Hybrid Imaging Reveals Improved Absorbed Dose from Selective Internal Radiation Treatment by Using an Anti-Reux Catheter

Selective internal radiation therapy (SIRT) is a promising technique for patients with hepatic malignancies. Several image-based investigations, e.g. volumetric and absorbed dose assessment, are mandatory for SIRT planning and treatment veri�cation based on national and international regulations. General treatment work�ows are described in guidelines, recommendations, and the package inserts of the manufactures. But, guidance to tackle particular clinical conditions can be ill-dened and different centers practice their own work�ow to analyze the treatment process. This case report includes an example of inconsistency between treatment simulation and observed treatment result, revealed by hybrid imaging. There is no universally accepted standard procedure de�ned in the literature for detecting and evaluating a possible mismatch between [ 99m Tc]Tc-MAA-based simulation and distribution of the therapeutic microspheres. In this setting, more advanced multi-modal image-based analysis may be bene�cial. A 78 year old patient with hepatocellular carcinoma underwent liver radioembolization with resin 90 Y-microspheres. Tumoral and non-tumoral dose–volume histograms were evaluated for simulated activity distribution using [ 99m Tc]Tc-MAA-SPECT and post-treatment activity measurement using 90 Y-PET. During simulation workup, [ 99m Tc]Tc-MAA particles were administered using a regular catheter. On the other hand, for the treatment session an anti-reux catheter was used. Our result, suggests that the use of an anti-reux catheter might improve tumor coverage, and as a result decrease non-tumoral liver uptake deposition.


Introduction
A work ow for selective internal radiation therapy (SIRT) using resin 90 Y-microspheres (SIR Spheres, SIRTEX, Sydney, Australia) is described in the package insert and training program of the manufacturer [1], guidelines and recommendations by different associations [2,3].In SIRT, the absorbed dose (i.e. the deposited energy per unit of mass) is used to quantify the irradiation of the tumor and non-tumoral healthy liver [4].
In the daily clinical practice, each patient has unique clinical characteristics and needs, necessitating a dosimetrybased tailored activity planning strategy, and post-treatment dose veri cation [4].This requires (1) a careful treatment planning using a simulation workup to estimate the intra-and extrahepatic distribution of the therapeutic microspheres, and (2) a post-treatment determination of the radionuclide distribution to identify potential adverse effects (treatment e cacy and safety).
In our previous studies, a personalized multi-modal dosimetry work ow was introduced [5, 6], and some patient population studies on comparing voxel-level pre-treatment predictive and post-treatment measured absorbed dose were presented [7].Our proposed work ow uses automated segmentation and non-rigid image registration.In daily clinical routine, we apply this work ow in parallel to the standard clinical work ow, which relies on manual segmentation and visual inspection of all images.
Here, we describe a case to illustrate the usefulness of multi-modal image analysis, e.g.volume of interest (VOI) de nition based on cone-beam CT (CBCT), in comparing predicted and measured dose.This case illustrates the value of using an anti-re ux catheter in SIRT for better tumor coverage and resulting e cacy/toxicity balance [8].

Case Description
MRI and CT images, a large central (diameter: 8.6 cm) and two satellite tumors, involving segments II and IV, were diagnosed.The case was subsequently discussed at the multi-disciplinary tumor board.The tumor load was considered too extensive for transarterial chemoembolization (TACE).Hence, SIRT was considered as an alternative aiming at controlling tumor progression.
The treatment was planned based on a pre-SIRT workup using technetium-99m macro-aggregated albumin ([ 99m Tc]Tc-MAA) as a surrogate for therapeutic microspheres.During the workup, a super-selective catheterization of the hepatic artery (HA) was performed, and the perfusion territories of the left and right HA were evaluated.Dualphase CBCTs indicated the bulky tumor in segment IV and two other smaller tumors.The tumors were irrigated from both the left and right HA.An accessory branch to segment I did not contribute to the blood ow of the tumors.Afterwards, a [ 99m Tc]Tc-MAA-SPECT/CT was performed, which predicted poor targeting, particularly within the right liver perfusion territory (LPT).In this LPT, almost no difference was found between the ventrally located tumor and dorsal healthy liver.An estimated lung shunt fraction (eLSF) of 4.8% was calculated.
The treatment was planned a month after [ 99m Tc]Tc-MAA workup by using the Medical Internal Radiation Dose (MIRD) formalism with an entire liver absorbed dose of 50 Gy (equivalent to 1 MBq/ml infused volume).The therapeutic microspheres were administered in a comparable manner to the workup, except for one difference; an anti-re ux catheter (Sure re Infusion System; Sure re Medical, Westminster, Colorado) was used during the treatment session.This was done because a very narrow right gastric artery did branch from the left hepatic artery just after its origin; this vessel was too narrow to insert a catheter and occlude it with coils.
We hypothesized that the microsphere distribution with the use of an anti-re ux catheter could be different, in an unpredictable manner, compared to a regular micro-catheter.Interestingly, on the day of the treatment, preferential tumor targeting was observed.We performed detailed dosimetry, particularly with respect to predictive versus measured dose comparison.
Three months after treatment, an arterial and venous phase contrast-enhanced CT imaging and contrast-enhanced MR images were performed.The two satellite lesions showed a marked partial response (according to modi ed RECIST; mRECIST).For the big tumor, no volume increase was observed, and it was as also a partial response (mRECIST) due to its increasing central necrosis (see Fig. 1).

