Feasibility study of ²²⁵Ac-PSMA-CY313 dosimetry in mCRPC patients using SPECT

Abstract

Purpose

Actinium-225 (²²⁵Ac)-labeled prostate-specific membrane antigen (PSMA) radiopharmaceuticals represent a promising therapeutic approach for metastatic castration-resistant prostate cancer (mCRPC), yet clinical implementation remains limited by the absence of accurate dosimetric assessment methods. The complex decay chain and non-imaging alpha emissions of ²²⁵Ac pose substantial challenges for quantitative imaging. We aimed to evaluate the feasibility of quantitative single photon emission computed tomography (SPECT)-based dosimetry for ²²⁵Ac-PSMA-CY313 therapy by exploiting gamma emissions from daughter radionuclides francium-221 (²²¹Fr) and bismuth-213 (²¹³Bi).

Methods

Four mCRPC patients received 185.8 ± 11.7 µCi ²²⁵Ac-PSMA-CY313 and underwent multi-timepoint SPECT/CT and whole-body planar imaging at 6, 24, 48, and 96 h post-injection. Quantitative SPECT reconstruction used ordered-subsets expectation-maximization with comprehensive corrections for attenuation, scatter, resolution blur, and crosstalk. Volume of interest were defined using co-registered ¹⁸F-PSMA-CY313 positron emission tomography /computed tomography (PET/CT). Time-activity curves were fitted with mono- or bi-exponential models, and absorbed doses were calculated using validated Monte Carlo-based software and International Commission on Radiological Protection reference phantoms.

Results

High-quality quantitative imaging was successfully achieved across all timepoints. Among normal organs, kidneys and liver exhibited the highest absorbed doses (1.55 ± 0.38 Gy and 1.07 ± 0.19 Gy, respectively), corresponding to dose coefficients of 0.23 ± 0.07 Gy/MBq and 0.16 ± 0.03 Gy/MBq. Soft-tissue lesions exhibited higher absorbed doses than bone metastases (5.03 ± 5.51 Gy versus 1.61 ± 2.28 Gy), with corresponding dose coefficients of 0.73 ± 0.80 Gy/MBq and 0.25 ± 0.33 Gy/MBq. Tumor-to-critical organ dose ratios indicated favorable therapeutic windows, with red marrow showing the highest ratio (14.84), followed by adrenal glands (6.35) and salivary glands (4.96), while the dose-limiting kidneys demonstrated a ratio of 1.55.

Conclusion

Quantitative SPECT-based dosimetry for ²²⁵Ac-PSMA-CY313 therapy is clinically feasible using standard imaging systems. This methodology demonstrates preferential tumor targeting with acceptable organ-at-risk dose distributions, supporting the therapeutic potential of ²²⁵Ac-PSMA-CY313 for mCRPC and providing a practical framework for personalized dosimetry in targeted alpha therapy.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40658-025-00831-7.

Introduction

Prostate-specific membrane antigen (PSMA)–targeted alpha therapy (TAT) has emerged as a transformative modality for metastatic castration-resistant prostate cancer (mCRPC), particularly with actinium-225 (²²⁵Ac)–labeled agents [1]. The physical properties of ²²⁵Ac—high linear energy transfer (LET) of approximately 100 keV/µm and short alpha-particle range of 50–80 μm—are well suited to eradicate micrometastases while sparing surrounding normal tissue [1]. Consequently, ²²⁵Ac-labeled PSMA radiopharmaceuticals are promising for patients with advanced disease who have exhausted conventional options.

Clinical implementation of ²²⁵Ac-based TAT, however, fundamentally depends on accurate dosimetry for treatment planning, efficacy prediction, and toxicity mitigation. Patient-specific dosimetry enables delivery of adequate tumor doses while constraining organs at risk to within tolerance limits. Despite this need, the inherent characteristics of ²²⁵Ac complicate dosimetric assessment: alpha emissions are not externally detectable due to their limited tissue penetration, and the complex decay chain with multiple daughter radionuclides hinders straightforward quantification. As a result, clinical data on patient-specific dosimetry for ²²⁵Ac-based TAT remain scarce, posing a significant barrier to individualized optimization of alpha-therapy protocols [2].

Single-photon emission computed tomography (SPECT) offers a promising avenue to address these dosimetric challenges by indirectly quantifying ²²⁵Ac activity in vivo through detection of gamma-emitting daughter radionuclides, particularly francium-221 (²²¹Fr) and bismuth-213 (²¹³Bi) [3]. This approach leverages the ²²⁵Ac decay chain to provide spatial and temporal information on radiopharmaceutical distribution and kinetics. However, conventional SPECT methodologies face substantial limitations in this context. The low photon yield from daughter radionuclides produces poor signal-to-noise ratios, and significant inter-peak crosstalk between energy windows compromises quantitative accuracy. Additionally, standard reconstruction algorithms often lack comprehensive physical corrections for attenuation, scatter, and resolution blur, further degrading image quality and quantification precision [2]. These factors have historically hindered the clinical translation of SPECT-based dosimetry for alpha therapy.

