Post-Treatment PET/CT Dosimetry to Predict Contralateral Lobe Hypertrophy After Transarterial Radioembolization for Hepatocellular Carcinoma

Abstract

Objective

Image-based dosimetry in transarterial radioembolization (TARE) has been proposed for the prediction of contralateral lobe hypertrophy as well as tumor response. This study aimed to evaluate the predictive value of post-treatment ⁹⁰Y PET/CT voxel-based dosimetry for post-TARE liver hypertrophy in treatment-naïve patients with hepatocellular carcinoma (HCC).

Materials and Methods

This retrospective single-center study included 40 treatment-naïve patients with right lobar HCC (age, 66.4 ± 12.5 years; male:female, 29:11) who underwent TARE with ⁹⁰Y-labeled glass microspheres. Voxel-based absorbed dose maps were generated from ⁹⁰Y PET/CT performed 16–24 hours post-treatment; dose histograms were obtained for the whole liver, right lobe (RL), and non-tumor RL (NRL). Mean absorbed dose and volume proportion receiving ≥50 Gy (P₅₀) and ≥100 Gy (P₁₀₀) for each region were calculated. All patients underwent follow-up CT-2–4 months after treatment, and the percentage increase in left lobe (LL) volume was measured using CT volumetry. The efficacy of dosimetric parameters and conventional clinical factors in predicting LL hypertrophy was analyzed using Spearman’s correlation and multivariable regression.

Results

The median LL volume increase was 12.7% (interquartile range, 5.7%–28.1%). The mean absorbed dose (ρ = 0.662, P < 0.001) and P₅₀ (ρ = 0.672, P < 0.001) of NRL showed the strongest correlations with LL hypertrophy. In multivariable regression integrating clinical factors, the mean NRL dose was the only independent dosimetric predictor (R² = 0.556, P < 0.001); platelet count also remained independently predictive (R² = 0.221, P = 0.003). The relationship between NRL dose and LL hypertrophy was markedly stronger in patients with preserved platelet counts (platelet ≥130 × 10³/µL; ρ = 0.847; P < 0.001) than in those with decreased platelet counts (ρ = 0.429; P = 0.129).

Conclusion

Post-TARE ⁹⁰Y PET/CT dosimetry, particularly in combination with platelet count, is an effective approach to predict contralateral hypertrophy in treatment-naïve patients with HCC. Further studies with larger prospective cohorts are required to validate these predictive strategies.

INTRODUCTION

The volume of the future liver remnant (FLR) determines the surgical eligibility and postoperative morbidity of patients with hepatic tumors [1,2,3,4]. The risk of posthepatectomy liver failure can be minimized using strategies that enhance contralateral lobe hypertrophy (CLH), such as portal vein embolization (PVE) [5,6,7]. Transarterial radioembolization (TARE), wherein ⁹⁰Y-labeled microspheres are injected into tumor-feeding arteries, is a bridging treatment to promote CLH and control tumor growth.

TARE initially emerged as a tumor-directed therapy for hepatic tumors in patients ineligible for surgical resection or transplantation [8,9,10]. It later gained recognition for its ability to produce CLH and to serve as a bridging therapy for patients with an insufficient FLR [11,12,13]. This effect is presumed to result from the radiation-induced atrophy of the treated lobe, leading to compensatory hypertrophy of the contralateral lobe [14,15]. Because the surgical feasibility and extent of resection in patients with insufficient FLR depend on the degree of hypertrophy, reliable biomarkers are required to predict FLR following TARE.

Dose requirements to induce CLH have traditionally been guided by pretreatment planning using ^99mTc-macroaggregated albumin (MAA) imaging [16,17]. Although not derived from large-scale evidence, some practice guidelines have recommended dose thresholds of >70 Gy with resin microspheres and ≥88 Gy with glass microspheres to the non-tumor liver, based on ^99mTc-MAA dosimetry [18,19,20].

