About 30% of all cancers among women are breast cancer, in the United States in 2019 [1]. Developing a biomarker that predicts tumor-treatment efficacy is of paramount interest to enhance quality of life, and to reduce medical expense. The conventional method in clinical settings is to detect the morphological variation by using traditional imaging modalities - magnetic resonance (MR) imaging (MRI), ultrasound imaging (US), and computed tomography (CT) - or palpation. However, it is said that tumor metabolic response precedes morphological changes [2]. Therefore, several indices (tumor oxygenation [2-5], tumor blood flow [2, 5], tumor vessel density [6], tumor pH [7], tissue optical index [2, 5, 8, 9], and mammary metabolic rate of oxygen [2, 5, 9]) have been studied to monitor tumor metabolic responses during treatment. Among them, measurements of blood flow provide insight about oxygen delivery and the clearance of metabolic by-products in tumor biology [2]. Therefore, novel approaches in medical imaging to assess blood flow have been recently introduced (e.g. Doppler ultrasound [10], MR angiography [11], and positron emission tomography (PET)/CT [12]). Among optical instrumentations, laser Doppler flowmetry (LDF) [13], diffuse correlation spectroscopy (DCS) [14], and Doppler optical coherence tomography (OCT) [15] have been increasingly evolved to monitor blood flow and perfusion dynamics.

Compared to the several techniques above, laser speckle flowmetry (LSF) has some advantages, in that it can provide noninvasive, label-free, real-time [16], cost-effective, high-spatial-resolution [17], and wide-field dynamics of blood flow in noncontact fashion. It has promise for quantification of blood flow velocity [18], for diagnosis of vascular-related diseases (e.g. cancer, stroke, skin, or peripheral artery disease), and for continuous monitoring and evaluation of therapeutic effects.

Meanwhile, there have been a few basic studies seeking a biomarker for early prediction of treatment efficacy of tumor in preclinical trials. S. Lee et al. demonstrated that the changes in tumor oxyhemoglobin during a breath-hold respond one day earlier than the tumor shrinkage during chemotherapy, by using near-infrared spectroscopy (NIRS) [4]. This methodology is simple and safe, so it has the potential to be translated into clinical practice. However, NIRS cannot reveal spatial heterogeneity of tumor blood volume and tissue oxygenation. Additionally, the outcome of the measurement is subject to operator-dependent variability, due to its probe contact pressure. H. Kim et al. showed that variations in tumor vascular morphometries (i.e. vessel skeleton density and fractal dimension) respond one day earlier than tumor shrinkage, by using OCT angiography (OCTA) [6]. This approach could provide high-spatial-resolution vascular morphology. However, OCTA has some disadvantages, being expensive and complex, with slow data acquisition and a relatively small field of view (FOV). Moreover, blood flow rate in a tumor could not be detected in the OCTA study above.

To address the limitations of the previous modalities, in this study we introduce LSF and report that significant changes in tumor blood-flow parameters (baseline SFI and normalized SFI) based on LSF occur one day earlier than the tumor shrinkage during chemotherapy. To the best of our knowledge, this is the first study that has monitored the longitudinal changes in tumor blood flow during tumor growth and chemotherapy through LSF. The main objective of this paper is to investigate the correlation between tumor volume and tumor blood flow for tumor-growth assessment using LSF.