Recent technological advances have begun to mitigate these challenges through the development of optimized acquisition protocols and reconstruction methodologies [4]. Multi-timepoint acquisitions enable more robust kinetic modeling, while advanced iterative reconstruction with full physical corrections has improved quantification accuracy [5]. Moreover, organ standardization using computational phantoms such as the International Commission on Radiological Protection (ICRP) reference models has enhanced the reliability of absorbed dose calculations [6]. Despite these advances, comprehensive validation studies demonstrating the clinical feasibility and accuracy of advanced SPECT-based dosimetry for ²²⁵Ac therapy remain scarce [2].

Against this technological and clinical backdrop, this study evaluates the feasibility of quantitative SPECT-based dosimetry in mCRPC patients treated with ²²⁵Ac-PSMA-CY313, a novel alpha-targeting radiopharmaceutical, by systematically exploiting detectable gamma emissions from the ²²¹Fr and ²¹³Bi daughter radionuclides. Our methodological approach incorporates key innovations to overcome the technical limitations inherent to ²²⁵Ac imaging. The imaging protocol combines dynamic single-bed SPECT/CT acquisitions with whole-body planar scans at multiple time points (6, 24, 48, and 96 h post-injection) to capture comprehensive pharmacokinetic information. Image reconstruction employs an advanced ordered-subsets expectation–maximization algorithm with comprehensive physical corrections, including attenuation compensation, resolution recovery, scatter correction, and crosstalk correction between energy peaks. Volumes of interest are defined using co-registered ¹⁸F-PSMA-CY313 PET/CT to ensure accurate anatomical localization, with subsequent normalization to the ICRP adult male computational phantom for standardized time–activity curve generation. Absorbed dose calculations are performed using validated Monte Carlo–based software (3D-RD) to ensure dosimetric accuracy. This integrated workflow represents one of the first comprehensive clinical demonstrations of advanced SPECT-based dosimetry for ²²⁵Ac-labeled radiopharmaceuticals, providing preliminary evidence to support the implementation of personalized dosimetry protocols in alpha-particle therapy. Successful validation of this approach could advance the field by enabling improved dose quantification, individualized treatment planning, and enhanced toxicity risk management, thereby facilitating broader clinical adoption of ²²⁵Ac-based TAT for mCRPC.

Methods

The specific process of the study is shown in Fig. 1.

Fig. 1.

Workflow of SPECT-based dosimetry for ²²⁵Ac-PSMA-CY313, from phantom calibration to image-guided VOI segmentation, TAC fitting, and absorbed dose calculation using Rapid 3D-RD-S

Imaging acquisition protocols

This dosimetric sub-study was embedded in a prospective, open-label, single-arm clinical trial involving four patients with metastatic castration-resistant prostate cancer (mCRPC). All patients received ²²⁵Ac-PSMA-CY313 at 200 µCi ± 10% every 8 weeks for a total of four cycles. Per protocol, imaging was performed at approximately 6, 24, 48, and 96 h after administration of ²²⁵Ac-PSMA-CY313. At each time point, SPECT/CT tomographic imaging covering at least one bed position was acquired to include the majority of tumor lesions, along with near whole-body planar imaging from head to toe. Table 1 summarizes body weight, administered activity, and SPECT/CT scan times for the four subjects.

Table 1.

Administrated doses and imaging time

Subject	Weight (Kg)	Injected [ 225Ac] Ac-PSMA-CY313 activity (µCi)	Imaging time after injection (hours)
01	78.2	198	6.3	27.2	45.2	93.1
02	77.0	170	8.2	28.8	46.7	94.6
03	86.5	185.5	8.5	26.0	45.4	98.8
04	63.8	185.8	10.8	28.7	47.7	101.0

SPECT/CT scans were performed on a GE Discovery NM/CT 670 Pro system using a non-circular orbit. A total of 120 uniformly spaced projections over 360° were acquired with 60 s per projection. The system was equipped with high-energy collimators. The projection matrix was 128 × 128 with a pixel size of 0.442 cm. Energy windows were configured as follows: for ²²¹Fr, a 218 keV window with 20% width (± 10% tolerance); for ²¹³Bi, a 440 keV window with the same parameters. Owing to the short half-life of ²²¹Fr (4.9 min), its biodistribution was assumed to mirror that of the parent radionuclide ²²⁵Ac. The 4.9-minute half-life of ²²¹Fr precludes significant biological redistribution, with > 99% of decays occurring within 30 min at the parent ²²⁵Ac location [7, 8]. This rapid decay ensures that ²²¹Fr imaging accurately reflects ²²⁵Ac distribution. In contrast, ²¹³Bi, with a half-life of 45.6 min, may redistribute following decay from the parent.

