Photoacoustic imaging for the prediction and assessment of response to radiotherapy in vivo

Radiotherapy is commonly used for cancer therapy, although its efficacy is reduced in hypoxic regions of tumours. Photoacoustic imaging (PAI) is an emergent, non-invasive imaging technique that allows the measurement of blood oxygen saturation (sO2) which inversely correlates with hypoxia in tissue. The potential use of PAI as a prognostic tool for radiotherapy outcome was investigated in a head and neck cancer model in vivo. PAI was performed before delivering a single fraction (10, 20 or 30 Gy) treatment. The results show that tumours with pre-treatment higher blood sO2 responded better than those with lower levels in the 10 and 20 Gy groups. For the 30 Gy group, treatment response was independent of blood sO2. The haemoglobin content of the tumours was not correlated with their response to any of the radiation doses studied. Changes in sO2, monitored at 24 h and 96 h following 10 and 20 Gy doses, showed that tumours that were subsequently unresponsive to treatment had an increase in blood sO2 at both time points compared to those which subsequently regressed after radiotherapy. The results suggest that sO2 values measured by photoacoustic imaging can be used before, and shortly after, irradiation to predict subsequent treatment response.

Title: Photoacoustic imaging for the prediction and assessment of response to radiotherapy in vivo Abstract: Radiotherapy is commonly used for cancer therapy, although its efficacy is reduced in hypoxic regions of tumours. Photoacoustic imaging (PAI) is an emergent, non-invasive imaging technique that allows the measurement of blood oxygen saturation (sO 2 ) which inversely correlates with hypoxia in tissue. The potential use of PAI as a prognostic tool for radiotherapy outcome was investigated in a head and neck cancer model in vivo. PAI was performed before delivering a single fraction (10, 20 or 30 Gy) treatment. The results show that tumours with pre-treatment higher blood sO 2 responded better than those with lower levels in the 10 and 20 Gy groups. For the 30 Gy group, treatment response was independent of blood sO 2 . The haemoglobin content of the tumours was not correlated with Introduction Here, we hypothesize that PAI can be used for assessing tumour oxygenation status in a preclinical head and neck cancer model before, and after, single fraction radiotherapy treatments, and that these blood sO 2 measurements would correlate with treatment outcome. We have also also explored the capability of individual haemoglobin components (Hb, HbO 2 and total haemoglobin, HbT) and ΔsO 2 (sO 2 (oxygen) -sO 2 (air)) to predict and monitor radiotherapy response. The last parameter can be obtained from an oxygen challenge test and it has been shown to allow the delineation between different tumours' vascular characteristics [29], as oxygen can be used as a contrast agent for hypoxia studies.

Results
Pre-irradiation imaging for predicting tumour outcome We assessed the potential of PAI to predict RT efficacy by imaging animals bearing subcutaneous CAL R tumours, 24 h and immediately before delivering a single boost RT dose (10,20 or 30 Gy) to the tumour. Data was acquired under both medical air and 100%-oxygen breathing conditions. One imaging dataset acquired prior to delivering a RT dose of 10 Gy was excluded due to an artefact masking the signal in the tumour (Fig. S1). Fig. 1 shows the normalised tumour growth curves (caliper measurements) for the control (CTRL), 10, 20 and 30 Gy cohorts, subdivided into groups by the level of response to RT dose: full-responders (FR), partial-responders (PR) and no-responders (NR). The NR growth curves, shown in red, are within the range of variability of the tumour growth for the control cohort, shown in black. There was no difference in the survival of the NR cohort in comparison to the control group. FR, represented in green, had a tumour volume decrease, reaching 0 mm 3 , i.e. no macroscopic signs of disease were palpable. The lack of palpable tumour mass was maintained for the remaining 60 day follow-up period. The average time after-RT at which tumour volume, V, decreased below the initial volume, V 0 , was 29±8 days for 10 Gy, 9±6 days for 20 Gy and 11±5 days for 30 Gy. The survival of the full responders was increased compared to the controls. The PR growth curves, represented in blue, showed that there was an increase in the survival of the mice compared to that of the control cohort, i.e. some tumours regrew and some maintained a stable volume above V 0 .
