The Predictive Value of SPECT/CT imaging in colorectal liver metastases response after 90Y-radioembolization

The aim of this study was to evaluate a modified method of calculating the 99mTc/90Y tumor-to-normal-liver uptake ratio (mT/N) based on SPECT/CT imaging, for use in predicting the overall response of colorectal liver tumors after radioembolization. A modified phantom-based method of tumor-to-normal-liver ratio calculation was proposed and assessed. In contrast to the traditional method based on data gathered from the whole tumor, gamma counts are collected only from a 2D region of interest delineated in the SPECT/CT section with the longest tumor diameter (as specified in RECIST 1.1). The modified tumor-to-normal-liver ratio (mT/N1) and 90Y predicted tumor absorbed dose (PAD) were obtained based on 99mTc-MAA SPECT/CT, and similarly the modified tumor-to-normal-liver ratio (mT/N2) and 90Y actual tumor absorbed dose (AAD) were calculated after 90Y-SPECT/CT. Tumor response was assessed on follow-up CTs. Using the newly proposed method, a total of 103 liver colorectal metastases in 21 patients who underwent radioembolization (between June 2009 and October 2015) were evaluated in pre-treatment CT scans and 99mTc-MAA-SPECT/CT scans and compared with post-treatment 90Y-SPECT/CT scans and follow-up CT scans. The results showed that the mT/N1 ratio (p = 0.012), PAD (p < 0.001) and AAD (p < 0.001) were predictors of tumor response after radioembolization. The time to progression was significantly lengthened for tumors with mT/N1 higher than 1.7 or PAD higher than 70 Gy. The risk of progression for tumors with mT/N1 lower than 1.7 or PAD below 70 Gy was significantly higher. The mT/N2 ratio had no significant correlation with treatment results. Conclusion The mT/N1 ratio, PAD, and AAD can be used as predictors of tumor response to SIRT treatment, and SPECT/CT imaging can be used for dosimetric assessment of radioembolization.


Introduction
The main advantage of Yttrium-90 in palliative treatment of liver tumors is its pure beta-radiation emission with high energy and short half-life [1,2]. The main limitation of radioembolization is the low tolerance of healthy liver tissue to radiation [3][4][5]. Therefore, proper calculation of the activity of 90 Y-microspheres, which can destroy liver tumors while sparing healthy liver tissue, is very important. The absence of primary gamma emission of 90 Y leads to significant problems not only with post-treatment imaging of therapy effects but also with calculation of dosimetry parameters for liver parenchyma [6][7][8]. Consequently, tumor dosimetry is not obtained in everyday practice. Dosimetry based on the MIRD equation (Medical Internal Radiation Dosimetry) and SPECT/CT imaging appears to be the best option for most centers performing radioembolization, but to attain wider use, it should be easier to calculate [7,9,10].
Patients undergoing radioembolization typically undergo a pre-treatment 99m Tc-MAA SPECT/ CT scan to calculate the 90 Y dose. The tumor-to-normal-liver uptake ratio (mT/N1) obtained in this procedure may serve as a basis for calculating further dosimetric parameters, including the predicted tumor absorbed dose (PAD) of 90 Y. A second SPECT/CT (bremsstrahlung) scan is performed after radioembolization to assess the post-treatment 90 Y-microsphere distribution. On this basis the tumor-to-normal-liver ratio (mT/N2) and actual adsorbed dose of 90 Y (AAD) can also be calculated and compared with pre-treatment and follow-up data [1,7,9,10].
With early dosimetric evaluation, patients with high potential risk of radiation-induced liver disease or patients with non-curative tumor doses can be recognized shortly before or immediately after undergoing a selective internal radiation therapy (SIRT) procedure. In view of recent reports on SIRT from prospective trials (SIRFLOX/FOXFIRE), it is important to develop a method for selecting patients who would benefit from the therapy [11,12]. In this study, we propose and assess a new approach involving SPECT/CT for predicting the effects of radioembolization in patients with liver tumors, based on a simplified method of calculating the tumor-to-normal-liver ratio.

Materials and methods
The study was approved by Ethical Committee Board of Military Institute of Medicine (decision: 24/WIM/2009). All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and o/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Informed consent was obtained from all individual participants included in the study.