After each SPECT acquisition, a CT-based attenuation correction (CTAC) scan was performed to generate attenuation maps. CT images were also used to define volumes of interest (VOIs) for specific organs, as detailed in VOI delimitation and quantification.

For whole-body planar imaging, the image matrix was 256 × 1024 with a pixel size of 0.221 cm. Energy window settings matched those used for SPECT imaging.

Phantom image and calibration factors

A NEMA phantom uniformly filled with ²²⁵Ac was imaged using the same SPECT/CT and whole-body planar protocols as the patient scans. This phantom study was conducted to optimize imaging parameters and to derive calibration factors for converting reconstructed image counts to activity concentration (µCi/cc). Phantom data were reconstructed and processed identically to patient data, as described in Sect. "SPECT image reconstruction method".

The total radioactivity in the phantom was 96.75 µCi with a total volume of 9.7 L. Table 2 summarizes the calibration factors for SPECT and planar imaging across the two energy windows. Specifically, a static (non–decay-corrected) calibration factor for the 218 keV window was applied to convert 218 keV image counts to ²²⁵Ac activity. Likewise, the calibration factor corresponding to the 440 keV window was used to derive ²¹³Bi activity from the 440 keV image counts.

Table 2.

Calibration factors

Calibration factors	218 keV energy window	440 keV energy window
SPECT	6.38E-05 µCi/count/cc	4.38E-05 µCi/count/cc
Planar	0.0016 µCi/count	0.0019 µCi/count

SPECT image reconstruction method

SPECT images were reconstructed using the quantitative single-photon emission computed tomography (QSPECT) method. QSPECT employs an ordered-subsets expectation–maximization (OSEM) algorithm with compensation for resolution blurring, photon attenuation, scatter, and crosstalk/downscatter among energy peaks. Attenuation correction maps were derived from CT images. Resolution blurring was modeled using a collimator–detector response function (CDRF) generated by Monte Carlo simulation, accounting for geometric collimation, septal penetration and scatter, scatter within the detector crystal, and backscatter. Scatter was addressed using the effective source scatter estimation (ESSE) method. QSPECT supports simultaneous reconstruction from multiple radionuclides. Reconstruction used five iterations with twelve subsets per iteration, and no post-reconstruction filtering was applied.

SPECT quantification accuracy was validated using simulation studies with a 22 cm diameter cylindrical phantom and a 6 cm diameter spherical insert. Activity levels were matched to typical patient acquisitions (approximately 264,000 total counts). Twenty noise realizations were generated using Poisson statistics and reconstructed with identical clinical parameters to assess quantitative accuracy and precision. Recovery coefficients were calculated to evaluate partial volume effects for different lesion sizes.

VOI delimitation and quantification

Organs visible on CT were segmented using an artificial intelligence–based method built on the MONAI framework and implemented via the 3D Slicer software (https://www.slicer.org/). AI-generated organ segmentations underwent rigorous quality control with manual review by two experienced nuclear medicine physicians. Segmentation accuracy was verified through comparison with anatomical landmarks and ICRP reference organ volumes, with all measured volumes showing < 15% variation from reference values. This approach aligns with validated MONAI framework implementations reporting DSC values > 93% for abdominal organ segmentation [9]. Depending on CT coverage, segmented organs included the heart, spleen, kidneys, gallbladder, liver, stomach, pancreas, prostate, adrenal glands, lungs, small intestine, and colon. The red marrow VOI was manually delineated on CT.

Tumor localization was first confirmed on PET/CT. For tumors visible on SPECT/CT, VOIs were manually delineated on fused SPECT/CT images using a threshold-based method (threshold set at 20%). Tumor VOIs were drawn at each imaging time point and applied to the corresponding SPECT images to quantify counts within each organ and tumor VOI in the 218 keV and 440 keV energy windows. These counts were converted to activity concentrations (µCi/cc) of ²²⁵Ac and ²¹³Bi using the calibration factors in Table 2.

For organs not visualized on SPECT/CT—primarily the thyroid and salivary glands, and occasionally the heart and lungs—VOIs were manually delineated on planar images using anterior and posterior views for localization. Total organ uptake was estimated using the geometric mean method and converted to ²²⁵Ac and ²¹³Bi activity (µCi) using the calibration factors in Table 2. Organ uptake values were then adjusted for body weight based on the adult male phantom described in ICRP Publication 103. Time–activity curves (TACs) were generated separately for ²²⁵Ac and ²¹³Bi.

TACs fitting and integration

Given the potential redistribution of ²¹³Bi, TACs for ²²⁵Ac and ²¹³Bi were fitted and integrated separately using the Rapid 3D-RD-S software (Radiopharmaceutical Imaging and Dosimetry, LLC, Baltimore, USA). Bi-exponential or mono-exponential models were applied to characterize uptake and clearance kinetics. The fitted models were analytically integrated to estimate the area under the curve (AUC) and derive the time-integrated activity coefficient (TIAC). If the model fit was suboptimal, a mono-exponential function was fitted to the last two data points, while the area from injection to the penultimate time point was estimated using the trapezoidal method; integration was then completed using the fitted mono-exponential function.

