Following patient selection, tumor and perfused volumes were segmented on baseline contrast-enhanced CT and dual-phase CBCT and co-registered to [99mTc]-MAA SPECT/CT. The values obtained on the different imaging modalities were used for treatment simulation and then were compared to those obtained on reference standard post-embolization 90Y PET/CT. Details are further described in each section (Fig. 1).

Fig. 1

Consecutio of the different study phases: tumor and perfused volume segmentation, the derived volumes co-registered with [99mTc]MAA SPECT for treatment simulation and verification with 90YPET

Patient population

Thirty-six lesions of 30 HCC patients that underwent TARE with resin 90Y-microspheres (SIR-spheres® SIRTex Medical Limited) from January 2020 to August 2023 were collected prospectively and retrospectively evaluated. Pre-procedural contrast-enhanced CT (CECT) performed within 30 days from procedure was primary inclusion criteria. Therapeutic indication for TARE was discussed and agreed in the dedicated multidisciplinary team prior to treatment. Patient data was collected after having signed a specific informed consent previously approved by the Institutions’ Ethical Committee (64/INT/2021).

Radioembolization protocol

Workup and treatment were performed following the current standard of practice at our Institution.

Radioembolization was divided into two sessions performed approximately one week apart. During the first session, mapping angiography and intraprocedural dual-phase CBCT was performed. The acquisition protocol was divided in an early and late arterial phase, to identify, respectively, the target lesion, its feeders and surrounding perfused parenchyma. Upon correct arterial access selection, [99mTc]macroaggregated albumin was injected to evaluate lung and gastrointestinal shunt fraction and Tumor to Normal tissue uptake ratio (TNR) [13]. 90Y activity calculation was based on a multicompartmental model [16], considering a tumor absorbed dose > 120 Gy, overall normal liver dose < 40 Gy and a lung dose < 30 Gy [9] [13] [16]. Prior to administration of 90Y-spheres, patients underwent a confirmatory angiography and CBCT to evaluate any change in perfusion. Standard of care confirmatory 90Y PET/CT study is acquired within 24 h after treatment.

Imaging acquisition protocolsContrast-Enhanced CT

Upper and lower abdominal scans were acquired on multidetector CT scanners (Siemens Somatom definition flash Syngo or Philips Brilliance 64). The scan included a non-contrast and a triple-phase acquisition to evaluate focal liver lesions (late arterial, portal and delayed). The arterial phase (30 s) was used to evaluate viable tumor tissue and vessels anatomy and to calculate tumor volume (TV).

Dual-Phase Cone-Beam CT Technique

Acquisition of images was performed using an angiographic system (Azurion 7 C20, Philips Healthcare) equipped with the XperCT Dual option, that allows dual-phase rotational acquisition (early and late arterial phases) with a single contrast injection. The C-arm rotation in head-end position (120°LAO- 185°RAO) reaches 25°/sec at a voltage of 120 kVp with a detector size of 12 inches; total rotation time is 5.2 s.

A specifically developed contrast agent injection protocol (Ultravist, 370 mg of iodine per milliliter) was standardized to deliver 10 mL at a rate of 1 mL/second in segmental branches of the hepatic artery and acquisition was performed during end expiration apnea. In cases of superselective catheterism of subsegmental branches, the injection protocol ratio remained unvaried, at a rate of 0,5 mL/second for 5 mL administered. The early arterial phase was triggered at 10 s after contrast injection and used to evaluate tumor volume; the late arterial phase, at 15–20 s after the first acquisition, for perfusion volume (Fig. 2). This injection protocol allows evaluation of both the feeding arteries and the hyper-vascular lesion/s as contrast lasts throughout the first acquisition. The perfused parenchyma is then visible in the late arterial phase (approximately at a delay of 30 s) (Fig. 3). The low rate of injection reduces the risk of contrast reflux to non-target areas. Automated image reconstruction is visible after completion of each scan.

Fig. 2

Dual-Phase CBCT acquisition protocol

Fig. 3

Dual-phase CBCT: a Early arterial phase (10–15 s) Lesion 1; b Late arterial phase (30 s) Lesion 1; c Early arterial phase (10–15 s) Lesion 2 and 3; d Late arterial phase (30 s) Lesion 2 and 3

[99mTc]macroaggregated albumin-SPECT/CT ([ 99mTc]-MAA SPECT/CT

Whole body planar scintigraphy (from thyroid to bladder) was acquired in anterior and posterior projections followed by a SPECT/CT scan (128 × 128, 120 steps, 20 s/step Discovery NM 670, GE Healthcare) within 30 min after conclusion of the angiography procedure. Planar images were used to define lung shunt fraction and presence of gastrointestinal shunt. SPECT data was obtained with a 3D ordered-subset expectation maximization (4 iterations, 10 subsets) with a Butterworth filter (cutoff,0.5 cycles/cm; order, 10) and CT-based attenuation, scatter correction and resolution recovery.