Phantom dosimetric study
We first performed a phantom study to determine if gamma counts obtained only from the middle-of-sphere section could be used for dosimetric calculations in 90 Y radioembolization. The IEC Body Phantom was used, filled with six different volumes of calibrated amounts of 90 Y resin microspheres. The inner diameters of the fillable spheres were 10 mm, 13 mm, 17 mm, 22 mm, 28 mm and 37 mm. The activity in the spheres was 3.8 MBq/ml, while the background activity was 0.038 MBq/ml. Gamma counts were collected from each phantom section containing spheres, and the sphere-to-background count ratio was calculated (phT/N).
The two smallest spheres of the IEC Phantom were not clearly visible and were excluded from the study.

Study population
From June 2009 to December 2015 twenty-one patients with unresectable liver metastases of colorectal cancer were enrolled. A total of 103 tumors were selected and evaluated.

Study protocol
Radioembolization was conducted in line with guidelines approved by a panel of experts and the local ethics committee. One session per patient of resin microsphere (Sirtex, Australia) treatment was used for the study. As a pre-treatment procedure, 120-180 MBq of 99m Tc-MAA was administered into the common hepatic artery. In the radioembolization procedure, resin microspheres were injected using the same microcatheter position. SPECT/CT was performed < 25 min after 99m Tc/ 90 Y administration. The Body Surface Area Method was used for calculation of 90 Y activity [9,10]. As the first step, a region-of-interest (ROI) analysis of the tumor-to-normal-liver ratio was used on the SPECT images to determine the standard ratio (sT/N)-a detailed description is available elsewhere [7,9]. The modified tumor-to-normal-liver uptake ratio (mT/N) was then calculated using a new method proposed by the authors, as described below.
The RECIST 1.1 criteria were used to evaluate target liver lesions. Correlations between dosimetric parameters of target liver lesions detected with 99m Tc/ 90 Y-SPECT/CT were investigated and compared with initial and follow-up CT results. The target liver lesions were those that had a longest diameter of at least 10 mm and were clearly visible on CT and 99m Tc/ 90 Y-SPECT/CT. Tumors with areas of necrosis were not excluded (9%) [13].
For SPECT/CT imaging the GE Infinia Hawkeye 4 was used, with 1-inch NaI crystal detectors and a low-dose four-slice CT protocol. The protocols for 99m Tc-MAA imaging and 90 Y bremsstrahlung image acquisition are shown in Table 1.
For image post-processing, a Xeleris 3.0423 Workstation (GE, USA) with Volumetrix MI software was used. The ordered subset expectation maximization (OSEM) method of iterative reconstruction was applied, with 2 iterations and 10 subsets. The prefilter HANN 0.9 and the 3D postfilter BUTTERWORTH 0.48 were used.

Patient dosimetric study
The main steps in calculating dosimetric parameters were as follows: 1. Selection of target tumors in pre-treatment CT scans and identification of them with 99m Tc/ 90 Y-SPECT/CT. 2. Calculation of the modified tumor-to-normal-liver ratio (mT/N1 and mT/N2). 3. Calculation of the predicted and actual absorbed dose for the target liver tumors (PAD and AAD, respectively). 4. Evaluation of tumor responses using data obtained in CT scans and of tumor dosimetric parameters [7,14].
To calculate mT/N1, the slice with the longest target tumor diameter was selected in 99m Tc-MAA SPECT/CT. Its borders were delineated on CT using semi-automatic tools: a 2D ROI was drawn manually on a CT image generated with an abdominal CT window (Volumetrix MI application on Xeleris 3.0423 Workstation). In the next step, the ROI was automatically duplicated for the fusion 99m Tc-MAA scans, and counts of gamma emission from within it were obtained (see Fig 1). Another ROI was placed in an area of healthy liver tissue adjacent to the assessed liver tumor, within the same liver segment, and the gamma emission counts from within this second ROI were determined. The mT/N1 ratio for 99m Tc-MAA SPECT/CT was then calculated using the following formula: 1) [7,[14][15][16].
mT/N1 -modified tumor to normal liver uptake ratio in 99m Tc-MAA SPECT/CT To determine the mT/N2 ratio for 90 Y-SPECT/CT, all the steps listed above were repeated. The liver absorbed dose (LAD1 or LAD2: predicted or actual, respectively), tumor predicted absorbed dose (PAD), and tumor actual absorbed dose (AAD) were calculated according to the following formulas, using the volume as a surrogate for the mass: 2-3) [7,[14][15][16].
LAD1/LAD2-90 Y liver absorbed dose (Gy) calculated separately for 99m TC-MAA and 90 Y-SPECT/CT 90Y A -90 Y activity to administer (GBq) LB -lung breakthrough mT/N -modified-tumor-to-normal uptake ratio (mT/N1 or mT/N2 respectively) n -number of liver tumors TD PAD/AAD -90 Y tumor predicted or actual absorbed dose (Gy), respectively and PFS (progression-free survival) were computed using the Kaplan-Meier estimator. The log rank test was used for comparisons of time to progression (TTP tumor ). Correlations between variables were assessed with Spearman's correlation coefficient. Multivariate correspondence analysis was used for tumors to assess complete response (CR) and disease progression (PD). Receiver operating characteristic (ROC) curves were generated for PAD and mT/ N1. We performed calculations for PAD based on the generally accepted threshold of 70 Gy [6]. P values less than 0.05 were considered significant. Statistical calculations were carried out using STATISTICA 12.0 software.