A trans-illuminated LSF (TILSF) system was fabricated in a free-space configuration to collect raw speckle images from a dorsal skin-fold window chamber (DWC) model with a long-coherence-length laser (DL785-100-SO, Crysta-Laser, USA; λ_c: 785 nm, coherence length: >5 m, optical power: 88 mW). The laser light was directed to the collimating lens (LA1951-B, Thorlabs, USA) to create a collimated beam of diameter ~6.5 mm, is then passed through a top-hat engineered diffuser (ED-1-C20-MD, Thorlabs, USA), and then the diffused light entered the sample. Through an objective (56-988, Edmund optics, USA; Mag. 1×, N.A. ~0.025) and an extension tube (56-992, Edmund optics, USA; length 152.5 mm), the forward-scattered speckle images were captured by a CCD camera (CoolSNAP MYO, Teledyne Photometrics, USA; 2/3” sensor, pixel size: 1940 × 1460, pixel pitch: 4.54 µm × 4.54 µm) for a field of view (FOV) of 8.8 mm × 6.6 mm. Therefore, the camera and objective were chosen to satisfy the Nyquist sampling criterion [19] of at least two pixels per speckle pattern. A neutral-density filter (NDF) was used to avoid saturation of the scattered light’s intensity, and a polarizer and analyzer pair (LPNIRE100-B, Thorlabs, USA) was inserted to minimize specular reflection. Raw speckle images were recorded at 0.5 frames per second using the µManager software [20]. Here the speed of data collection was purposely reduced, to efficiently store large amounts of data without affecting biological analysis for the longitudinal study. To acquire speckle-contrast images even for microvessels having relatively slow speed flow, the exposure time of the camera was empirically set to 10 ms by taking account of the exactly analytical blood flow rate (BFR) in a specific velocity range [21, 22].

A DWC model was applied to balb/c/nu (female, ~8 weeks old, body weight: 22 g) mice to model ectopic breast cancer [6]. The animals were divided into three groups (six for the normal, ten for the treated, and six for the untreated tumor) to find a correlation between tumor blood flow and tumor volume. All of the animals were cared for in individual cages with food and water ad libitum during the experiment. Around 200,000 13762-MAT-B-III (CRL-1666, ATCC, Manassas, VA) rat breast-cancer cells in 20 µL of McCoy’s 5A medium (ATCC, Manassas, VA) were inoculated between the fascial layer and the dermis for the treated and untreated tumor mice, while the normal mouse received the same amount of McCoy’s 5A medium without tumor cells. The tumor size was measured by its horizontal (W) and vertical (L) lengths from the epithelial layer within the window chamber by a Vernier caliper, and then was converted to the tumor volume (V) according to the following equation: [23]. All data are presented as the mean ± standard error of the mean (SEM). The procedures in this study were approved by the Institutional Animal Care and Use Committee of the Gwangju Institute of Science and Technology.

The LSF raw-data recording for the dorsal skin-fold vasculature was initiated from the second day after DWC surgery. The LSF image was taken on a customized imaging mount to reduce motion artifacts [6]. The animal was anesthetized for imaging with 1.5% isoflurane mixed-gas intervention (200 sccm flow rate) according to the protocol shown in Fig. 1. The inhaled gas was initially set to normoxia for fifteen minutes to stabilize the physiological state, then switched to hypercapnia for five minutes, and finally returned to normoxia for ten minutes (Fig. 1). The physiological states (heart rate and arterial oxygen saturation) were recorded on the lower abdomen by a mouse pulse oximeter (MOUSEOX, STARR, USA). End tidal CO₂ (E_tCO₂), fraction of inspired oxygen (FiO₂), isoflurane, and minimum alveolar concentration (MAC) were measured continuously by a gas monitor (B40, GE Healthcare, UK) and transferred to a PC via data-transmission software (Datex-Ohmeda S/5 Collect, GE Healthcare, USA) for analysis. The body temperature was maintained at ~38°C with a heating pad (SRFG-203/10-P, OMEGA, USA) attached to the imaging mount. For the treated tumor group, cyclophosphamide (40 mg/kg body weight) was given to the animal intraperitoneally every other day eight times, while the untreated tumor group received the same amount of distilled water.

Therapeutic initiation was determined by monitoring the tumor growth of over 20 mm³. The LSF raw data were recorded every day until seventeen days for the normal, twenty-one days for the treated, and fourteen days for the untreated tumor group. Using a smartphone camera, a daily digital image was also captured immediately after recording the LSF raw data.

The raw speckle image obtained from the camera was converted to a spatially correlated speckle-contrast (SC) image by computing the ratio of standard deviation to mean in speckle intensity at each pixel through a sliding window of 5 × 5 pixels in MATLAB (R2017b). By using the SC image, we adopted the speckle flow index (SFI) to simply acquire a pseudoabsolute velocity image [21, 24]. The values of SFI and SC have the following relationship [16]:

where T and K refer to the exposure time of the camera and SC value respectively. Theoretically, SFI is approximately proportional to the flow velocity v according to the following equations: [25] and , where τ_c, a, and k₀ respectively refer to the correlation time of the speckle fluctuation, a factor depending on the Lorentzian width as well as scattering properties of the tissue, and light wavenumber [26].