90Y PET/CT

Two bed positions of 15 minutes each, including complete liver coverage, were acquired on a Time Of-Flight (TOF) PET/CT system (Discovery PET/CT 690, GE Healthcare). PET data was reconstructed using a Bayesian reconstruction algorithm (VPFX, 2iterations 16 subsets) with a gaussian post-reconstruction filter of 5 mm in full width at half maximum including time of flight information. Normalization, dead time, activity decay, random coincidences, attenuation, scatter and resolution recovery (imaging corrections) were used.

Tumor and perfused volume delineation

All tumor- and perfused volumes (TV, PV) and overall healthy total liver volume were segmented using MIM software (Beachwood, Ohio- 7.2.8). Segmentation was performed using a semi-automated method which was then corrected manually, if needed.

TV on CECT and PV and TV on CBCT were evaluated by two interventional radiologists (EDG and LA). TV and PV from [99mTc]MAA SPECT/CT and 90Y PET/CT were evaluated by a nuclear medicine physician, using a threshold method.

Tumor volumes, defined as the viable tumoral tissue showing uptake in arterial phase, were retrospectively segmented on the arterial phase of CECT and early phase of intraprocedural CBCT. Necrotic areas were excluded from volume delineation.

TVs were also delineated on [99mTc]MAA SPECT/CT and 90Y PET/CT images using a threshold method. Delineation was carried out on the fusion images (SPECT and CT, PET and CT) and for each volume, the threshold value was visually adjusted in order to match the contours based on the activity distribution with the CT lesion borders. Mean threshold value for TVs on [99mTc]MAA SPECT/CT and 90Y-PET/CT images was 30% [16].

Perfusion volumes (defined as the parenchyma of the segment vascularized by the arterial branch selectively catheterized) including the lesion and the surrounding healthy parenchyma enhanced following contrast injection, were segmented on one or more (in case of multiple feeding arteries) late arterial phases of intraprocedural CBCT acquisition.

Due to the broad distribution of single phase CBCT protocols, PV was also segmented in the early phase to confirm accuracy compared to the late arterial phase. In addition, PVs were delineated on [99mTc]MAA SPECT/CT and 90Y PET/CT images using a threshold method. The mean threshold value for PVs on [99mTc]MAA SPECT/CT and 90Y PET/CT images was 5% [7].

Treatment planning simulation and verification

TVs and PVs delineated on CECT and CBCT were transferred to the SPCET/CT images using non-rigid co-registration. Treatment planning simulations were performed independently considering volumes segmented on the three imaging modalities: CECT, CBCT and [99mTc]MAA SPECT/CT. For CECT alone, the overall healthy total liver was used to calculate perfused volume toxicity.

Fixed tumor- and perfused parenchyma doses were set at 120 Gy and 40 Gy, respectively. Simulations of PET dose distribution were evaluated by normalizing PET images to the three fictitious activities derived from the treatment planning obtained with volumes delineated on the three imaging modalities. All phases of the treatment planning simulation were performed using the MIM software.

Comparisons

90Y PET/CT volumes were considered as reference standards. TVs obtained on CECT, CBCT early arterial phases and [99mTc]MAA SPECT/CT were compared to those obtained on post-embolization 90Y PET/CT. PVs from CBCT early and late arterial phases and [99mTc]MAA SPECT/CT were compared to those obtained on post-embolization 90Y PET/CT. Early arterial dual-phase CBCT, CECT and [99mTc]MAA-SPECT/CT were compared head-to-head using a one-way ANOVA. In order to evaluate the agreement of the predicted dosimetry derived from the treatment planning simulations and post-treatment dosimetry, the dosimetric metrics considered were V120Gy, defined as the percentage of target volume receiving at least 120 Gy, and V40Gy, percentage of healthy parenchyma volume receiving 40 Gy [9].

Statistical analysis

Statistical analysis was performed using MedCalc software (Medcalc Software Ltd, Ostend, Belgium – v20.216). The coefficient of determination R2 of the regression model was calculated to compare PVs and TVs derived in CECT, CBCT and [99mTc]MAA SPECT/ to those obtained on post-embolization 90YPET/CT. Interobserver agreement (for tumor and perfused volume) on all imaging was evaluated using the Dice coefficient. Correlation of volumes obtained between techniques is depicted using Bland–Altman plots. Quantitative descriptive statistics are expressed as medians and interquartile ranges; tumor size is reported in mean and standard deviation. A p-value less than 0.05 was considered significant. Head-to-head comparisons among modalities was evaluated with one-way ANOVA. Agreement between V120Gy and V40Gy obtained in CECT, CBCT, SPECT, values were calculated on PET normalized imaging. Concordance correlation coefficient (CCC) for dose comparison were calculated with a 95% Confidence Interval (CI).