Dosimetry calculation

Time-integrated activity coefficients (TIACs) for organs were used to compute absorbed doses according to the Medical Internal Radiation Dose (MIRD) schema, with all calculations performed using Rapid 3D-RD-S. The ICRP 103 adult male reference phantom was used. Tumors were modeled as spheres defined by their volume and density. Absorbed doses and dose coefficients were calculated for ²²⁵Ac, ²¹³Bi, and their progeny (²²¹Fr, ²¹⁷At, ²¹³Po, ²⁰⁹Pb, and ²⁰⁹Tl). Final absorbed doses were obtained by weighted summation of contributions from all radionuclides according to their branching ratios. A relative biological effectiveness (RBE) of 5 was applied for alpha particles, and an RBE of 1 for electrons and photons.

Results

Imaging feasibility and VOI definition

Quantitative SPECT/CT and whole-body planar imaging were successfully acquired at 6, 24, 48, and 96 h post-injection in all four patients. High-quality reconstructions were achieved using ordered-subsets expectation–maximization with comprehensive corrections for attenuation, scatter, resolution blurring, and crosstalk. These corrections yielded clearly defined uptake regions in tumors and major organs, providing a reliable basis for dosimetric quantification. Phantom validation studies demonstrated excellent quantitative accuracy with an estimated bias of -4.9% and coefficient of variation of 0.64%, confirming high precision across noise realizations.

For anatomical localization and volume-of-interest (VOI) definition, a multimodal image-guided workflow was implemented. Tumor VOIs were delineated on ¹⁸F-PSMA-CY313 PET/CT using a 20% threshold to account for resolution differences between PET and SPECT (Fig. 2). Only lesions within the SPECT field of view were included. For organ segmentation, an AI-assisted algorithm based on the MONAI framework was applied to CT images from the SPECT/CT acquisitions, enabling accurate and reproducible VOI definition for major organs (Fig. 2, left).

Fig. 2.

Tumor and organ VOI definition. Right: PET/CT image of patient 01 showing three representative tumor lesions (Tumor 1–3) segmented using a 20% threshold. Left: Organ segmentation was performed on CT images using MONAI-based AI algorithms, with manual refinement where necessary

Following VOI segmentation, quantitative SPECT/CT images were reconstructed across all time points. Representative serial images for ²²⁵Ac (via ²²¹Fr, 218 keV) and ²¹³Bi (440 keV) are shown in Fig. 3. These images demonstrate dynamic radiotracer distribution and retention at 6.3, 27.2, 45.2, and 93 h, highlighting progressive tumor uptake and clearance patterns. The lesion labeled Tumor 1 in patient 01 showed clear focal uptake, traceable across both radionuclides and multiple time points.

Fig. 3.

Reconstructed SPECT/CT images of patient 01 at multiple time points post-injection, showing dynamic uptake of ²²⁵Ac (left column at each time point) and ²¹³Bi (right column). Tumor 1 is clearly visualized and tracked over time. Image reconstruction includes attenuation, scatter, and resolution corrections to improve quantification accuracy

This integrated imaging and VOI definition strategy ensured accurate temporal sampling and anatomical localization for subsequent TAC fitting and absorbed dose calculations.

Time-activity curves integration and residence times

To quantitatively assess tissue exposure to ²²⁵Ac, the area under the time–activity curve (AUC) was calculated for each tissue using fitted mono- or bi-exponential models. Figure 4 illustrates the integration process for kidney and tumor TACs (see Supplementary Fig. 1 for detailed data), where the shaded area represents the cumulative activity over time. Both the fitted curve and measured data points are shown, and the trapezoidal rule was applied between sampling points prior to model fitting. As shown in Table 3, the liver exhibited the longest mean residence time for ²²⁵Ac (251.02 ± 141.38 min), followed by the lungs (105.95 ± 56.15 min), kidneys (63.56 ± 43.16 min), heart wall (55.26 ± 19.15 min), and red marrow (45.23 ± 27.54 min). The average tumor retention time for ²²⁵Ac was 11.59 ± 19.21 min, with high inter-patient variability; some soft-tissue lesions exceeded 30 min.

Fig. 4.

Plot of Integral for ²²⁵Ac in kidney and tumor 1 of patient 01

Table 3.