Before extracting SFI, we defined the region of interest (ROI) by considering a characteristic vessel pattern on the whole daily SC map for each group (normal, treated, and untreated tumor), as shown in Fig. 2. For later normalization of SFI, we also defined the nonvascular part with the vessel region excluded. The mean SFI in a ROI (SFI_vessel) was normalized to that in an avascular region (SFI_avascular), to minimize background artifacts such as tissue motion or variation in system illumination, according to the equation as follows [27]:

Here SFI_norm refers to the relative measures of flow velocity and is a ratio of correlation times. It is difficult in practice to extract an absolute velocity from the correlation time in LSF, because the number of moving particles with which the light interacts and their orientations are unknown [25]. Therefore, SFI_norm is a corrected pseudoabsolute velocity parameter in LSF.

In addition, we introduced ∆rBFR by modulating the inhaled gas. Tumor BFR in dorsal skin-fold vasculature typically decreases when switched to hypercapnia from normoxia (Figs. 3 and 4). Therefore, ∆rBFR was calculated as follows:

where A and B respectively refer to the SFI_norm at the moment just before being switched to hypercapnia from normoxia, and the minimum SFI_norm during hypercapnia.

Figure 2 shows a few of the daily LSFAs and their corresponding color photos, for the normal, treated, and untreated tumor groups. The FOV of the LSFA is indicated by a white dotted rectangle with a size of 8.8 mm × 6.6 mm. The primary criterion for choosing the ROI (as indicated by a red dotted area) in each group was to consider its characteristic vessel pattern with the naked eye throughout all of the daily LSFAs. Although the daily ROIs are variable due to the daily change in tumor size, it was easy to compare the variation of blood-flow morphology within each landmark boundary (i.e. each ROI) for each group during the course of observation. Mean SFI values were extracted within the ROIs in the daily LSFAs.

Meanwhile, the tumors were not perfectly treated, due to administration of low doses of anticancer drug. For example, the DWC photo on day 10 after the first CTX administration for the treated tumor group was still reddish, and the corresponding speckle-contrast image still showed high BFR-encoded dark color within the tumor boundary (Figs. 2(b) and 5(a)).

Figure 3 shows the general decrease in tumor BFR on a dorsal skin-fold layer when switched to hypercapnia. The inhalational gas is modulated according to the protocol in Fig. 1. By monitoring the E_tCO₂ signal, the duration of hypercapnia was determined, from 5 to 10 minutes. A smoothing filter was applied to all signals. Tumor volume was measured at ~8.3 mm³.

To examine the wide-field variations in a SFI image of a dorsal skin-fold vascular response to the inhalational-gas challenge, we compared SFI images at three different points in time (Figs. 4(a)~4(c)). Each set of ten SFI images was averaged to minimize speckle variation. Here we highlighted the spatially distributed subtle changes in the tumor microvascular BFR images by manually adjusting the maximum limit of the color bar. The spatiotemporal SFI images showed not only hypercapnia-induced reduction in tumor blood flow on dorsal skin (Figs. 4(a) and 4(b)), but also reactive hyperemia at the moment of returned normoxia (Fig. 4(c)). To emphasize the inhaled-gas-induced effect on tumor vasculature, we extracted differential images between the SFI images by purposely adjusting the maximum limit of the color bar (Figs. 4(d) and 4(e)). Figure 4(d) clearly showed the reduction of hypercapnia-induced BFR within tumor-bearing dorsal skin. Furthermore, the hypercapnia-induced BFR on the bigger blood vessel was significantly decreased in the subtraction map (Fig. 4(d)). Likewise, Figure 4(e) presented markedly compensatory increase in BFR to reestablish the steady state in the tumor region.

Figure 5 shows in part a positive correlation of a few BFR parameters with tumor volume for all the groups (normal, treated, and untreated tumor), although the animals had not fully recovered from the tumor during chemotherapy for the treated tumor group, as shown in Fig. 2(b).

