As tumors cannot grow beyond 2 mm simply with nutrients supplied through diffusion, angiogenesis becomes a critical process for sustained tumor growth. Angiogenesis is capable of differentiating between malignant and benign tumors and can be used as a discriminating characteristic of aggressiveness [16]. The wall of neovascularity tends to be leaky, and the increased permeability results in early and rapid contrast-enhancement on MRI. Pharmacokinetic parameters are very useful in the characterization of angiogenesis in breast cancer and have been shown to be associated with the spread of breast cancer and patient prognosis.

The THC measured by DOT represents blood volume, which has been reported to have high values in malignant tumors [15, 20, 26]. Therefore, high THC measured by DOT is generally associated with tumors showing contrast-enhancement identified by MRI [27–30]. Several MR-compatible DOT systems have been developed for breast imaging, which has the advantage of improving the quality of reconstructed DOT images by using the morphological information provided by DCE-MRI as a priori information [28–33]. Since tumors were co-registered, the obtained information by DCE-MRI and DOT could be easily compared. The suspicious malignant lesions on DCE-MRI were reported to show higher mean absorption coefficient than benign lesions [32].

DCE-MRI is an established clinical imaging modality for breast cancer. For research, pharmacokinetic analysis is commonly applied to obtain parameters. K^trans is the inflow transfer constant, which is related to the delivery of contrast agent to the tumor through vascular perfusion and permeability, while k_ep is the out-flux transfer rate constant for the contrast agent to diffuse from the extracelluar extravascular space back to the plasma compartment [26]. The signal enhancement ratio measures the washout slope based on signal intensities at three time points, which is also related to perfusion and permeability [34]. These DCE-MRI parameters as well as the THC measured by DOT have been correlated with the histologic microvascular density count [35].

The mean THC results could be affected by tumor heterogeneity and the partial volume effect (i.e. inclusion of normal issues in the measurement) [36]; therefore, in this study we chose to analyze the maximal THC, which was measured as the maximum THC value within the tumor ROI box. A suggested cutoff value of THC for malignancy was 140 μmol/L in a previous report [20], but it was also reported that many malignant tumors could have a lower THC value around 100 μmol/L [36]. In our results, the mean value of the THC was 181.3 μmol/L, comparable with results in the previous report [20]. In the correlation analysis between US-DOT parameters and clinicopathologic characteristics, several poor prognostic biomarkers, including high histologic grade, ER negativity and high Ki-67 expression, were significantly correlated with a high THC. Histologic grade is one of three strongest prognostic determinants, which include LN mets, tumor size and histologic grade [37]. Ki-67 is a marker of cell proliferation including the S-phase fraction, mitotic index and bromodeoxyuridine uptake [38]. High Ki-67 expression has been regarded as a characteristic of more aggressive proliferation as well as neovascularization; it is also associated with a good chance of clinical response to chemotherapy [39, 40]. In this study, the THC in high Ki-67 ≥ 20 % cancers (mean ± SD, 207.61 ± 80.15 μmol/L) was higher than in low Ki-67 < 20 % cancers (mean ± SD, 153.52 ± 64.07 μmol/L). With our results, it could be suggested that breast cancers showing a high THC have poorer prognosis than those with a low THC. There was no difference between THC values in HER-2 positive and HER-2 negative groups (p > 0.05), different from Brown et al. [41] and Choi et al. [21]. In our study population, only 6 cancers were HER-2 positive while 31 were HER-2 negative. The low rate of HER-2 positive cancers was possibly from case selection bias (because both US-DOT and DCE-MRI scans were required), and the insufficient case number might affect the results. We also found a higher THC value in larger tumors (≥2 cm) than in smaller ones (<2 cm), but the difference was not statistically significant.

As the use of breast US-DOT in current clinical practice increases [15, 19, 21, 35, 36], there has been efforts to assess the correlation of THC with parameters measured by other imaging modalities [42, 43]. Zhu et al. compared US-DOT with the color Doppler flow imaging and found that the THC value did not differ significantly in malignancies with or without vascular tissue shown on Doppler flow imaging. Doppler flow imaging is based on detection of blood flow motion with relatively high velocity in large vessels, while optical imaging is mainly sensitive to the capillary blood volume within the tumor, which might explain the disagreement [43]. Similarly, in our study we did not find a high correlation between MR parameters and US-DOT results, presumably because of the different imaging principles on which the two techniques are based, as well as the analysis methods. Since a low molecular weight contrast agent (Gd-DTPA) is used for DCE-MRI, the agent can easily leak from the plasma compartment to the extravascular-extracellular compartment, and it is well known that the DCE kinetics are heavily dependent on the vascular permeability and the distribution volume in the extravascular-extracellular space [22, 26]. For example, DCE-MRI can miss breast cancers showing low angiogenesis such as in low-grade DCIS [44]. In contrast, the THC measured by optical imaging is mainly related to the total blood volume without involvement of vascular permeability or distribution space, therefore the fundamental differences in the contrast mechanism could explain the lack of a high correlation between DCE-MRI and US-DOT parameters.