Average blood sO 2 results Fig. 2A shows representative 'oxymaps', i.e. maps of the calculated blood sO 2 , for the tumour central slice, for the 10 Gy cohort, from the FR, PR and NR groups, immediately pre-RT, during air-breathing. The 'oxymaps' are shown overlaid on greyscale photoacoustic images, in order to visualise where, within the tumour, the blood sO 2 is being calculated.
'Oxymaps' of tumours of FRs had majority of pixels corresponding to high values of sO 2 , while those of PR and NR tumours had substantial pixels corresponding to low blood sO 2 .
The blood sO 2 values were averaged over three central tumour slices and measured 24 h and immediately before delivering 10 Gy irradiation. As seen in Fig. 2B and Table 1, the FR cohort had statistically significantly higher levels of blood sO 2 (0.660.02 and 0.730.02 for air-and oxygen-breathing, respectively, immediately pre-RT) in comparison to the PR (0.510.08 and 0.550.1 for air-and oxygen-breathing, respectively) and NR (0.440.05 and 0.470.04 for air-and oxygen-breathing, respectively) cohorts. FRs had a significantly higher average blood sO 2 values compared to the control group, apart from immediately pre-RT measurements during air-breathing (Table 1). On the other hand, NR tumours, imaged immediately pre-RT, had significantly lower average blood sO 2 (0.440.05 and 0.470.04 for air-and oxygen-breathing, respectively) than the controls (0.580.06 and 0.640.06 for airand oxygen-breathing, respectively). The 20 Gy group had only one PR with average blood sO 2 levels similar to those of NR. Therefore all the PAI data for this animal was combined with the NR data, and this combined group was compared with the FR cohort. The tumour averaged blood sO 2 values, obtained 24 h and immediately before delivering 20 Gy irradiation, are shown in Fig. 3 and Table 1.
During both air-and oxygen-breathing imaging conditions, tumours which went on to have a full response (n=7) had statistically significantly higher blood sO 2 (0.630.09 and 0.710.11 for air-and oxygen-breathing, respectively, immediately pre-RT) than those with partial-and no-response (n=3, 0.500.02 and 0.550.01 for air-and oxygen-breathing, respectively, immediately pre-RT). This difference was also observed 24 hours pre-RT. Twenty-four hours pre-RT, FR had significantly higher average blood sO 2 (0.630.08 and 0.730.07 for air-and oxygen-breathing) than the control cohort (0.520.07 and 0.620.07 for air-and oxygenbreathing). NR tumours had significantly lower sO 2 immediately pre-RT compared to the control cohort, but only during oxygen-breathing (0.550.01 and 0.640.06, for NR and control cohorts, respectively).
For tumours irradiated with 30 Gy the association between pre-treatment average blood sO 2 and radiation response was weaker. This could be because majority of the tumours responded to the high dose, irrespective of their blood sO 2 . One tumour had a low average blood sO 2 (0.35±0.01 24 h pre-RT and 0.41±0.01 immediately pre-RT) and went on to respond fully to the radiation ( Fig. 4 and Table 1). Although the average blood sO 2 for FR was higher (0.560.07 and 0.610.10 for air-and oxygen-breathing, immediately pre-RT) than for PR (0.450.07 and 0.500.03 for air-and oxygen-breathing, immediately pre-RT), there were no statistically significant differences between the two groups. Nevertheless, PR had significantly lower average blood sO 2 than the control cohort (Table 1).

ΔsO 2 and haemoglobin results
The difference in measured blood sO 2 during air-and oxygen-breathing (sO 2 ) and its relationship with treatment outcome was also investigated, 24h and immediately pre-RT. No significant differences were found between FR, PR and NR groups, at 10, 20 or 30 Gy (Tables S1-S3).