Among the three kinds of BFR indices, first, to obtain the baseline SFI (Figs. 5(a) and 5(b)), we selected the SFI values during air breathing and collected them for all groups. As shown in Fig. 5(a), the baseline SFI increased gradually from ~150 to ~600 until day 3 as the tumor grew, and then it decreased toward ~400 until day 10, as the tumor volume regressed due to chemotherapy in the treated group. From day 11, however, it started to rise toward ~500 again, even though the tumor volume continued to decrease. For the normal group, it remained relatively stationary between ~150 and ~200 (Fig. 5(a)). In the untreated tumor group, the baseline SFI continued to rise from ~180 to ~1800 as the tumor progressed (Fig. 5(b)). Interestingly, the trend of the untreated tumor group is slightly different from that of the vessel skeleton density (VSD) using OCTA, as reported in our previous study [6]. For the uncured-tumor case in that paper, the VSD exhibited a W-shaped pattern at the end of tumor progression, perhaps because a necrotic or apoptotic region progressed [6, 28]. The BFR parameters, however, continued to go up, with no W-shaped trend (Figs. 5(b) and 5(d)). In short, we found that the tumor BFR continued to rise with tumor progression, although the microvessel density (i.e. VSD) goes up and down most likely due to necrosis or apoptosis. Therefore, we speculate that our spatial-resolution-limited LSF modality did not discern the vascular-destruction effect, while OCTA could [6]. Meanwhile, the increased variation in the baseline SFI for the treated tumor group is noted from day 3, presumably due to the decreased sensitivity and signal-to-noise ratio (Fig. 5(a)) [5].

Second, we normalized the baseline SFI to the SFI value in an avascular region during air breathing (i.e. baseline SFI_norm). Compared to the graphs for the baseline SFI (Figs. 5(a) and 5(b)), interestingly, the baseline SFI_norm of the normal group became considerably stable (Figs. 5(c) and 5(d)). In addition, the maximum value of baseline SFI_norm for both treated and untreated tumor groups was coincidentally limited to ~6.5, so that the baseline SFI_norm can be easily compared among all groups, unlike the baseline SFI (Figs. 5(c) and 5(d)).

Third, we compared the values of ∆rBFR for all of the groups (Figs. 5(e) and 5(f)). For the normal group, the small-positive integer values of ∆rBFR correspond to the insignificant decrease in relative BFR on a dorsal skin-fold vasculature during hypercapnia. In the treated and untreated tumor groups, however, ∆rBFR rose gradually as the tumor grew. The result is consistent with that reported by T. J. Dunn et al. [29] using LDF and an intravital microscope. They showed that the reduction in tumor BFR and clear vasoconstriction occur on a rat’s DWC model during carbogen (5%CO₂/95%O₂). Meanwhile, increased fluctuation in ∆rBFR was seen for all groups, compared to both the baseline SFI and baseline SFI_norm (Fig. 5). This might result from hypercapnia-induced motion artifact, or peculiar variation in BFR.

Figure 6 shows that tumor volume has a strong correlation with baseline SFI during tumor growth in the treated group as well as in the untreated group, while it has a poor correlation during tumor regression in the treated group. The poor correlation between them might be due to incomplete tumor treatment, as addressed in Sec. 4.3 of the ‘Discussion’.

We have demonstrated a partially positive correlation of tumor volume with several BFR parameters by using LSF. In addition, we found that the baseline SFI has the potential to be an indicator for early treatment monitoring. For example, the physiological changes in baseline SFI are detected one day earlier than the anatomical regression of the tumor, for the treated tumor group (Fig. 5(a)). Meanwhile, we observed that baseline SFI would be relatively suitable for monitoring tumor treatment, as compared to ∆rBFR. For example, there was a two-day time lag between the trend using ∆rBFR and those using two kinds of SFI (baseline SFI and baseline SFI_norm) (Fig. 5). The features of these three BFR parameters (baseline SFI, baseline SFI_norm, and ∆rBFR) are discussed in details as follows.