Residence time (min) of ²²⁵Ac and ²¹³ Bi in target organs/tissues

Residence time (min)	225Ac	213 Bi
Adrenals	1.89 ± 1.73	1.62 ± 1.52
Heart wall	55.26 ± 19.15	21.80 ± 16.41
Kidneys	63.56 ± 43.16	49.39 ± 34.31
Liver	251.02 ± 141.38	127.35 ± 61.42
Lungs	105.95 ± 56.15	25.66 ± 16.81
Pancreas	17.17 ± 6.95	6.23 ± 2.68
Red marrow	45.23 ± 27.54	37.75 ± 33.28
Salivary glands	4.38 ± 4.31	5.86 ± 5.24
Spleen	17.36 ± 9.75	4.86 ± 1.06
Thyroid	1.60 ± 1.26	0.97 ± 0.58
Tumor	11.59 ± 19.21	3.94 ± 6.06

Absorbed dose and absorbed dose coefficients

Table 4 presents the RBE-weighted absorbed doses and dose coefficients (Gy/MBq) of ²²⁵Ac-PSMA-CY313 for normal organs and tumors. Among normal organs, kidneys and liver received the highest RBE-weighted absorbed doses, with mean values of 1.55 ± 0.38 Gy and 1.07 ± 0.19 Gy, corresponding to dose coefficients of 0.23 ± 0.07 Gy/MBq and 0.16 ± 0.03 Gy/MBq, respectively. In contrast, minimal absorbed doses were observed in the brain, testes, and eye lens, indicating limited off-target exposure in these critical organs.

Table 4.

RBE-weighted absorbed doses and corresponding dose coefficients of ²²⁵Ac-PSMA-CY313 in normal organs and tumors

	RBE-weight absorbed dose (Gy)	RBE-weight absorbed dose coefficient (Gy/MBq)
Adrenals	0.39 ± 0.28	0.06 ± 0.04
Brain	(6.16 ± 2.90)E-06	(9.17 ± 4.68)E-07
Breast	(4.89 ± 0.35)E-05	(7.18 ± 8.11)E-07
eye-lens	(4.25 ± 1.47)E-06	(6.28 ± 2.38)E-07
Gb-wall	(7.94 ± 3.18)E-03	(1.17 ± 0.47)E-03
Ht-wall	0.75 ± 0.36	0.11 ± 0.06
Kidneys	1.55 ± 0.38	0.23 ± 0.07
Liver	1.07 ± 0.19	0.16 ± 0.03
Muscle	(2.30 ± 0.45)E-05	(3.40 ± 0.86)E-06
Pancreas	1.02 ± 0.57	0.15 ± 0.08
Prostate	1.99 ± 2.73	0.30 ± 0.40
Red marrow	0.16 ± 0.09	0.02 ± 0.01
Salivary glands	0.50 ± 0.47	0.07 ± 0.07
Skin	(1.31 ± 0.16)E-05	(1.93 ± 0.34)E-06
Spleen	0.54 ± 0.28	0.08 ± 0.05
Stomach	(1.21 ± 1.06)E-03	(1.77 ± 1.54)E-04
Testes	(2.46 ± 1.39)E-06	(3.68 ± 2.24)E-07
Thymus	(7.26 ± 1.62)E-05	(1.07 ± 0.30)E-05
Thyroid	0.56 ± 0.39	0.08 ± 0.05
Urinary bladder	(2.58 ± 1.51)E-05	(3.85 ± 2.45)E-06
Tumor-bone	1.61 ± 2.28	0.25 ± 0.33
Tumor-soft tissue	5.03 ± 5.51	0.73 ± 0.80
Tumor	2.41 ± 3.44	0.36 ± 0.50

*RBE, relative biological effectiveness

For tumor lesions, RBE-weighted absorbed doses were analyzed separately for bone and soft-tissue sites. Soft-tissue lesions—including primary tumors, lymph node metastases, and distant soft-tissue metastases—showed a higher mean absorbed dose (5.03 ± 5.51 Gy) and dose coefficient (0.73 ± 0.80 Gy/MBq) compared with bone lesions (1.61 ± 2.28 Gy and 0.25 ± 0.33 Gy/MBq, respectively).

Tumor-to-organ ADC ratios

Figure 5 shows tumor-to-organ ADC ratios. The red marrow exhibited the highest ratio (14.84), followed by the adrenal and salivary glands (6.35 and 4.96, respectively). Ratios for the heart wall, liver, pancreas, spleen, and thyroid ranged between 2 and 4. The prostate and kidneys showed the lowest ratios, approximately 1.55.

Fig. 5.

Tumor-to-organ absorbed dose coefficient ratios

Absorbed dose and standard uptake values mean (SUVmean) of tumor lesions

Table 5 summarizes the absorbed dose and SUVmean for all tumor lesions across the four subjects. Tumor absorbed doses varied widely, from 0.06 to 13.09 Gy. SUVmean also showed considerable variability, ranging from 6.4 to 21.4. However, no one-to-one correspondence between SUVmean and absorbed dose was observed.

Table 5.