The average oxy-(HbO 2 ), deoxy-(Hb) and total haemoglobin (HbT) were measured for each tumour, at the two time points before treatment, to study if haemoglobin levels could also be used as a predictive factor to radiotherapy outcome. Mostly, there were no statistically significant differences in these PAI parameters between the FR group and PR/NR irrespective of the radiation dose (Tables S1-S3). However, for tumours irradiated with 10 Gy, there was a significant difference between FR and PR groups 24 h pre-RT (p-value of 0.041* during air-breathing and 0.0019** for oxygen-breathing; Table S1). FR had, on average, less HbO 2 (63 ± 17 A.U. for air-breathing, 54 ± 17 A.U. for oxygen-breathing) than PR (143 ± 44 A.U. and 120 ± 17 A.U. for air-and oxygen-breathing, respectively). This significance was only observed for this comparison, so it is likely to be an outlier.

Post-irradiation imaging for treatment monitoring
The potential of PAI to monitor RT treatment response was investigated by estimating blood (air-breathing), for one FR and one NR. Immediately pre-RT, the FR 'oxymaps' had predominantly red pixels, indicating an average sO 2 > 0.5, while NR 'oxymaps' had largely white and blue pixels, suggesting sO 2  0.5. Ninety-six hours post-RT, an increase in blue pixels at the centre of the FR tumour (Fig. 5), was observed, suggesting a decrease in the mean blood sO 2 . On the other hand, in the NR tumour, an increase in the red pixels or high blood sO 2 regions was observed 96 h post-RT, particularly in the margins of the tumour. The remaining 'oxymaps' for the 10 Gy cohort are shown in Fig. S1 The percentage change in average blood sO 2 measured at 24 h and 96 h post-RT with respect to the baseline (i.e. that measured for the same tumour immediately pre-RT) for the control and 10 Gy cohorts are shown in Fig. 5B and Table 2. The mean percentage change in average blood sO 2 for the control animals was close to zero, for both air-and oxygenbreathing, at both time points. As seen in Fig. 5 the mean percentage change in blood sO 2 for the FR and the NR cohort, was consistently negative and positive, respectively at both time-points post-RT. There was no statistically significant difference between the control and FR cohorts. Interestingly, the NR group showed a statistically significant increase in blood sO 2, at both 24 h and 96 h post-RT, in comparison to either the FR or control groups, during air-or oxygen-breathing imaging ( Table 2). The partial-responders (PR) had a positive percentage change in average sO 2 at both 24 h and 96 h post-RT, although there were no statistically significant differences between these and the remaining response groups or control cohort.
The percentage change in blood sO 2 at 24 and 96 h post 20 Gy irradiation, with respect to blood sO 2 pre-RT, is shown in Fig. 6 and Table 2 Tables S4-S6). One exception, however, was that 2 PR tumours, irradiated with 30 Gy, had a significant decrease of 65% or 44% in the level of deoxy-haemoglobin (Hb) at 96 h post-RT compared to immediately pre-RT, during airbreathing.

Discussion
In this study, we have tested the hypothesis that 'PAI can predict and monitor tumour response to RT'. The tumoural blood sO 2 , ΔsO 2 and haemoglobin levels (specifically Hb, HbO 2 and HbT) were estimated using PAI, prior to and shortly after delivering RT doses of 10 Gy, 20 Gy and 30 Gy to test the hypothesis.
Radiotherapy response categories were chosen to reflect the different types of tumour growth behaviour seen in response to radiation treatments. From Fig.1, it is possible to observe that while for 10 Gy, only 3/13 tumours regressed, the majority of tumours treated with 20 Gy (7/10) and 30 Gy (8/10) did so. The regression took longer in the 10 Gy cohort (298 days) than in the 20 Gy (96 days) and 30 Gy (115 days) groups. This was expected as the higher the ionising dose, the greater the cellular damage, with more tumour cells dying [30,31]. The proportion of NR was greater for 10 Gy (5/13) than for 20 Gy (2/10), and all 30 Gy treatments resulted in full-or partial-response.