First of all, the baseline SFI is considered to be a reliable enough indicator to estimate tumor progression without the need for external inhaled-gas stimulation (Figs. 5(a) and 5(b)), although the tumor volume seldom reflected the tumor status, even in the partially chemotherapeutic situation. Here we speculated that tumor vasculature on day 10 might still remain physiologically aggressive, even though the anatomical size continued to regress for the treated tumor group (Fig. 5(a)). Nonetheless, more precise conversion of tumor volume should be taken into account. Meanwhile, the decrease in baseline SFI on day 4 was followed by tumor regression on day 5 (Fig. 5(a)), implying that baseline SFI would be an indicator to predict tumor treatment response.

To verify the feasibility of quasiquantification for baseline SFI, additionally we attempted to compare the indices of baseline SFI for the treated and untreated tumor groups. For instance, the baseline SFI for the treated tumor group was ~600 on day 3, and the corresponding tumor volume was ~55 mm³ (Fig. 5(a)), while it was ~690 at the identical tumor volume for the untreated tumor group (Fig. 5(b)), as shown in Table 1. Therefore, there was a discrepancy (∆≅+15%) in baseline SFI values between the treated and untreated tumor groups at the identical tumor volume. This might come from the nonlinearity of SFI with absolute blood-flow velocity. Therefore, we will show that the normalized SFI could reduce slightly the discrepancy [25]. Nevertheless, the baseline SFI would be a surrogate indicator of pseudoabsolute blood-flow velocity, although a more precise process such as calibration should be requested.

Second, we found that the baseline SFI_norm has the potential for early detection of the efficacy of tumor treatment, as the baseline SFI does above. To verify the feasibility of quasiquantification, for example, the baseline SFI_norm for the treated tumor group was ~4 on day 3, and the corresponding tumor volume was ~55 mm³ (Fig. 5(c)). For the untreated tumor group, the baseline SFI_norm was ~3.5 at the identical tumor volume (Fig. 5(d)), as in Table 1. Therefore, there is a lower discrepancy (∆≅−12.5%) in the values of baseline SFI_norm between the treated and untreated groups at identical tumor size, compared to the baseline SFI above. Therefore, baseline SFI_norm would be more reliable as a pseudoabsolute BFR parameter by using the ratio of correlation times, compared to the baseline SFI above [25].

Third, we investigated whether ∆rBFR might be an indicator to evaluate tumor response to treatment, by measuring the degree of tumor vascular reactivity to a respiratory challenge. It is based on the hypothesis that the variation in tumor blood-flow reactivity to an inhaled-gas intervention would reflect the degree of tumor characteristics (such as benign or malignant state, progression, or regression), by using an LSF system. Here we observed a reduction in hypercapnia-induced tumor BFR for a dorsal skin-fold vasculature (Figs. 3~5). For the normal group, ∆rBFR remained relatively stable with a small-positive integer (which refers to a small decrease in BFR), and it is consistent with the result of the normal DWC model from M. Neeman et al. [30] using LDF. For the treated and untreated tumor groups, however, ∆rBFR went up gradually as the tumor grew. We speculate that this may result from the failure of the autoregulated mechanism against the external-gas stimulus in tumor. In other words, the malfunction of the mechanism would occur due to the lack of both smooth muscle and pericytes in the immature neovasculature of the tumor [31, 32]. Likewise, to verify the feasibility of quasiquantification for ∆rBFR, we simply compared the values of ∆rBFR for both the treated and untreated groups. For instance, ∆rBFR for the treated tumor group was ~18 on day 4, and the corresponding tumor volume was ~60 mm³ (Fig. 5(e)). For the untreated tumor group, ∆rBFR was ~17 at the identical tumor volume (Fig. 5(f)), implying that ∆rBFR has the potential to be a quantitative parameter for all groups. However, the tumor regression on day 5 was followed by a decrease in ∆rBFR on day 6 (Fig. 5(e)). Therefore, ∆rBFR is considered inappropriate as an indicator of early prediction of tumor response to treatment, unlike both baseline SFI and baseline SFI_norm above.