Recently, post-treatment ⁹⁰Y PET has been increasingly used to confirm the distribution of injected radioactivity and to quantify the actual delivered dose after TARE [21,22,23], offering a more patient-specific dosimetric assessment compared to pre-treatment simulations. The ^99mTc-MAA distribution observed during planning angiography might differ from the actual ⁹⁰Y-microsphere distribution [24]. Additionally, the contribution of radiation from tumors to the adjacent non-tumor parenchyma was underestimated, considering the long beta particle range of ⁹⁰Y (11.8 mm). Therefore, the radiation dose measured on post-treatment ⁹⁰Y PET may provide distinct information for the prediction of CLH after TARE. Despite this advantage, the predictive value of ⁹⁰Y PET-based dosimetry for CLH has been explored in only a few studies. Prior studies have evaluated its predictive utility in patients treated with resin microspheres [25] or in mixed cohorts of glass and resin microspheres [26]; these studies were conducted in populations with heterogeneous tumor pathologies.

This study aimed to evaluate the effect of the radiation dose calculated by voxel-based ⁹⁰Y PET/CT dosimetry on CLH in a homogeneous cohort of treatment-naïve patients with hepatocellular carcinoma (HCC) treated with glass microspheres.

MATERIALS AND METHODS

Patients and Treatment

We retrospectively reviewed data of patients diagnosed with HCC who underwent TARE between May 2022 and December 2024 at a single institution. Patients were included if they met all of the following criteria: 1) no prior treatment for HCC, 2) a single nodular HCC in the right lobe (RL), 3) TARE using glass microspheres (TheraSphere^®, Boston Scientific, Marlborough, MA, USA) with microsphere distribution confined to the RL, 4) available follow-up contrast-enhanced CT scan performed 2–4 months after the TARE, and 5) no other treatment for HCC between TARE and the follow-up CT scans (Fig. 1). Patients were excluded if they had acute inflammatory conditions in the liver or developed new tumors during the follow-up period. Forty patients (age, 66.4 ± 12.5 years; male:female = 29:11) were finally included. This study was approved and the requirement for written informed consent was waived by the Institutional Review Board (IRB No. H-2504-019-1628).

Fig. 1. Overall flow of the study. Inclusion and exclusion criteria for retrospective patient enrollment and the overall study flow are summarized. TARE = transarterial radioembolization, HCC = hepatocellular carcinoma, VOI = volume of interest.

TARE was performed by experienced interventional radiologists, as previously described [27]. Briefly, pretreatment gamma camera scanning was conducted after injecting ^99mTc-MAA into the tumor-feeding arteries for treatment planning. Based on this scan, anticipated absorbed doses for the normal liver, tumors, and lungs were calculated. TARE was performed with a boosted dose to ensure a predicted tumor dose of >150 Gy, while maintaining the lung dose of <25 Gy (male) or <20 Gy (female) and preserving the left lobe (LL) from treatment exposure. Because the primary purpose of the procedure was tumor control, additional dosing to the non-tumor liver to induce CLH was usually not performed.

Clinical Factors

Clinical factors related to demographic characteristics and liver conditions were collected through a review of hospital records. These factors included age, sex, hepatitis virus status, presence of hepatic encephalopathy, ascites, splenomegaly, portal vein thrombosis, and maximal tumor diameter. Ascites, splenomegaly, and tumor diameter were assessed on the basis of the most recent pretreatment CT scan, with splenomegaly defined as a splenic length of >12 cm. Liver function-related laboratory parameters included serum aspartate aminotransferase (AST), alanine aminotransferase (ALT), total bilirubin, albumin, platelet count, and the international normalized ratio of prothrombin time (PT INR). Additionally, data on tumor markers such as alpha-fetoprotein (AFP) and proteins induced by vitamin K absence or antagonist-II (PIVKA-II) were recorded.