The PR group contained both tumours that regrew and those that stopped growing after RT, with no tumour regression. One of the consequences of hypoxia is that cells enter a dormant state, in which they proliferate more slowly than normoxic cells, but also remain viable for long periods of time [32]. If new vasculature is formed or reoxygenation occurs after radiotherapy, these cells may leave their dormant state and proliferate [33]. This may be the mechanism involved in the different behaviour of PR tumours seen after RT, with tumours that regrew possibly being reoxygenated during the 60 day follow up period of this study, while tumours with growth inhibition were not.
Hypoxia is a microenvironmental feature which results in radioresistance and, consequently, increases the risk of unsuccessful RT treatment. This study has investigated whether tumour blood sO 2 measured with PAI, 24 h or immediately before treatment, could predict response to 10, 20 or 30 Gy irradiation doses. The results showed a clear trend for FRs having higher average pre-treatment blood sO 2 , in the 10 and 20 Gy radiation groups, than the PRs or NRs ( Figs. 2 and 3). The FR tumours from the 10 Gy cohort also had significantly higher baseline blood sO 2 values (Fig. 2) than the controls, suggesting that high pre-treatment sO 2 (>0. 65) might be predictive of a good response to radiotherapy.
In line with this, blood sO 2 levels at baseline of 10 and 20 Gy non-/partial-responders were significantly lower than those of the controls (Figs. 2 and 3). This indicates that tumours with relatively low blood sO 2 might respond poorly to treatment. Nevertheless, while for 10 Gy only a partial-response was observed, for the 20 Gy cohort a full-response was obtained, probably due to the greater level of radiation damage to the tumour. The results for the 30 Gy group show a full response to treatment for a wider initial sO 2 range (0.38 to 0.68 pre-RT for air-breathing), and hence an increased percentage of tumour remissions, 80%, was observed ( Figs. 1 and 4). Since 30 Gy is a high single fraction dose, it is likely that even hypoxic cells are not resistant to the dose. This treatment efficacy was also seen for one animal in the 20 Gy radiation group which had a low sO 2 value of 0.43±0.02 immediately pre-

RT.
The potential to use PAI in the clinic to screen patients before radiotherapy and thus to personalise their radiotherapy exposure to their measured blood sO 2 would allow the reduction of dose in some patients to avoid side effects and its increase in others to avoid reduced efficacy. Other research groups have demonstrated, in clinical and pre-clinical models, that tumour oxygenation levels prior to radiotherapy correlate with treatment outcome [34,35]. For example, Nordsmark et al. [36], in 1996, found that patients' head and neck tumours with low (< 2.5 mmHg) partial oxygen (pO 2 ), measured using oxygen electrodes, had significantly lower loco-regional tumour control. Similarly, a better RT outcome for better oxygenated tumours (pO 2 > 2.5 mmHg) has been shown in subcutaneous C3H mammary carcinomas, in mice [37]. HÖckel et al. [38], using needle electrodes, found that cervical cancer patients with low tumour pO 2 (median threshold <10 mmHg) tumours had significantly worse overall survival than those with better oxygenated tumours (80 months). Non-invasive imaging techniques, such as positron emission tomography (PET) [37,[39][40][41] and magnetic resonance imaging (MRI) [42][43][44][45] have also been used before delivering RT to infer hypoxia, demonstrating that tumours with higher hypoxic fractions were associated with worse response to treatment than those with lower hypoxic fraction, similar to our findings. To the authors' knowledge, there are no studies which compare the capability of PAI, MRI and PET for predicting RT effectiveness but this would be worth investigating.