Nonetheless, ∆rBFR is a qualified parameter that has already been verified in the literature [22]. In detail, H. Cheng et al. [22] demonstrated that the hyperoxia-induced percent change in the SFI map relative to baseline (normoxia) in rat retinal blood flow showed the same result as that using analytical modeling for a Lorentzian velocity distribution. Moreover, the ∆rBFR (percent change of SFI for H. Cheng et al. [22]) can make confounding factors (so-called β terms, which affect the speckle contrast by means of scattering properties of the tissue, degree of polarization, illumination angle, and ratio of pixel size to speckle size) cancel out with ease [22].

Among the three kinds of BFR parameters, we cannot conclude which is better than the others for monitoring tumor treatment. Because of slightly different approaches for the same data, each might have its own features. In summary, despite the identical tumor size, baseline SFI in the untreated tumor group was a little larger than in the treated group. On the contrary, both baseline SFI_norm and ∆rBFR for the untreated tumor group were slightly lower than for the treated group, at the identical tumor size. In addition, baseline SFI and baseline SFI_norm could be used as indicators for early detection of tumor response to treatment. Especially, we suggest that baseline SFI_norm would be a better indicator for early prediction of tumor response to chemotherapy without the need for inhalational-gas intervention.

Meanwhile, we attempted to calculate the ratio of tumor blood-flow rate (BFR) to tumor volume (TV) in Table 1. As the tumor grows, interestingly the growth rate of blood flow per unit volume is shown to decrease and then approach a plateau at 55 mm³ of TV, for both the treated and untreated groups. This means that as the tumor grows, BFR itself increases, while the growth rate of BFR per unit tumor volume (TV) decreases. The decrease in growth rate is most likely associated with the development of either a hypoxic or necrotic region in the central part of the tumor, which has poor blood perfusion [28] as the tumor grows bigger. Besides, we expect that the quantification of BFR could be improved by correcting several confounding factors that hinder its linear response to flow velocity in tissue, as discussed in Sec. 4.3.

We suggest that the baseline SFI has the potential to estimate pseudoabsolute BFR in dorsal skin, although a more precise process such as calibration should be requested later [33]. In the literature, the analytical relationship between the measured SFI and the actual flow rate in a specific velocity range was described for practical application (such as a dorsal skin-fold window chamber model) [21, 24]. B. Choi et al. reported that arterioles and venules in the rodent dorsal skin-fold and cranial window models were estimated as ~4 mm/s and ~5 mm/s respectively, and thus the SFI-based DWC model study could be reliable enough to measure such a velocity range [21].

To make sure that our DWC-model-based SFI value is valid, additionally we have checked that our daily speckle contrast values K were measured within 0 < K < 0.6 [24], especially for the initial day of DWC data recording. After that, the K value headed toward zero (which refers to higher tumor BFR) as the tumor grew. This implies that the measured SFI from our DWC model can provide a pseudoabsolute BFR with high accuracy. Furthermore, regardless of the approximate model of speckle fluctuation (i.e. a Lorentian or Gaussian model, or Goodman’s approach [34]), the K condition mentioned above warrants a linear relationship between the measured SFI and the actual flow velocity, if the exposure time is set to 10 ms [21, 22, 24]. Nonetheless, the limitations of the SFI approach and its evolution for actual velocity measurement will be presented in further detail in Sec. 4.3.

Here the range of baseline SFI was usually between 140 and 1800, considering the tumor BFR for all groups (Figs. 5(a) and 5(b)). This is consistent with that reported by Chao Zhou et al. [2]. In brief, they demonstrated that faster decay rates of autocorrelation curves are detected in a human breast cancer (indicating the higher blood flow in the tumor) compared to a normal breast, by using DCS. They insist that the results from DCS agree with those from US, PET, and MRI [2]. Meanwhile, the range of baseline SFI for healthy cerebral vasculature was between 2500 and 4500 during air breathing (n = 34, not shown). Therefore, it implies that the baseline SFI would be an alternative to an absolute BFR parameter, because it is reliable enough to reflect that the BFR of cerebral vasculature is typically higher than that of dorsal skin-fold vasculature.