Phantom data
As shown in Table 2 and Fig 2, gamma counts from the middle (widest diameter) section closely matched those obtained from the entire sphere, indicating that they could be used to calculate the liver tumor absorbed dose.

Patient data
The SIRT procedures were performed with technical success, and no serious adverse events were observed. Overall survival (OS) was 17.3 months and progression-free survival (PFS) was 6 months. The clinical and treatment data are summarized in Table 3.

Tumor and dosimetric data
Initially 112 liver tumors within 21 patients were found, but for 9 tumors the target lesion did not meet the inclusion criteria. Thus, data for 103 liver tumors were analyzed. (Table 4) The overall tumor response rate (ORR tumor ) was 36% (37 tumors). Their median values of descriptive parameters were as follows: T0 = 19 mm, mT/N1 ratio = 2.1, mT/N2 ratio = 1.8, PAD = 96 Gy, AAD = 89 Gy. The median time to tumor progression (TTP tumor ) obtained using the Kaplan-Meier estimator was 7 months.

Predicted and actual tumor 90 Y absorbed dose
Detailed information on PAD and AAD is presented in Table 4. For 37 tumors that were responsive to treatment, the mean PAD was 113 Gy and the mean AAD was 98 Gy (p < 0.001, Wilcoxon test). For 66 non-responding tumors the means were 82 Gy and 77 Gy, respectively (p<0.05). Negative correlations were found for PAD with T0 (R: -0.45, p < 0.001), PAD with TUMOR VOLUME (R: -0.43, p < 0.001), AAD with T0 (R: -0.41, p < 0.001), and AAD with TUMOR VOLUME (R: -0.4, p < 0.001). Both PAD and AAD showed significant correlations with tumor response after the SIRT procedure (Kruskal-Wallis test and ANOVA, respectively).
In NIR ANOVA post-hoc tests performed for AAD, significant differences between CR, PR, SD and PD were revealed (e.g., 82 Gy in CR vs 50 Gy in PD, p = 0.031). Results of calculations conducted for tumors with absorbed dose higher or lower than 70 Gy and size smaller or larger than 20 mm are shown in Table 4 and Fig 4. TTP tumor was significantly longer for tumors with PAD higher than 70 Gy (6.9 vs 2.1 months, p = 0.004, log rank test). The risk of progression was elevated for tumors with PAD lower than 70 Gy (HR: 2.0, CI: 1.3-3.2, p = 0.001). The multivariate correspondence analysis performed for tumors with complete response (CR) and progression disease (PD) revealed a large impact of PAD and AAD on these variables. (Fig 5)

