MRI parameter maps; however, the possibility of finding a homogeneous tumor region in the IVIM and DCE perfusion-related parameter maps to examine the correlation would require further investigation and validation.

Imprecision in the estimates of microvascular blood flow and blood volume fraction, in particular, is a potential issue that may have masked correlations between the IVIM and DCE-MRI perfusion parameters. A previous report recognized that the precision with which pseudo-diffusion is estimated is poor (34), and the estimate of blood volume fraction in another study was reported to be very imprecise (35), which was reflected in our calculated upper estimate of its repeatability (Appendix C, Supplementary Material ). The estimation of blood volume fraction, against which the perfusion fraction derived from IVIM is compared, becomes difficult when tumor capillaries are excessively leaky (24). In this study, out of 108 DCE-MRI datasets, a one-compartment model was preferred in 13 cases, and an estimate of blood volume fraction and capillary permeability–surface area product was not possible in those 13.

It is also possible that IVIM and DCE-MRI reflect different underlying physiology. IVIM does not estimate perfusion in a classical way but estimates flow in the direction of the diffusion encoding gradient, whereas DCE-MRI measures the delivery of blood and subsequent distribution of contrast agent in the tissue, on a different time scale (36). Furthermore, it has been suggested that a single pseudo-diffusion coefficient is insufficient to describe the complex diffusion properties of the vascular signal (37). The inconsistent patterns of response to treatment seen in the median values of perfusion fraction versus blood volume fraction and microvascular blood flow versus blood flow may support this suggestion ( Figure 4 ).

In contrast, this study found moderate positive between-subject and within-subject repeated measures correlations between the diffusion parameters (ADC and D_t) and interstitial volume fraction, as well as a moderate positive within-subject repeated measures correlation between the diffusion parameters and extracellular volume fraction. These positive results are important, as this is the first time they have been observed in breast cancer (8), and support the current understanding of these imaging parameters. A positive between-subject correlation between ADC and interstitial volume fraction was previously determined in head and neck cancers (9) suggesting that these parameters are related to tissue microstructure. The ADC and D_t values reflect the diffusion of water molecules in tissue, which is affected by cellular density, membrane permeability and extracellular volume (7), and v_d is a direct measure of the extracellular volume fraction (24) while v_e is a parameter that reflects the volume fraction of the interstitial space within the tissue, which can be influenced by such factors as cellular density and extracellular matrix deposition. A prior study revealed that tumor cellularity is inversely proportional to v_e, v_d, and ADC values (38). Therefore, the observed between-subject correlation of the diffusion coefficients and interstitial space may suggest that breast tumors with a high cellular density tend to have a small interstitium and increased diffusion restriction, whereas tumors with a low cellular density tend to have a large interstitium and less diffusion restriction. The observed positive within-subject repeated measures correlations could result from the fact that ADC/D_t, v_e, and v_d exhibited similar patterns of change in response to treatment, wherein the values were increasing during the three MRI time-points ( Figure 4 ).

Furthermore, a moderate positive between-subject correlation between the diffusion coefficients and tumor T₁ was observed in this study. Tumor T₁ measures tissue relaxation time, which can be affected by tissue water and fat content, macromolecule concentration and hydration state (39). Thus, this positive correlation may be because breast tumors with high cellular density and a small extracellular space have a decreased free-water content, resulting in low diffusion coefficient values and short tumor T₁ (12, 39).

The present study has some limitations. First, this study was performed on a limited sample size using a 1.5 T MRI scanner, which may limit the statistical power of the results. However, this is the first study that assesses both between-subject and within-subject repeated correlations between the perfusion parameters estimated by IVIM and DCE-MRI in a cohort of breast cancer patients undergoing NACT with a primary focus on hypothesis generation rather than testing; therefore, the results can be used to direct future studies. Second, the DWI data were acquired with only 6 b-values, four of which were low (≤ 200 s/mm²). In a clinical protocol, it is not practical to acquire DWI data with a large number of b-values. Nevertheless, the simulation study showed that using 6 b-values will not result in appreciably worse outcomes for most parameters, though the precision of microvascular blood flow was lower than with 12 b-values (details provided in Appendix B, Supplementary Material ). Further, a previous study showed that a small number of b-values is not the main source of errors in IVIM parameter estimates. Intra-patient variability is significant; they found that the precision in the estimates of the IVIM parameters with only 4 b-values was better than the test-retest repeatability of those same parameters estimated with 16 b-values (34). Third, a pixel-wise comparison of IVIM and DCE-MRI parameter maps was not performed in this study, although it might be valuable. Instead, the images were analyzed by following the recommended approaches of the International Breast Diffusion-Weighted Imaging Working Group (22), which included volumetric sampling and focused regions of interest (i.e., smaller single-slice regions on the darkest part of the ADC map). No correlations were observed between perfusion fraction versus blood volume fraction and microvascular blood flow versus blood flow in these smaller regions, but they showed more variability in the estimates instead ( Supplementary Figure A1 ), suggesting that a pixel-wise analysis might yield similar outcomes. Fourth, rigid registration was employed for aligning the DCE and DWI images and this approach may not have been sufficient to correct DWI distortions. As such, the accuracy of spatial co-registration could have been affected, potentially influencing the findings reported, particularly in the smaller regions. Therefore, future work incorporating pixel-wise analysis following rigorous DWI and DCE-MRI image registration is needed to further investigate these relationships. Finally, the repeatability of the DWI and DCE-MRI parameters was not formally investigated. It was challenging to justify performing a repeated baseline DCE-MRI scan that required an additional injection of gadolinium contrast because the patients were due to undergo multiple NACT cycles and MRI scans. Instead, an upper estimate of the repeatability of the DWI and DCE-MRI parameters was calculated from a selection of baseline and cycle 1 studies (Appendix C, Supplementary Material ).

In conclusion, this preliminary study investigated both between-subject and within-subject repeated measures correlations between DWI and DCE-MRI parameters in a cohort of patients with breast cancer imaged before and after one and three cycles of NACT. No statistically significant correlations were observed between the perfusion parameters estimated by IVIM (perfusion fraction and microvascular blood flow) and those estimated by DCE-MRI (blood flow and blood volume fraction). The two techniques may reflect different underlying physiology, and/or estimates of the IVIM and DCE-MRI parameters in the current study are largely imprecise. Therefore, care should be taken when interpreting the IVIM perfusion parameters as surrogates for those measured using DCE-MRI until their underlying pathophysiologic interpretation and relationship to the DCE-MRI perfusion parameters are elucidated by further research. However, the moderate positive within-subject repeated measures correlations found between the diffusion parameters and DCE-MRI measures of the tissue’s interstitial and extracellular volume fractions confirms the expectation that as these volumes increase, water diffusion increases.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by Yorkshire & The Humber - Bradford Leeds Research Ethics Committee (Ethics approval: 15/YH/0246). The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

ZA: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Methodology, Investigation, Formal Analysis, Data curation. SB: Writing – review & editing, Software, Methodology. DW: Writing – review & editing, Supervision, Funding acquisition. NS: Writing – review & editing, Funding acquisition. TD: Writing – review & editing, Supervision. DB: Writing – review & editing, Validation, Supervision, Software, Resources, Project administration, Methodology, Funding acquisition, Data curation, Conceptualization.