Post-Treatment ⁹⁰Y PET/CT Imaging

⁹⁰Y-microsphere PET/CT imaging was performed 16–24 hours post-treatment. Non-enhanced CT and PET images (acquisition time: 10 minutes for a single-bed position) were acquired using a dedicated PET/CT scanner with an axial field of view of 216 mm (Biograph mCT 64; Siemens Healthineers, Erlangen, Germany) to cover the entire liver. PET images were reconstructed using an ordered subset expectation maximization algorithm (2 iterations and 21 subsets) by applying time-of-flight and point-spread function recovery techniques. CT-based attenuation correction, scatter correction, and post-reconstruction Gaussian filter (full width at half maximum; FWHM 5 mm) were implemented. The matrix size was set to 200 × 200, and the voxel size was 4 × 4 × 5 mm³.

CT Liver Volumetry and Follow-Up

CT volumetry and voxel-based PET dosimetry were performed using a dedicated analysis software (SurePlan LiverY90, MIM Software Inc., Cleveland, OH, USA) by a nuclear medicine physician (Y.K.). Baseline contrast-enhanced CT was performed within 30 days of TARE. The volumes of the RL, tumor, and LL were delineated on the portal-phase images. A region of interest (ROI) was manually drawn on each transaxial plane slice and interpolated to create a single volume of interest (VOI). The volume of the non-tumor RL (NRL) was determined by subtracting the tumor volume from the RL volume. The same measurements were performed on follow-up CT scans. The post-treatment LL volume increase was the primary endpoint indicating CLH. The VOIs defined in the pretreatment scans were also used for subsequent PET dosimetry analyses.

PET Dosimetry

Pretreatment contrast-enhanced CT images and the corresponding VOIs of the normal liver and tumor were fused with ⁹⁰Y-microsphere PET/CT images using rigid and deformable registration techniques. Voxel-wise absorbed-dose maps were generated on the fused images using the voxel S-value (VSV) kernel approach. Briefly, cumulative activity maps were derived from ⁹⁰Y-microsphere PET images based on their physical half-lives, and these cumulative activity maps were convoluted with the VSV kernel to produce absorbed dose maps. The VSV represents the mean absorbed dose of a target voxel per radioactivity in a source voxel. On this dose map, the voxel dose distributions for each VOI (RL, LL, tumor, and NRL) were assessed using dose histograms. For each VOI, we calculated the mean absorbed dose and the proportion of voxels receiving >50 Gy and >100 Gy, denoted as P₅₀ and P₁₀₀, respectively.

Statistical Analysis

Clinical parameters were reported using numbers and proportions for categorical variables and median values and interquartile ranges (IQR) for continuous variables. LL volume changes according to group stratified by categorical variables were evaluated using the Mann-Whitney U test. Associations among continuous clinical variables, dosimetric parameters, and LL volume changes were evaluated using Spearman rank correlation coefficients. Multivariable regression analysis was used to evaluate clinical and dosimetric parameters for the prediction of CLH. Multivariable analysis initially included variables that demonstrated significance in group comparisons or univariable regression analyses using a stepwise variable selection approach. P-values <0.05 were considered significant. Analyses were performed using MATLAB R2024a (MathWorks, Natick, MA, USA) and R (version 4.3.2; https://www.R-project.org).

RESULTS

Patients

Patient characteristics are summarized in Table 1. All patients were diagnosed with treatment-naïve HCC and underwent TARE using glass microspheres. Twenty-two patients (55.0%) had hepatitis B, and five (12.5%) had hepatitis C. Twenty-five patients (62.5%) had liver cirrhosis, but all patients had Child-Pugh scores of 5 or 6. Portal vein thrombosis was observed in 4 patients (10.0%), all of whom had partial thrombosis, with no cases of total occlusion. All tumors were single nodules and were located within the RL. The median tumor size was 5.9 cm (IQR, 3.8–7.3 cm).

Table 1. Clinical characteristics of patients.