When it comes to the safety evaluation of hypercapnia in clinical settings, it is said that hypercapnia is permissively tolerable to our body [35]. The physiological effects of hypercapnia are increasingly well understood, in terms of both advantageous and potentially detrimental aspects. In detail, it could reduce experimental acute lung injury because of adverse ventilatory strategies, and hypercapnic acidosis could also attenuate the inflammation response at the genomic level. In summary, M. N. Chonghaile et al. [35] suggest that humans can tolerate extreme levels of hypercapnia for relatively prolonged periods without adverse effects.

There are some limitations of the experimental procedure and methodology for this study. First, the dosage of the anticancer drug should be augmented later, for full recovery from the tumor (Figs. 2(b) and 5(a)). According to the protocol reported by H. Kim et al. [6], the metronomic CTX dosage empirically needs to be not 40 but ~100 mg/kg body weight, or a single high dose of 200 is recommended for complete chemotherapy in mice [36].

Second, we did not consider the multiscattering light model associated with tissue optical properties (absorption and scattering at 785 nm, in our case), to quantify flow velocity more precisely from the observed speckle contrast in the depth-integrated LSF configuration [5, 37]. Instead, we simply adopted normalization of BFR against an avascular region, to minimize the decorrelation from the tissue background even in a TILSF configuration. Therefore, the depth-integrated LSF system that we used poses a limitation on more accurate BFR. Alternatively, multiexposure laser-speckle imaging (MESI) [38], to extract the decorrelation time τ_c of the speckle fluctuation directly related to flow velocity, has been recently studied, to realize a more linear correlation of LSF with actual blood-flow velocity by compensating for some key factors (static scattering tissue layer being out of focus, vessel diameter, blood content, and volume fraction of RBCs) [18]. It is based on compound theories, such as DCS and diffuse optical spectroscopic imaging (DOSI) [5]. Therefore, we need to compare the SFI-based flow approximation of Eq. (1) to the corrected decorrelation time τ_c approach above for our animal study. These factors should be addressed in future studies.

Finally, LSF in clinical settings is only suitable for monitoring blood flow rate in superficial tissue, because it has a limited penetration depth of 0.4 to 1 mm [39]. For example, intraoperative LSF for imaging free flaps [40] has been successfully utilized in clinical trials. However, LSF still has limitations in human breast cancer, because the cancer lesion is located at a site deeper than LSF can monitor. Nonetheless, LSF can be a useful tool in basic research on breast cancer with a dorsal skin-fold chamber model, and may provide insight for clinical application to some extent. Meanwhile, minimally invasive LSF combined with microendoscopy has been suggested for proof of concept for imaging the deep brain [41] and human breast cancer [42], as one of the needle-based optical-imaging approaches (e.g. OCT, confocal/multiphoton microscopy, and fluorescence microscopy). By integrating with and guiding standard core-needle biopsy procedures, LSF has the potential to be translated into clinical applications with a FOV of 400 to 710 µm in diameter, for more accurate targeting of breast cancer.

The final goal of this study is to find a robust indicator to predict tumor response to treatment. For further work, therefore, we will utilize a combined spectral-imaging and LSF modality (e.g. dual-wavelength LSF [43] or hemoglobin oxygen-saturation imaging [44]) to extract tumor metabolism (such as evaluating mammary metabolic rate of oxygen and total optical index [9]) by acquiring both blood flow and oxygenation. Thereafter, the OCT angiography [6] that we utilized provides a highly spatial-depth-resolved blood-flow morphology (blood-vessel diameter and vasomotion effect) or absolute flow velocity [45, 46], to complement tumor hemodynamics. Ultimately we expect that this work may provide a better understanding of tumor behavior and metabolism, which may contribute to predicting treatment outcome, and thus to designing anticancer strategies.

We have showed that the variations in some tumor blood-flow parameters (baseline SFI, normalized SFI, and ∆rBFR) have partially positive correlations with the change in tumor volume during tumor growth and chemotherapy. Among those parameters, changes in the baseline SFI and normalized SFI responded one day earlier than the change in tumor size, compared to changes in the ∆rBFR. This work indicates that blood-flow parameters based on LSF would aid in evaluating tumor response to treatment, and would contribute to a better understanding of tumor biology in preclinical trials.