Understanding the tumor microenvironment for angiogenesis can be complicated, and results obtained using different methods may not be well correlated. For example, although US-DOT, Color Doppler flow imaging, and DCE-MRI are all based vascular properties for measurements, Color Doppler imaging shows no significant correlation with microvessel counts [45], while some DCE-MRI parameters were reported to be correlated with microvessel density but not specifically with VEGF, a potent factor to stimulate angiogenesis [46]. Another major reason leading to the poor correlation of parameters is the high heterogeneous nature of the tumor. In this study, for DCE-MRI we included all enhanced tumor tissues from multiple imaging slices as ROI to evaluate kinetics on MR imaging, therefore, it is more like a “whole tumor analysis” approach. In US-DOT the maximum THC value in the ROI box was obtained and used for analysis, thus it is more like a “hot spot analysis” approach. Therefore, it is unlikely to have a high correlation between parameters obtained from “whole tumor” and “hot spot” analyses. However, it was not possible to do co-registered regional analysis due to the diffuse nature of the optical imaging. Optical imaging is very sensitive to the depth information, and tissues near the sensitive region of optical fibers will have more contribution to the measurement results. For example, as the case illustrated in Fig. 2, while DCE-MRI clearly shows a necrotic/cystic core, THC maps shows an averaged high blood volume, presumably due to the sensitivity to the strongly enhanced tissue near the surface closer to the source and detector fibers.

DCE-MRI parameters have been shown to be associated with poor prognostic factors such as high histologic grade and ER negativity [22]. Nevertheless, there have been inconsistent results due to different case numbers and the study population, e.g. Fernández-Guinea et al. [47]. In our study high Ki-67 expression and triple-negative breast cancers showed higher K^trans than their counterparts with marginal significance, suggesting that more aggressive tumors have a higher angiogenesis as measured by DCE-MRI. This result is consistent with a recent report which showed that the mean K^trans was higher in Ki-67-positive tumors than in Ki-67-negative tumors [48]. For further detailed analysis considering tumor heterogeneity, a histogram or pixel-by-pixel analysis can be considered [49]. However, this type of analysis is not meaningful in US-DOT due to the diffuse nature of the optical imaging methods.

There were some limitations in this study. First, the study population was limited to a small number of patients with malignant tumors who received both US-DOT and DCE-MRI. Since no benign tumors with lower angiogenesis were included in the analysis, the dynamic range was small and less likely to show a good correlation result as published in other studies using a diagnostic population. Second, the DCE-MRI was acquired using a typical clinical protocol with 7 dynamic frames and 66 s temporal resolution. This coarse temporal resolution was not sufficient to obtain vascular volume characteristics at a very early time after contrast injection, which is expected to have a better correlation with THC measured by US-DOT [50]. Also, we did not measure the pre-contrast T1 relaxation time T10, and could not measure the arterial input function from each individual patient. The T10 and the arterial input function may vary between patients, and if these parameters can be accurately measured from each patient and used in the pharmacokinetic model fitting, more precise K^trans and k_ep may be obtained. However, these measurements are difficult to do and not practical in a clinical setting; also variations in the resulting Ktrans and kep values are expected to be small and are not thought to affect the correlation with THC. Nonetheless, the pharmacokinetic analysis obtained using assumed T10 and the population blood curves is a common approach and can yield characteristic K^trans and k_ep that are highly correlated with parameters directly calculated from DCE kinetics. The DCE-MRI was done within 2 weeks after US-DOT. The vascularity of breast tissues is known to vary in different phases of a menstrual cycle; therefore, this may introduce a small variation in 16 pre-menopausal women [51]. However, the vascularity of the tumor is much higher compared to normal tissues, and it is unlikely to change much in 2 weeks. Since we were focusing on tumors, the effect of imaging time differences was expected to be very small.