Characteristic		Data
Pathology
	Hepatocellular carcinoma	40 (100)
Age, yr		66.4 ± 12.5
Sex, male:female		29 (72.5):11 (27.5)
Etiology
	Hepatitis B	22 (55.0)
	Hepatitis C	5 (12.5)
	Others	13 (32.5)
Hepatic encephalopathy		0 (0)
Ascites		4 (10.0)
Splenomegaly		15 (37.5)
Portal vein thrombosis		4 (10.0)
Child-Pugh score
	5	34 (85.0)
	6	6 (15.0)
Serum AST, U/L		30.5 (24.0–45.5)
Serum ALT, U/L		22.5 (16.0–33.0)
Serum total bilirubin, mg/dL		0.6 (0.4–0.9)
Serum albumin, g/dL		4.1 (3.9–4.4)
Platelet count, 103/µL		161.5 (112.5–203.0)
PT INR		1.07 (1.02–1.14)
AFP, ng/mL		13.3 (4.9–538.5)
PIVKA-II, mAU/mL		115.5 (29.0–2116.5)
Tumor diameter, cm		5.9 (3.8–7.3)
Left lobe volume change, %		12.7 (5.7–28.1)
Radioactivity of injected microspheres, GBq		2.9 (2.0–4.4)
Mean absorbed dose, Gy
	Whole liver	82.8 (52.2–121.6)
	Left lobe	2.8 (1.5–8.1)
	Right lobe	113.4 (78.8–162.5)
	Non-tumor right lobe	65.8 (45.0–91.5)
	Tumor	447.3 (321.9–736.4)
Interval between TARE and follow-up CT, day		99 (87–112)

Data are presented as number of patients with percentage in parentheses, mean ± standard deviation, or median (interquartile range).

AST = aspartate aminotransferase, ALT = alanine transaminase, PT INR = prothrombin time international normalized ratio, AFP = alpha-fetoprotein, PIVKA-II = proteins induced by vitamin K absence or antagonist-II, TARE = transarterial radioembolization

For TARE, the median injected radioactivity was 2.9 GBq (IQR, 2.0–4.4 GBq), and no aberrant extrahepatic accumulation was observed in any patient. Of 40 patients, five underwent surgical resection at a median of 120 days (range: 72–147 days) after TARE. Among the patients who underwent resection, only one patient underwent TARE as a radiation right lobectomy, with an additional dose to the NRL to ensure sufficient FLR hypertrophy. In the remaining 39 patients, the primary intent of TARE was tumor control.

Patients underwent follow-up CT at a median of 99 days (IQR, 87–112 days) after TARE. CT volumetric analysis showed a median LL volume increase of 12.7% (IQR, 5.7%–28.1%). NRL showed significant post-treatment atrophy, with a median volume decrease of 8.8% (IQR, 1.9%–18.5%; P < 0.001).

Univariable Relationship Between Clinical Characteristics and CLH

First, the degree of CLH was compared according to the categorical clinical characteristics (Table 2). There was a trend toward greater hypertrophy in patients without splenomegaly than in those with splenomegaly, despite borderline significance (median, 18.0% vs. 7.8%, P = 0.046). Other factors, including viral status, ascites, portal vein thrombosis, and Child-Pugh scores, did not show significant associations.

Table 2. Relationship between clinical variables and contralateral lobe hypertrophy.

Clinical variables		Patient (n)	LL volume increase %	P
Hepatitis virus				0.845
	B	22	11.9 (5.5–23.1)
	C	5	24.5 (0.7–35.1)
	Others	13	12.8 (7.4–33.1)
Splenomegaly				0.046
	Present	15	7.8 (0.3–13.0)
	Absent	25	18.0 (8.7–41.4)
Ascites				0.367
	Present	5	4.3 (-3.6–42.0)
	Absent	35	12.9 (6.0–28.1)
Portal vein thrombosis				0.652
	Present	4	18.0 (5.3–57.9)
	Absent	36	12.4 (5.7–28.1)
Child-Pugh score				0.272
	5	34	14.7 (5.8–31.1)
	6	6	7.5 (0.8–12.2)

Data are presented as number of patients or median (interquartile range).