Discussion
The aim of this study was to assess the value of SPECT/CT in predicting the effects of radioembolization for liver tumors, using a simplified method of calculation of the tumor-to-normalliver ratio (mT/N). SPECT/CT scans are performed two times during the radioembolization procedure. The results showed that if 99m Tc-MAA and 90 Y-microspheres are similar in size, it is possible to predict the actual distribution of the latter in the liver [9,10].
SPECT/CT imaging can give valuable dosimetric data, crucial for both planning and evaluating SIRT treatment. We have proposed a fast and easy method for dosimetric calculations [13]. The method relies on the RECIST 1.1 criteria, in which the longest tumor diameter (not the tumor volume) is used for further evaluation on follow-up CT scans. These criteria, adopted for SPECT/CT imaging, reflect the notion that counts from the largest lesion diameter accurately reflect the counts in the entire tumor. We confirmed in the SPECT phantom study that gamma counts from the section with the longest tumor diameter are usable to calculate the tumor absorbed dose.  The predictive value of SPECT/CT imaging in mCRC response after 90 Y-radioembolization The first step to achieving this goal in the liver is to measure the gamma counts in the 99m Tc-MAA and 90 Y-Bremsstrahlung SPECT/CT slices with the longest tumor diameters and calculate the tumor-to-normal-liver ratios (mT/N1 and mT/N2 respectively). Other dosimetric parameters, such as predicted tumor dose (PAD) and actual tumor dose (AAD), can then be calculated [7,[13][14][15][16]. The process of gathering dosimetric parameters is fast, taking only 1-2 minutes for liver tumors. From a dosimetric point of view, the tumor-to-normal-liver ratio is a very interesting parameter because it reflects blood flow in the tumor, as the hepatic artery is the main source of blood supply for liver lesions [1,17,18]. This value also forms a basis for calculating further dosimetric parameters, especially PAD for pre-treatment procedures and AAD after radioembolization [7,14,16].
As we observed, the mean mT/N1 was 2.1, which means two times greater blood flow in the tumor than in the healthy liver. The mT/N1 ratio was higher for responsive tumors and showed a positive correlation with TTP tumor , thus it may be used as a predictor of tumor response. This knowledge should give clinicians an opportunity to plan better treatments, by re-evaluation of the 90 Y dose before radioembolization. We should remember, however, that calculations made using mT/N1 may sometimes be less accurate due to differences in the distribution of 99m Tc-MAA and 90 Y-microspheres in the liver caused by various factors (e.g., in vivo blood flow variations) [10,18]. If we take these limitations into account, the use of mT/N1 is simple and gives us the possibility of quick evaluation of conditions of planned 90 Y treatment. The mT/N2 ratio, which is calculated after radioembolization, would potentially be much more helpful because it reflects the actual 90 Y dose distribution within liver tumors. Unfortunately, mT/N2 (p = 0.057) is not adequate as a prognostic factor, although its p-value The predictive value of SPECT/CT imaging in mCRC response after 90 Y-radioembolization is close to statistical significance and a study with a larger population might give positive results. It is worth mentioning that AAD calculated using mT/N2 showed significant correlation with treatment results.