Dosimetry Report
By using a projected absorbed dose of 50 Gy (or 1 MBq per ml) to the entire liver, including tumor volume (TV) and non-tumor volume (NTV), 0.522 and 1.027 GBq of therapeutic microspheres were prescribed for the left and right liver LPT, respectively.
Fractional uptakes were derived from the [ 99m Tc]Tc-MAA-SPECT and 90 Y-PET for the predictive and post-treatment dosimetry, respectively.We performed dosimetry based on the local deposition model by utilizing these VOIs [7]: liver: the liver was delineated by applying a convolutional neural network on the CT image from the [ 99m Tc]Tc-MAA workup with some manual adjustment [6], tumor: tumors were segmented on cone-beam CTs (CBCTs), LPTs: left and right LPTs were segmented on the CBCT images.
For the post-treatment dosimetry, all the VOIs were aligned to the post-treatment 90Y-PET using non-rigid registration [5,7].A comparison between segmented volumes used in the pre-and post-treatment dosimetry, together with the manual segmentation (which was used for prescription) are provided in Table 1; screenshots of three different transverse slices of various images, as well as de ned VOIs are shown in Fig. 2.

Discussion
In SIRT, the pre-treatment workup is the basis for the treatment planning.This investigation consists of two essential imaging studies: (a) an angiography to assess the gross hepatic vascular anatomy and to investigate tumoral perfusion.In most centers nowadays, this angiography is followed by a dual-phase contrast-enhanced CBCT (early and late arterial phase), focusing on the perfusion territory of each catheter position.These images also indicate the hyper-or hypovascular nature of the tumor(s), (b) a planar imaging and SPECT/CT imaging after administration of surrogate particles ([ 99m Tc]Tc-MAA) at each proposed catheter position.We use these images to predict the intra-and extrahepatic distribution of the therapeutic microspheres.
A good predictive power of the [ 99m Tc]Tc-MAA simulation in SIRT treatment planning was demonstrated in some studies.Strigari and Flamen reported a good correlation between [ 99m Tc]Tc-MAA particles and therapeutic microsphere distribution for 80% and 100% of the cases, respectively [8,9].Kao et al. con rmed this correspondence later for tumor absorbed dose [10,11].
On the other hand, the inaccuracy in the prediction made using [ 99m Tc]Tc-MAA has been reported in some studies.
Possible reasons for a mismatch between [ 99m Tc]Tc-MAA and therapeutic microspheres are: (i) the discrepancy between [ 99m Tc]Tc-MAA and therapeutic microsphere in shape, size, and size distribution [12], the number of administered particles/microspheres [13], in vivo stability, and infusion rate, (ii) the administration close to an arterial bifurcation or a small arterial branch [12], (iii) temporary vessel spasm during the pre-treatment workup [14], and (iv) the difference in catheterization, such as the catheter tip positioning and the used catheter type [15].
In our previous study, we compared pre-treatment dose estimation and post-treatment dose measurement.In this study, VOIs were segmented using CBCT images, which was demonstrated to improve the accuracy of LPT delineation compared to anatomical lobe segmentation using contrast-enhanced CT images.We demonstrated a stronger agreement for NTV compartment and total LPT than for the TV.
Fortunately, 90 Y-PET imaging revealed a more extensive tumor uptake with greater TV absorbed dose than that predicted by the workup (by a factor of 2.0), for the described patient.As a result, decreased activity deposition in the areas of NTV was measured compared to the pre-treatment workup.As the catheterization was similar for this patient, we believe that this large mismatch between [ 99m Tc]Tc-MAA and therapeutic microsphere activity distributions was due to the use of anti-re ux catheter in the therapy session.Pasciak et al. also reported a similar effect with the use of an anti-re ux catheter, with an increase of tumor uptake with on average of 68% (range 33-90%) in 9 patients when comparing to an infusion with a conventional end-hole catheter [16].

Conclusion
Here, we present a case for which a multi-modal image processing facilitated the use of more information from all available images (e.g.CBCT, [ 99m Tc]Tc-MAA-SPECT/CT, and 90 Y-PET/MR) to create more detailed reports.These details can help the physician to extract clinically relevant information, maximizing both therapy e cacy and safety.
MIRD Medical Internal Radiation Dose.

Figures
Figure 1 [A] MR image before SIRT, performed 1.5 month before the treatment session, [B] a follow-up MR performed three months after the SIRT session.These images demonstrate the appearance of a prominent central necrosis (hypointense; see red arrow).

Table 1
Segmentation analysis; comparing our segmentation results and clinical records.thepredictivedosimetryandmeasured dose distribution is provided in Table2.Cumulative dose-volume histogram are also provided in Fig.3.The voxel-level dosimetry revealed that: TV mean dose in the post-treatment is notably higher than the pre-treatment evaluation for the right LPT (72 versus 189 Gy).In the left LPT, the underestimation of the TV dose by [ 99m Tc]Tc-MAA was less notable (82 versus VOI : volume percentage of the VOI that receive at least X Gy

Table 2
Overview of the predictive and measured dosimetry comparison.