Standard uptake values mean (SUVmean) and absorbed dose of tumor lesions

	Absorbed dose(Gy) 225Ac-PSMA-CY313 SPECT/CT		SUVmean 18F-PSMA-CY313 PET/CT
01	lesion1	0.48	18.5
	lesion2	0.44	21.4
	lesion3	0.92	21.0
02	lesion1	0.32	9.9
	lesion2	2.33	17.8
	lesion3	3.5	11.6
	lesion4	0.67	12.7
	lesion5	2.23	11.5
03	lesion1	0.08	7.0
	lesion2	0.13	6.4
	lesion3	0.06	7.1
	lesion4	3.88	10.6
	lesion5	3.76	11.0
04	lesion1	7.88	9.9
	lesion2	0.72	8.2
	lesion3	0.54	9.6
	lesion4	13.09	10.3

Discussion

This study demonstrates the feasibility of quantitative dosimetry for ²²⁵Ac-PSMA-CY313 using a combined approach of multi-timepoint SPECT/CT and whole-body planar imaging. As a feasibility study, our cohort of four patients was designed to establish technical proof-of-concept for SPECT-based ²²⁵Ac dosimetry rather than to derive statistically powered dose-response relationships. This sample size aligns with established precedents in early-phase alpha-therapy dosimetry studies [10–12]. Sample sizes in early-phase dosimetry studies are typically based on feasibility rather than formal calculations due to radiation exposure considerations. The consistent acquisition of high-quality quantitative images across all patients and timepoints validates our methodological approach. Based on these encouraging results, we are currently expanding enrollment to a target of 30 patients for comprehensive dose-response analysis, with stratification by tumor type (soft-tissue vs. bone metastases) and baseline PSMA expression levels, as recommended by recent consensus guidelines for dosimetric studies. Accurate absorbed dose calculations were achieved through systematic acquisitions at 6, 24, 48, and 96 h post-injection with high-quality OSEM reconstructions. Notably, soft-tissue lesions received substantially higher mean absorbed doses than bone metastases (5.03 ± 5.51 Gy vs. 1.61 ± 2.28 Gy), corresponding to dose coefficients of 0.73 ± 0.80 Gy/MBq and 0.25 ± 0.33 Gy/MBq, respectively. Tumor-to-organ absorbed dose coefficient ratios indicated a favorable therapeutic window, with red marrow showing the highest ratio (14.84), followed by adrenal glands (6.35) and salivary glands (4.96), while kidneys and liver showed ratios of 1.55 and 2.25, respectively. These findings support the therapeutic potential of ²²⁵Ac-PSMA-CY313 in mCRPC, demonstrating preferential tumor targeting with acceptable organ-at-risk dose distributions.

The optimization of our multi-timepoint SPECT/CT protocol represents a significant technical advance in ²²⁵Ac dosimetry, addressing the inherent challenges of alpha-emitter imaging. Our acquisition strategy captures both ²²⁵Ac (via ²²¹Fr, 218 keV) and ²¹³Bi (440 keV) emissions, enabling independent TAC fitting with mono- or bi-exponential models. OSEM reconstruction with comprehensive corrections for attenuation, scatter, resolution blur, and, critically, crosstalk and downscatter contamination markedly improved image quality and quantification. This approach effectively mitigates challenges highlighted in prior work—particularly low gamma yield and high SPECT noise in ²²⁵Ac imaging, as described by Tulik et al. [3], who emphasized the complexity of quantitative ²²⁵Ac SPECT due to its decay scheme and multiple energy emissions. By separately tracking daughter nuclides, our protocol provides superior dosimetric accuracy compared with methods relying solely on surrogate isotopes or single–energy-window acquisitions. By separately tracking daughter nuclides, our protocol provides superior dosimetric accuracy compared with methods relying solely on surrogate isotopes or single–energy-window acquisitions.

Our absorbed dose results show good consistency with published studies of other ²²⁵Ac-PSMA agents while revealing important differences. The kidney dose coefficient of 0.23 ± 0.07 Gy/MBq aligns with the range reported in the systematic review by Ma et al. [13], which found ²²⁵Ac-PSMA-617 to be associated with a 3.0% rate of Grade III nephrotoxicity in a meta-analysis of 201 patients. However, our liver dose coefficient (0.16 ± 0.03 Gy/MBq) appears somewhat lower than values reported for ²²⁵Ac-PSMA-617, potentially reflecting structural differences in the CY313 ligand. The residence times observed for critical organs—particularly the liver (251.02 ± 141.38 min) and kidneys (63.56 ± 43.16 min)—are consistent with the biodistribution patterns described by Kratochwil et al. [14]. The prolonged retention of ²²⁵Ac-PSMA-CY313 in specific tissues correlates with higher absorbed doses. The ²²¹Fr and ²¹³Bi daughter-nuclide contributions in our study show expected patterns, with ²²¹Fr demonstrating longer residence times due to its longer half-life, confirming the importance of tracking individual daughter isotopes for accurate dosimetry, as emphasized in the literature.