Besides the analysis of blood sO 2 , the individual components of haemoblogin, deoxy-(Hb) and oxyhaemoglobin (HbO 2 ), as well as the total amount of haemoglobin (HbT), were analysed to investigate their relationship with treatment outcome. The total haemoglobin findings were concordant with those obtained in another study by Rich et al. [25], suggesting this parameter is not a predictive factor for tumour response to radiotherapy. Nordsmark and Overgaard [46], while investigating the relationship between hypoxia and tumour control after radiotherapy, showed that oxygenation in tumours, measured using Eppendorf oxygen electrodes, was independent of total haemoglobin levels, obtained from a venous blood sample acquired before the treatment. Also, Joseph et al. [47] found that a higher coefficient of variation (CoV) was expected for haemoglobin than for sO 2 over either short periods of time (hours) or longer ones (days), as we have confirmed in a previous study [48]. This higher CoV makes Hb, HbO 2 and HbT less predictive parameters than blood sO 2 .
Another parameter which was not predictive of treatment outcome was sO 2 . Tomaszwwski et al. [29] has shown in 8 nude mice bearing PC3 prostate cancer that sO 2 is linearly correlates with an increased uptake of indocyanine green (ICG), suggesting also an increase in HbT. This implies that if haemoglobin was not a predictive factor in radiotherapy, sO 2 should also not be one.
The differences between average blood sO 2 before and after treatment were also analysed.
In the control cohort, whilst 3 tumours showed an increase in oxygenation (of 3 to 25%, airbreathing) from baseline to 96h post-RT, for all the others (n=7) the opposite effect (-3 to -22%, air-breathing) was observed. It is likely that in the latter cases, tumours were outgrowing their blood supply, whereas in the former we hypothesise that neoangiogenesis (the production of new vasculature) has occurred and oxygenation has met or exceeded the demands of these tumours. In fact, we have shown [28] that tumours with faster growth rates have a larger decrease in blood sO 2 over time compared to those with lower growth rates.
It is important to highlight that radiation dose, along with type of tumour and level of oxygenation pre-treatment, can affect the level of blood oxygenation post radiotherapy [49]. It has been shown that while conventionally fractionated low-dose radiotherapy (~2 Gy/fraction) tends to preserve vasculature, hypofractionated-doses (>10 Gy/fraction) are likely to cause vascular collapse, due to endothelial cell death [50,51], and subsequently a decreased blood sO 2 . This is one of the arguments for using hypofractionated RT, as vascular collapse results in a decrease in oxygen and nutrients to the tumour and consequently in secondary cell death [49]. Garcia-Barros et al. [52], for example, showed that for large single doses (10 to 20 Gy), microvascular damage was observed in murine fibrosarcoma (MCA/129) and melanoma (B16F1) murine xenograft models within 6 hours of irradiation with subsequent tumour cell death.
In our study, for both 10 and 20 Gy exposure groups, i.e. hypofractionated regimes, the FR cohort showed no change, or a mean decrease, in blood sO 2 both 24 h (0.3 ± 5 to -10 ± 6%) and 96 h post-RT (-8 ± 18 to -13 ± 10%). One possible explanation for this trend is endothelial cell death causing blood vessel collapse. Interestingly, Rich et al. [53] used patient-derived xenograft head and neck squamous cell carcinoma (HNSCC) models and showed a positive correlation between increased PAI measured blood sO 2 24 h post-RT (15 Gy) and tumour growth inhibition 2 weeks later. We had similar observations for the FR tumours of the 30 Gy cohort. In contrast to the 10 and 20 Gy cohorts, the FRs in the 30 Gy group had a mean increase in blood sO 2 at 24 h (6±17% and 6±11%, for air-and oxygen-breathing, respectively) and 96 h post RT (7±18% and 13±18%, for air-and oxygen-breathing, respectively). It can be speculated that for a high dose (30 Gy), the CAL R tumour cells are drastically damaged reducing the oxygen demand. The increase in blood sO 2 might also be due to acute inflammation arising due to a large radiation dose being delivered in one single fraction [49].