LL = left lobe

When clinical factors were tested as continuous variables for the prediction of CLH (Table 3), platelet count (Spearman’s ρ = 0.644 [95% confidence interval {CI}, 0.416–0.796], P < 0.001), AFP (ρ = -0.315 [-0.570–0.004], P = 0.048), and injected radioactivity (ρ = 0.561 [0.302–0.743], P < 0.001) demonstrated significant correlations with LL volume increase. Other factors, including serum AST, ALT, total bilirubin, albumin, PT INR, PIVKA-II, tumor size, and tumor volume, were not significant predictors.

Table 3. Univariable correlation between clinical variables or dosimetry parameters and contralateral lobe hypertrophy.

Variable		Spearman’s ρ [95% CI]	P
Clinical variables
	Age, yrs	0.565 [-0.394–0.224]	0.565
	Serum AST, U/L	-0.224 [-0.501–0.094]	0.164
	Serum ALT, U/L	-0.215 [0.493–0.104]	0.183
	Serum total bilirubin, mg/dL	-0.130 [-0.424–0.189]	0.424
	Serum albumin, g/dL	0.128 [-0.191–0.423]	0.430
	Platelet count, 103/µL	0.644 [0.416–0.796]	<0.001
	PT INR	-0.193 [-0.476–0.126]	0.233
	AFP, ng/mL	-0.315 [-0.570– -0.004]	0.048
	PIVKA-II, mAU/mL	0.003 [-0.308–0.315]	0.983
	Tumor size, cm	0.187 [-0.132–0.471]	0.248
	Tumor volume, cm3	0.201 [-0.118–0.482]	0.214
	Injected radioactivity, MBq	0.561 [0.302–0.743]	<0.001
	Interval between TARE and follow-up CT, days	-0.008 [-0.319–0.304]	0.961
Dosimetry parameters
	Mean dose of the whole liver, Gy	0.594 [0.347–0.764]	<0.001
	Mean dose of RL, Gy	0.586 [0.336–0.759]	<0.001
	Mean dose of NRL, Gy	0.662 [0.442–0.807]	<0.001
	P50 of RL, %	0.546 [0.283–0.733]	<0.001
	P50 of NRL, %	0.672 [0.455–0.813]	<0.001
	P100 of RL, %	0.560 [0.301–0.742]	<0.001
	P100 of NRL, %	0.563 [0.305–0.744]	<0.001

CI = confidence interval, AST = aspartate aminotransferase, ALT = alanine transaminase, PT INR = prothrombin time international normalized ratio, AFP = alpha-fetoprotein, PIVKA-II = proteins induced by vitamin K absence or antagonist-II, TARE = transarterial radioembolization, RL = right lobe, NRL = non-tumor right lobe, P₅₀ = volume fraction of absorbed dose exceeding 50 Gy, P₁₀₀ = volume fraction of absorbed dose exceeding 100 Gy

Univariable Correlation Between PET Dosimetry and CLH

Patients received a median mean absorbed dose of 82.8 Gy (IQR 52.2–121.6) for the whole liver. The mean absorbed doses to RL, LL, NRL, and tumor were 113.4 Gy (IQR, 78.8–162.5), 2.8 Gy (1.5–8.1), 65.8 Gy (IQR, 45.0–91.5), and 447.3 Gy (321.9–736.4), respectively. The dose for LL was far lower than that for RL, confirming the successful targeting of RL.

Dosimetry parameters significantly correlated with CLH (Table 3, Fig. 2). The mean absorbed dose to NRL showed the strongest correlation with LL among the mean absorbed dose parameters (ρ = 0.662 [0.442–0.807], P < 0.001). Mean absorbed doses to RL (ρ = 0.586 [0.336–0.759], P < 0.001) and the whole liver (ρ = 0.594 [0.347–0.764], P < 0.001) demonstrated slightly lower correlations.

Fig. 2. Univariable correlations between PET dosimetry parameters and contralateral lobe hypertrophy. A-G: Correlations between representative dosimetric parameters (the mean dose for all VOIs and volume fraction parameters for NRL and RL) and contralateral lobe hypertrophy (LL volume increase) are demonstrated using scatter plots and Spearman’s rank correlation coefficients (ρ). VOI = volume of interest, NRL = non-tumor right lobe, RL = right lobe, LL = left lobe, P₅₀ = volume fraction of absorbed dose exceeding 50 Gy, P₁₀₀ = volume fraction of absorbed dose exceeding 100 Gy.