The substantially higher absorbed doses observed in soft-tissue lesions (5.03 ± 5.51 Gy) compared to bone metastases (1.61 ± 2.28 Gy) can be attributed to interconnected biological and physical mechanisms. Alpha particles exhibit high linear energy transfer (80–100 keV/µm) and a short range (50–100 μm), making dose deposition highly dependent on local PSMA expression and tissue microarchitecture. Soft-tissue lesions typically demonstrate greater vascularization and more homogeneous PSMA expression, facilitating superior radiopharmaceutical penetration and retention. In contrast, bone metastases often exhibit heterogeneous PSMA expression due to sclerotic changes and altered microvasculature, consistent with the differential uptake patterns observed on ¹⁸F-PSMA-CY313 PET. The bone microenvironment may also influence alpha-particle energy deposition due to higher tissue density and altered cellular organization. Additionally, targeting specificity varies between lesion types, with soft-tissue metastases showing more consistent and intense PSMA expression, as reflected in the variable SUVmean values (6.4–21.4) across lesion types in our cohort.

The observed disconnect between ¹⁸F-PSMA-CY313 SUVmean and ²²⁵Ac-PSMA-CY313 absorbed dose reflects fundamental differences in tracer pharmacokinetics and tissue retention mechanisms. While SUV represents a snapshot of PSMA expression and tracer accessibility at a single timepoint, absorbed dose integrates multiple factors including: (1) prolonged tissue retention governed by PSMA recycling kinetics, (2) differential blood flow and vascular permeability affecting initial delivery and subsequent washout, (3) tissue-specific differences in interstitial pressure and lymphatic drainage, and (4) potential differences in binding kinetics between fluorinated and actinium-chelated compounds.

Recent studies have demonstrated that PSMA internalization rates vary significantly between tumor types, with soft-tissue metastases showing 2–3 fold higher internalization rates compared to bone metastases. Tumor cell internalization is a key factor for the efficiency of therapeutic radiopharmaceuticals [15]. A physiologically-based pharmacokinetic (PBPK) model study demonstrated that for imaging, the combination of properties leading to the highest tumour uptake was kon = 0.1 L/nmol/min, koff = 0.01 min − 1 for typical ligand amounts (1–10 nmol). For therapy, the higher the internalization rate, the larger was the required ligand amount for optimal tumour-to-kidney ratios [16]. This differential internalization, not captured by static SUV measurements, directly influences therapeutic radiopharmaceutical retention and absorbed dose.

Furthermore, the tumor microenvironment (TME) evolves in parallel with the cancer clones, altering extracellular matrix composition (ECM), vasculature architecture, and recruiting specialized tumor-supporting cells that favor tumor spread and colonization at distant sites [17]. In bone metastases specifically, the tumor microenvironment creates unique challenges for drug delivery.

Our data support these mechanisms: Patient 01-lesion 2 showed high SUVmean (21.4) but low absorbed dose (0.44 Gy), likely reflecting rapid washout, while Patient 04-lesion 4 demonstrated moderate SUVmean (10.3) but the highest absorbed dose (13.09 Gy), suggesting superior retention kinetics. These findings underscore the importance of multi-timepoint dosimetric assessment rather than relying solely on diagnostic PET imaging for therapeutic planning.

Future studies incorporating immunohistochemical validation of PSMA expression levels and assessment of tumor vascular patterns will help elucidate these mechanisms. Clinical outcome correlations, including progression-free survival and toxicity assessments, are being collected as part of our ongoing long-term follow-up protocol and will be reported separately upon maturation of the data.

The tumor-to–critical organ dose ratios observed in this study suggest a favorable safety profile for ²²⁵Ac-PSMA-CY313. The kidney ratio of 1.55 is particularly encouraging, as the kidneys are the primary dose-limiting organs in PSMA-targeted therapy [18], with EANM guidelines suggesting kidney dose limits of 23–27 Gy for fractionated therapy [19] The exceptionally high red marrow ratio (14.84) indicates minimal marrow toxicity risk, consistent with the low hematological toxicity rates reported by Ma et al. [13], where Grade III anemia, leukopenia, and thrombocytopenia occurred in 7.5%, 4.5%, and 5.5% of patients, respectively. The salivary gland ratio of 4.94 suggests manageable xerostomia risk, although it remains a concern, as 77.1% of patients in the meta-analysis experienced some degree of xerostomia. Importantly, no serious adverse events or Grade III toxicities were observed during the study period, supporting a preliminary safety profile; however, longer follow-up and larger cohorts are required for definitive assessment. These ratios compare favorably with ICRP-recommended constraints and suggest potential for dose escalation or fractionated-therapy protocols.