On the other hand, a trend for an increase in post-RT average blood sO 2 relative to baseline (13 ± 8 to 26 ± 15 %) was observed for PR and NR. This could be due to a lack of endothelial cell death, which then cannot counteract physiological phenomena such as inflammatory processes due to acute radiation treatment or the development of a new matured vascular network after radiation. Both processes are common after radiation and can influence tumour oxygenation [54,55]. In fact, some studies have suggested that formation of new vessels post-RT is often observed before a tumour regrows and drastically affects tumour control [56][57][58][59].
The lack of histological evidence here allows only speculation about the underlying physiological mechanisms occurring after radiotherapy, such as endothelial cell death, tumour reoxygenation or inflammation. Histopathology was not possible for this study as the main aim was to obtain long-term (60 days) tumour volume measurements for the accurate evaluation of response to treatment. In the future, animals will be imaged at regular intervals within the first 24-96 hrs, and oxygenation changes will be evaluated histologically to investigate the timing of these changes. It was also not possible to validate the hypoxic state of tumours before treatment, which could be done using exogenous markers, such as pimonidazole, or other imaging techniques, such as BOLD-MRI or PET. However, our group has shown the good agreement between PAI measurements of blood sO 2 and pimonidazole staining in another paper [28]. While initial results are encouraging, further investigation is necessary with a larger number of animals, as some radiation exposures resulted in response groups with a small number of animals (<5).
This study shows the potential for using the average blood sO 2 measured in tumours by PAI, under both medical air (21% O 2 ) and 100% oxygen-breathing conditions, before RT to predict their response to radiotherapy, which has been demonstrated here for the first time.
In addition, the potential to monitor treatment outcome was demonstrated. PAI measured changes in average blood sO 2 at 24 h and 96 h post-RT compared to those pre-RT showed a trend for a decrease in this parameter to be correlated with better tumour response, particularly for the later time point. Thus the potential for using photoacoustic imaging as both a predictive and a treatment assessment tool for radiotherapy has been showed in this paper.

Cell line and xenograft model
A human HNSCC cell line, CAL R , was cultured in vitro as described previously [60].
Xenografts were established by injecting 5 x 10 5 cells subcutaneously, into the right flank of female mice (FOXnu n1 , 6 weeks old, ~25 grams). All research was conducted under the Guidelines for Animal Welfare as per recommendation by the UK Home Office Animals (Scientific Procedures) Act 1986 [61] and the ARRIVE guidelines [62].
Tumour size was measured non-invasively using callipers every 2 or 3 days, in 3 orthogonal directions (length (l), width (w) and height (h)). An ellipsoidal shape was assumed and the tumour volume (V) calculated using equation (1).  Partial response (PR): either a reduced tumour growth rate following RT or tumour regression after RT followed by regrowth at a later time.
 Full response (FR): tumour size decreased with some tumours becoming undetectable.

Photoacoustic Imaging
Animals were imaged using the MultiSpectral Optical Tomography small animal imaging device (MSOT, iThera, Germany), as described in [48]. Before imaging, animals were

Competing interests
The authors declare no competing interests.