Voxel proportions exceeding the predefined dose thresholds were also significantly correlated with LL hypertrophy. Among these, P₅₀ of NRL demonstrated the strongest correlation with LL hypertrophy (ρ = 0.672 [0.455–0.813], P < 0.001). P₁₀₀ of NRL showed a slightly lower but still significant correlation (ρ = 0.563 [0.305–0.744], P < 0.001). Similar correlations were observed for RL with P₅₀ and P₁₀₀ showing ρ values of 0.546 (95% CI 0.283–0.733; P < 0.001) and 0.560 (95% CI 0.301–0.742; P < 0.001), respectively.

Multivariable Prediction of CLH

Multivariable analysis was performed to assess the independent predictive power of clinical and dosimetry parameters for CLH (Table 4). Splenomegaly, platelet count, and injected radioactivity were selected as candidate clinical predictors based on previous group comparisons and correlation analyses. Among the dosimetry parameters, the mean absorbed dose of the NRL was the most robust predictor of CLH (R² = 0.556; P < 0.001). Other dosimetry factors, including the mean dose to the whole liver and RL, and voxel threshold metrics (P₅₀ and P₁₀₀), did not retain independent predictive values. In addition, the platelet count was the only clinical variable that showed an independent predictive value (R² = 0.221, P = 0.003).

Table 4. Multivariable regression analyses for prediction of contralateral lobe hypertrophy.

Variable	R2	P
Splenomegaly	-	N/S
Platelet count, 103/µL	0.221	0.003
AFP, ng/mL	-	N/S
Injected radioactivity, MBq	-	N/S
Mean dose of the whole liver, Gy	-	N/S
Mean dose of RL, Gy	-	N/S
Mean dose of NRL, Gy	0.556	<0.001
P50 of RL, %	-	N/S
P100 of RL, %	-	N/S
P50 of NRL, %	-	N/S
P100 of NRL, %	-	N/S

N/S = non-significant, AFP = alpha-fetoprotein, RL = right lobe, NRL = non-tumor right lobe, P₅₀ = volume fraction of absorbed dose exceeding 50 Gy, P₁₀₀ = volume fraction of absorbed dose exceeding 100 Gy

Patients with higher platelet count (≥130 × 10³/µL) exhibited significantly greater CLH than those with a lower platelet count (median 23.8%, IQR 12.8%–41.2% vs. 0%, IQR -2.2%–7.1%; P < 0.001). When stratified analysis based on platelet count was performed (Fig. 3), the correlation between the NRL mean absorbed dose and CLH was stronger and more significant in the subgroup with higher platelet count (ρ = 0.847 [0.684–0.929]; P < 0.001). In contrast, the subgroup with lower platelet count exhibited a weak effect of NRL dose on LL hypertrophy (ρ = 0.429 [-0.132–0.781]; P = 0.129).

Fig. 3. Correlation between the mean absorbed dose to NRL and contralateral lobe hypertrophy stratified by platelet count. A: In patients with platelet count ≥130 × 10³/µL, a strong and statistically significant positive correlation was observed, indicating a marked hypertrophic response with increasing NRL dose. B: In contrast, patients with platelet count <130 × 10³/µL showed a weaker and non-significant correlation, with overall lower degrees of hypertrophy. Each point represents an individual patient; solid lines represent linear regression fits. NRL = non-tumor right lobe, LL = left lobe.