Several important limitations must be acknowledged in interpreting our results. The small sample size of four patients significantly limits the statistical power and generalizability of our findings. This sample size is insufficient to establish definitive dose-response relationships or identify predictive biomarkers for therapeutic response. The inherent technical limitations of SPECT imaging in assessing ²²⁵Ac’s complex decay chain represent another significant constraint, as noted by Tulik et al. [3], who highlighted challenges in accurate quantification due to background noise and daughter nuclide redistribution. The high inter-patient variability in residence times, particularly for tumors (11.59 ± 19.21 min for ²²⁵Ac), reflects both biological heterogeneity and technical uncertainty. Our use of the ICRP 103 standard phantom represents a compromise between practicality and accuracy. While patient-specific computational phantoms would reduce systematic errors by accounting for individual anatomical variations, recent comparative studies have shown that standard phantom-based calculations typically differ by < 20% from patient-specific models for normal organs. A study evaluating patient-specific dosimetry through Monte Carlo simulations and deep learning demonstrated that patient-specific dose calculations were higher than the dose calculated using the ICRP 110 reference phantom model [20]. Another comparative study of ¹⁷⁷Lu-DOTATATE dosimetry found that patient-specific dosimetry resulted in significantly different estimated absorbed doses compared to those obtained through conventional dosimetry [21].

For our cohort, CT-based organ volume measurements showed coefficients of variation < 15% relative to ICRP reference values, suggesting acceptable approximation for feasibility assessment [22]. Future methodological refinements should prioritize: (1) Implementation of patient-specific voxel-based dosimetry using Monte Carlo simulation directly on patient CT data; (2) Development of hybrid imaging protocols combining SPECT quantification with PET spatial resolution through simultaneous ¹⁸F/²²⁵Ac imaging; (3) Integration of radiomics features and machine learning algorithms to predict dose distribution from baseline imaging, potentially reducing the need for multiple post-therapy scans; and (4) Exploration of novel reconstruction algorithms incorporating deep learning for enhanced signal extraction from low-count ²²⁵Ac daughter emissions.

The assumption of secular equilibrium for daughter nuclides may not hold in all tissue compartments, and the potential for daughter nuclide migration could affect dose distribution accuracy.

The promising preliminary results of this study warrant expansion to multi-center validation studies with larger patient cohorts to establish robust dose-response relationships and optimize treatment protocols. Future research should focus on developing more sophisticated individualized dose calculation models that incorporate patient-specific anatomy, tumor heterogeneity, and real-time biokinetic modeling. The methodology established here could be readily adapted to other ²²⁵Ac-labeled radiopharmaceuticals, potentially advancing the entire field of targeted alpha therapy dosimetry. Integration of advanced imaging techniques, such as the AI-based organ segmentation using MONAI algorithms employed in our study, represents a promising direction for automated and standardized dosimetric workflows. Understanding the sequencing effects of different radiopharmaceutical therapies will be crucial, necessitating studies comparing treatment-naive patients with those previously exposed to ¹⁷⁷Lu-PSMA therapy [23, 24] The development of predictive models combining dosimetric parameters with molecular biomarkers and imaging characteristics could enable personalized treatment planning and improved patient selection criteria.

Conclusion

This study makes significant methodological contributions to the emerging field of ²²⁵Ac-based targeted alpha therapy dosimetry, establishing a practical and reproducible approach for quantitative dose assessment using clinical imaging systems. Our findings demonstrate that ²²⁵Ac-PSMA-CY313 exhibits favorable dosimetric characteristics with preferential tumor targeting and acceptable organ-at-risk dose distributions, supporting its continued clinical development for mCRPC treatment. The substantially higher absorbed doses achieved in soft tissue lesions compared to bone metastases provide important insights for patient selection and treatment expectations. These results directly address the research questions posed in our introduction regarding the feasibility of accurate ²²⁵Ac dosimetry and the comparative effectiveness across different metastatic sites. Our work advances the field of nuclear medicine and molecular imaging by providing a validated methodology for ²²⁵Ac dosimetry that can inform dose optimization, safety assessment, and treatment personalization. The established therapeutic window and preliminary safety profile support the potential for ²²⁵Ac-PSMA-CY313 to become a valuable addition to the theranostic armamentarium for advanced prostate cancer, while the dosimetric methodology developed here contributes to the broader goal of precision medicine in targeted radiopharmaceutical therapy.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

All authors contributed to the study’s conception and design. Yekuan Shi, Huaijia Tang, Huajian Gu and Fei Luo performed material preparation, data collection and analysis. Hao Zhang wrote the first draft of the manuscript. Suping Li, Daiyuan Ma and Tielong Tang completed the revision of the manuscript. All authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

This work was supported by the Sichuan Science and Technology Program (No. 2024NSFSC0668), Research and development project of Affiliated Hospital of North Sichuan Medical College (No. 2022LC004), and the Opening Project of Medical Imaging Key Laboratory of Sichuan Province (No. MIKL202409).

Data availability

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

Declarations

Ethics approval consent to participate

This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Ethics Committee of The Affiliated Hospital of North Sichuan Medical College (Date 22/03/2024/No.2024ERJ89-1). Informed consent was obtained from all individual participants included in the study.

Consent for publication

The authors affirm that human research participants provided informed consent for publication of the images in Figs. 2 and 3 and the supplementary materials.

Clinical trial registration

Chinese clinical trial registry: ChiCTR2400083275, Registered 19 April 2024.(https://www.chictr.org.cn).

Competing interests

The authors have no relevant financial or non-financial interests to disclose.