Availability of materials and data
All data is available for the readers upon request.  Figure 2: A: Three examples of 'oxymaps' obtained pre-RT for the 10 Gy cohort, during air-breathing, for FR, and NR tumours. The 'oxymaps' show the calculated blood sO 2 pixel-by-pixel and are overlaid in a greyscale photoacoustic image. Blue regions represent blood sO 2 below 0.5 and pink-to-red regions blood sO 2 above 0.5. Black regions represent pixels in which the calculated total amount of haemoglobin falls below 0, i.e. the amount of haemoglobin was below the system's noise threshold. Yellow bars = 10 mm; B: body of the mouse; W: water, used as an acoustic coupling medium. B: Average blood sO 2 measured 24 h and immediately before delivering 10 Gy irradiation. Black circles represent the average of three central PAI slices for each tumour. Black horizontal lines in the boxplots represent the median sO 2 with the box extremities representing the 25% and 75% percentiles and the whiskers representing the maximum and minimum values. Full responders (FR, n=3) had a statistically significantly higher average blood sO 2 than partial-(PR, n=5) and no-responders (NR, n=5) during both air-and oxygen-breathing, both 24 h and immediately pre-RT. * p-value < 0.05; ** pvalue < 0.01; *** p-value < 0.001 using a paired Student's t -test. Figure 3: Average blood sO 2 measured 24 h and immediately before delivering 20 Gy irradiation. Black circles represent the average of three central PAI slices for each tumour. Red circle represents the partial-responder. Black horizontal lines in the boxplots represent the median average, with the box extremities representing the 25% and 75% percentiles and the whiskers representing the maximum and minimum values. Full responders (FR, n=7) had a statistically significantly higher average blood sO 2 than combined partial-(PR, n=1) and no-responders (NR, n=2) during both air-and oxygen-breathing, both 24 h and immediately pre-RT. * p-value < 0.05; ** p-value < 0.01; *** p-value < 0.001 using a paired Student's t -test. Figure 4: Average blood sO 2 measured 24 h and immediately before delivering 30 Gy irradiation. Black circles represent the average of three central PAI slices for each tumour. Black horizontal lines in the boxplots represent the median average, with the box extremities representing the 25% and 75% percentiles and the whiskers representing the maximum and minimum values per group. No statistical differences (Student's t-test) were observed between full-(FR, n=8) and partial-responders (PR, n=2). Figure 5: A: Examples of tumour 'oxymaps' obtained immediately pre-RT and 96 h post-RT for 10 Gy cohort, showing the differences in blood sO 2 between FR and NR at the two time points (airbreathing). Green arrows indicate blue regions (low blood sO 2 ) observed 96 h post-RT for the FR, which were not present pre-RT. White arrows indicate red regions, i.e. regions with high blood sO 2 , for the NR 96 h post-RT, which were not observed pre-RT. Yellow bar = 10 mm. B: Percentage change in average blood sO 2 for the 10 Gy RT group between 24 hours, left column, and 96 hours post-RT, right column, and immediately pre-irradiation. Black circles represent the change in average blood sO 2 for each tumour measured over 3 central slices. Black horizontal lines in the boxplots represent the median average, with the box extremities representing the 25% and 75% percentiles and the whiskers representing the maximum and minimum values per group. FR = full-responders; PR = partialresponders; NR = no-responders. * p-value < 0.05; ** p-value < 0.01 using a paired Student's t-test. Figure 6: Percentage change in average blood sO 2 for the 20 Gy RT group between 24 hours, left column, and 96 hours post-RT, right column, and immediately pre-irradiation. Black circles represent the change in average blood sO 2 for each tumour measured over 3 central slices. Red circle corresponds to the partial-responder. Black horizontal lines within the boxplots represent median blood sO 2 change per group, with the box extremities representing the 25% and 75% percentiles and the whiskers representing the maximum and minimum values per group. FR = full-responders; PR+NR = partial-and no-responders. * p-value < 0.05; ** p-value < 0.01 using a paired Student's ttest.   30 Gy RT group between 24 hours, left column, and 96 hours post-RT, right column, and immediately pre-irradiation. Black circles represent the change in average blood sO 2 for each tumour measured over 3 central slices. Black horizontal lines within the boxplots represent median average blood sO 2 change per group, with the box extremities representing the 25% and 75% percentiles and the whiskers representing the maximum and minimum values per group. FR = full-responders; PR = partial-responders. * p-value < 0.05; ** pvalue < 0.01 using a paired Student's t-test. Tables   Table 1: Average blood sO 2 ( standard deviation, s.d.), acquired by PAI, for tumours irradiated with 0, 10, 20 or 30 Gy. Treatment response groups are: FR (full-responders); PR (partial-responders) and NR (no-responders) and images were acquired during both air-and oxygen-breathing 24 h pre-and immediately pre-RT. Statistically significant differences between FR and the other response groups are indicated by: * p-value < 0.05; ** p-value < 0.01; *** p-value < 0.001 using a paired Student's ttest.