DISCUSSION

In this study, we conducted voxel-based ⁹⁰Y PET/CT dosimetry to predict CLH after TARE in a selected cohort of treatment-naïve patients with solitary right lobar HCC treated with glass microspheres. Among the PET dosimetry factors, the mean absorbed dose (ρ = 0.662 [0.442–0.807], P < 0.001) and P₅₀ (ρ = 0.672 [0.455–0.813], P < 0.001) of NRL demonstrated the highest correlations with LL hypertrophy. Dosimetric parameters for the RL and the whole liver also showed correlations but had lower predictive values. Multivariable analysis revealed that the mean NRL dose was the sole independent dosimetric predictor (R² = 0.556, P < 0.001). Platelet count, which indicates the liver function reserve, was also identified as a significant predictor among the clinical variables (R² = 0.221, P = 0.003), forming two complementary contributors alongside radiation exposure. Stratified analysis further revealed that the degree of hypertrophy increased more markedly with NRL dose increment in patients with preserved platelet counts (≥130 × 10³/µL; ρ = 0.847 [0.684–0.929]; P < 0.001) than in those with decreased platelet counts (ρ = 0.429 [-0.132–0.781]; P = 0.129).

Dose planning to induce CLH in TARE is guided by ^99mTc-MAA imaging, but the prediction of CLH using the actual delivered dose measured on ⁹⁰Y PET remains an incompletely explored area and has only been evaluated by a few studies [25,26]. While ^99mTc-MAA mapping remains the most established tool for pretreatment planning, potential discrepancies between the predicted and actual dose distributions [24] make the PET-based assessment of the actual dose distribution a complementary methodology for CLH prediction. Given that the FLR is a critical determinant of post-hepatectomy liver failure, early prediction of CLH based on ⁹⁰Y PET dosimetry may help identify surgical candidates and assist in planning resection shortly after TARE. It also enables the early recognition of patients who are unlikely to achieve sufficient hypertrophy, facilitating the prompt consideration of alternative treatment strategies. Notably, in settings where CLH was not the primary goal at planning, as was the case for most patients in our cohort, hypertrophy tended to be inadequate owing to the absence of intentional radiation to the normal liver. In such cases, PET-based prediction of CLH can guide timely decisions regarding the feasibility of subsequent surgery.

Although not yet fully elucidated, radioembolization is thought to induce CLH through the atrophy–hypertrophy complex, which is a regenerative response following hepatocyte loss from various types of injuries [15]. Radiation to the treated lobe causes progressive atrophy via tissue fibrosis, diverting portal venous flow toward the contralateral lobe and stimulating regenerative signaling that ultimately results in compensatory hypertrophy [14,28]. This mechanism is supported by prior dosimetry studies demonstrating correlations between non-tumor liver doses and CLH, despite minor differences in ROI definitions and used dosimetric parameters. Palard et al. [16] and Grisanti et al. [25] analyzed the nontumor liver within the treated lobe, whereas Demir et al. [26] evaluated the entire nontumor liver. In line with the hypothesis that CLH reflects radiation-induced fibrosis of the treated lobe, we selected the NRL as our target ROI rather than the entire non-tumor liver. Regarding dosimetric parameters, the mean absorbed dose and volume fractions exceeding certain thresholds (e.g., >30 Gy [25] and >20–100 Gy [26]) have shown predictive value but lack a clear consensus on preference. In our study, the mean absorbed dose and voxel-based metrics in the NRL were correlated with hypertrophy. Among these factors, we propose the mean absorbed dose as the representative metric because it remains significant in multivariable analysis, is more clinically accessible, and is less vulnerable to variations in image quality and processing [29].

CLH can depend on pre-existing conditions, such as previous therapy and liver function, in addition to treatment [11,30,31,32,33,34]. Previous studies included heterogeneous patients, in terms of pathology and prior systemic therapies [25,26]. In contrast, we included a homogeneous group of treatment-naïve patients with HCC to minimize the potential effects of such confounding factors. Moreover, our cohort consisted entirely of patients with HCC, among whom chronic liver disease was the most frequent. Given that portal flow redistribution is a key mechanism in the atrophy–hypertrophy complex [14,28], baseline portal hypertension may influence the FLR response. In this context, our cohort enabled a more effective assessment of the impact of portal hypertension, as reflected by surrogate markers such as platelet count, thereby enabling more precise CLH prediction. Additionally, all 40 patients were treated with glass microspheres, in contrast to a previous PET/CT-based study that analyzed resin microspheres [25] and a PET/MRI-based study that included both types [26]. These two microsphere types differ in key properties that may influence their downstream biological effects including CLH [18]. In particular, the effect of radiation on CLH can be evaluated more specifically in the glass microsphere-treated group, because glass microspheres have a low embolic potential. Considering this, the exclusive use of glass microspheres further clarified the interpretation of the findings.