Absorbed dose evaluation
Based on mT/N1 and mT/N2 it is possible to calculate PAD and AAD for tumors and liver absorbed dose (LAD) [7,14]. As we observed in the multiple correspondence analysis, the magnitude of predicted or actual absorbed dose is a major factor determining the tumor response ( Fig 5). The mean PAD was 95 Gy, and PAD showed a strong correlation with tumor response. TTP tumor was longer when PAD was higher than 70 Gy. There were 75 tumors with PAD higher than 70 Gy (minimal absorbed dose for tumor destruction). Within this group, there were 7 lesions with CR (all in the entire study) and only one lesion with PD. Interestingly, the median PAD was higher in tumors with PR than with CR. Complete destruction of tumors with lower PAD may be partially explained by a more uniform distribution of microspheres in smaller lesions. Larger, partially necrotic tumors are more resistant, even if they absorb higher 90 Y doses. The mean AAD was 84 Gy, and AAD was also a predictor of tumor response. Moreover, in terms of AAD all 5 tumors with PD were below the threshold of 70 Gy, whereas for PAD only 4/5 were below this threshold. It makes sense that AAD would better predict tumor response, because this value is calculated after SIRT treatment and reflects the actual 90 Y-microspheres deposition. We should, however, remember that AAD is determined by the mT/N2 ratio, which was not statistically significant in our study. PAD remains only value calculated before treatment that may predict the response to treatment.
As shown in the Results section, the difference between PAD and AAD in non-responding tumors was not statistically significant. A possible explanation is that non-responders fail to The predictive value of SPECT/CT imaging in mCRC response after 90 Y-radioembolization absorb sufficient doses in both the preparatory ( 99m Tc) and treatment phases ( 90 Y). In our opinion, the prognostic evaluation of SIRT treatment should be undertaken using all dosimetric parameters, because there is no single prognostic factor that is strongly and reliably predictive of treatment results. Other factors, for example tumor size or volume, may also be important. Consistent with previous reports, our study shows that tumors smaller than 20 mm gather higher mean absorbed doses than larger tumors and show a better response to radioembolization [2,19,20].
Information about 90 Y-microsphere distribution within the liver may be a key to improvement of SIRT treatment results. Using this information, patients prone to potential radiationinduced liver disease (RILD) or patients who accumulate non-curative tumor doses may be identified immediately after the SIRT procedure. Moreover, these data may be useful during treatment planning for much more selective tumor destruction. In other words, mT/N1 and PAD are important, because they allow changes in radioembolization before the procedure itself. Several papers have shown the usefulness of the T/N ratio in prediction of SIRT treatment results. This applies especially in hepatocellular carcinoma, which is known for its developed tumor vasculature and high T/N ratio [21][22][23]. In comparison to hepatocellular carcinoma, colorectal metastases are characterized by low blood flow and a lower T/N ratio, so there is no easy way of applying dosimetric parameters for assessing tumor response after treatment [18,24].
As we can observe in previous reports, there are several methods of measuring dosimetric data in SPECT/CT. In the partition model, analysis of 99m Tc-MAA distribution between healthy liver and tumor compartments is performed on SPECT images (2D slices reconstructed from 3D projections of the organ), and the activity of 90 Y is calculated [1,7,10,18,22,25]. In some cases, it is difficult to establish proper tumor ROIs (especially in liver with multiple lesions). Moreover, the presence of "hot spots" does not mean there is a tumor in each case; these may be healthy liver areas with higher macroalbumin accumulation. Some authors have used ROIs (or VOIs) with strictly defined size (e.g., 1 cm 2 or cm 3 , smaller than the tumor size) and defined them on the basis of the researcher's impressions), while others have administered angiotensin shortly before radioembolization, which should have a significant influence on microsphere deposition in the liver and would considerably alter the pre-treatment procedure [18]. In another study, the authors used only a visual (qualitative) scale of 90 Y deposition [26]. An interesting new proposal is to use functional dual-tracer SPECT imaging with fusion 99m Tc-MAA and 99m Tc-Sulphur colloid with semiautomatic tumor segmentation. A study performed by Lam on 122 patients proved the usefulness of 99m Tc-MAA-based dosimetry in prediction of radioembolization results [27]. Within this group, there were 29 patients with metastatic colorectal cancer (which corresponds to 58 tumors investigated based on the RECIST 1.1 criteria). The mean tumor absorbed dose in this study was 30 Gy, although the generally accepted therapeutic dose is much higher, i.e., 70 Gy. This discrepancy may be explained by differences between anatomo-functional and functional methods of gamma count measurement [27,28]. This implies that there is no simple and direct comparison between these methods. Moreover, the highly sophisticated and time-consuming process of data gathering is in our opinion a barrier to the widespread use of this method.
The above-mentioned methods are based on a standard approach to calculating the tumorto-normal-liver ratio. The sT/N1 and sT/N2 ratios were proved to be of limited predictive value of tumor response in our study, as they were in majority of previous reports [1,7,10,18,22,25]. We have presented a simpler method of dosimetric calculation on a similar cohort of patients (21 patients, 103 tumors). In our opinion, the advantages of gamma count measurement from the scan section with the largest diameter are simplicity, repeatability and representativeness. Our method for calculating the tumor to liver ratio proved to be a valuable tool for predicting treatment results and could serve as a basis for further dosimetric calculations. The resulting information about tumors dosimetric parameters obtained prior to treatment should improve treatment planning, and may help in selecting tumors with insufficient PAD. At this stage, it may be possible to calculate a new slightly higher level of 90 Y-microsphere activity that may be sufficient to destroy all tumors and still be safe for liver tissue. The alternative approach is to plan much more selective 90 Y-microsphere injection within tumor areas with PAD lower than 70 Gy to achieve a curative 90 Y dose. Knowing about insufficient AAD on the day of radioembolization may also allow immediate planning of treatment using other methods (especially radiofrequency ablation), focused only on resistant tumors.
In this study, when we focus on the liver absorbed dose, both predicted and actual doses were almost the same and no serious adverse events were observed. As we can see in the standard guidelines, the limit of liver tolerance is 30 Gy, although some authors have reported that higher doses (up to 70 Gy) were well tolerated by liver tissue [4,5,14].
Our study has several limitations. First, we assumed uniform distributions of macroalbumin and microspheres within the liver and tumor compartments. Some tumors had central necrosis areas that did not accumulate the tracer, but these were not excluded from the analysis. Second, we hypothesized that if the method proved to be applicable for varied tumor morphologies, then in a more uniform group, the prognostic value of the results would be greater. Third, even though ROC curve analysis suggested that the cutoff point for PAD should be 87 Gy, we performed calculations using the generally accepted threshold of 70 Gy. Fourth, in accordance with our assumptions regarding the uniformity of dose distribution, gamma counts from the middle (widest diameter) section closely matched those obtained from the entire sphere.

Conclusions
The mT/N1 ratio and PAD based on 99m Tc-MAA SPECT/CT and AAD calculated from 90 Y-SPECT/CT can be used as predictors of radioembolization results. TTP tumor was significantly longer with mT/N1 greater than 1.7 and for tumors with PAD greater than 70 Gy. The risk of progression was elevated for tumors with mT/N1 lower than 1.7 and PAD lower than 70 Gy. The mT/N2 ratio had no significant correlation with treatment results.
Supporting information S1