Our study identified platelet count as a predictor of CLH and a representative marker of liver function reserve. Platelet count has been associated with CLH following PVE [33] and TARE [11,30], and thrombocytopenia is an early manifestation of portal hypertension [35]. Notably, CLH in response to increasing NRL doses was more prominent in patients with adequate platelet counts. This suggests the importance of liver function reserve when interpreting the dose-hypertrophy relationship after TARE, even in patients without advanced cirrhosis. Other liver function-related parameters were not significant in our analysis, likely because of the relatively preserved liver function in our cohort. In our study, all patients had Child-Pugh scores of 5 or 6.

A limitation of this study is the relatively short time interval between TARE and CLH evaluation, with CT performed at a median of 98 days post-TARE, owing to the characteristics of our patient cohort. We applied strict inclusion criteria and enrolled only treatment-naïve patients with a solitary right lobar HCC. Many patients with these conditions undergo additional local treatment within a few months of TARE. Given the strict inclusion criteria and patient management patterns, extending the follow-up interval would have limited the number of patients with available follow-up CT scans without additional treatment. Therefore, the median follow-up period was relatively short. This was shorter than the time intervals in previous studies using ^99mTc-MAA SPECT/CT [16] and ⁹⁰Y PET/CT [25], wherein the follow-up periods were >6 months, and was comparable to that in a PET/MR-based study [26]. Considering that maximal hypertrophy is typically achieved at 6–12 months post-treatment with volume increases of 35%–40% [13], our follow-up interval was suboptimal for predicting the degree of maximal hypertrophy and was earlier than the typical timing of surgery following radiation lobectomy. However, other studies have reported that the degree of hypertrophy reaches 57%–81% of the maximum by three months [30,36,37,38], and a previous time-dependent analysis showed near-maximal hypertrophy 2–4 months after TARE [25]. Thus, despite the shorter interval, the follow-up period may still be sufficient to evaluate the predictive value of the PET dosimetry parameters, although this does not allow for the determination of precise dose thresholds.

The second limitation of this study is that most patients underwent TARE primarily for tumor control rather than for the explicit goal of inducing CLH. Consequently, the cohort may not be representative of all patients who underwent radiation lobectomy for CLH. In addition, this can be a limitation in evaluating the dose thresholds for CLH, particularly in those with low platelet counts, who may require a higher dose based on our study results. Nevertheless, the cohort reflects the real-world clinical practice in regions where TARE is performed as a primary treatment rather than as a bridging therapy.

A third limitation was the small sample size. Although this small sample size reduced the statistical power of the findings, strict inclusion criteria allowed for the selection of a homogeneous patient cohort. This clinical uniformity enhances the internal validity of the results and enables a clearer interpretation of the relationship between ⁹⁰Y PET/CT-based dosimetry and CLH. However, despite this strict cohort design, the retrospective nature of the study limits control over the standardization of treatment protocols and introduces the possibility of unmeasured confounding variables. In addition, because the study included patients with large tumors who received boosted-dose therapy, it may not represent the broader HCC population. Future studies with larger prospective cohorts are needed to optimize the predictive models that incorporate PET-based dosimetry and thrombocytopenia.

In conclusion, this study demonstrated that post-TARE ⁹⁰Y PET/CT-based dosimetry is a predictive tool for CLH after TARE using glass microspheres in a homogeneous cohort of treatment-naïve patients with HCC. The NRL mean absorbed dose served as the primary dosimetric predictor; the platelet count, reflecting the liver functional reserve, influenced the dose–hypertrophy relationship as an additional independent predictive factor. However, prospective studies involving larger cohorts are required to validate this combined predictive strategy.

Availability of Data and Material

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.