In vivo real-time positron emission particle tracking (PEPT) and single particle PET

The study addresses the challenge of translating positron emission particle tracking (PEPT) to biomedical applications. While PET images average distributions of many molecules or particles, PEPT can localize and track an individual positron-emitting particle with high spatiotemporal resolution, offering unique potential to study blood flow dynamics, cell migration, and complex multiphase flow in vivo. Progress has been limited by the lack of biocompatible tracers that can be radiolabelled with sufficiently high specific activity and by practical methods to isolate and inject a single sub-micrometre particle without risk of capillary embolization. The authors propose that inorganic sub-micrometre silica particles radiolabelled with 68Ga could overcome constraints of living cell tracers (limited radioactivity tolerance and radiolysis), enabling higher activity per tracer and real-time tracking. The aims were to synthesize suitable silica particles, maximize 68Ga labelling to achieve very high specific activity per particle, develop a robust method to isolate and manipulate a single particle, and implement a PEPT reconstruction compatible with preclinical PET list-mode data to track a single particle in vivo.

PEPT has been extensively applied in industrial systems for tracking particles in opaque, dense, or multiphase environments, with cameras such as the Forte enabling tracking at ~1 m s−1 with ~0.5 mm precision at ~250 Hz. Advancements like superPEPT promise even better spatiotemporal resolution. Biomedical interest includes quantifying blood flow velocity, density, and dynamics, but has been constrained by tracer limitations. Prior related biomedical work demonstrated whole-body PET tracking of a single breast cancer cell radiolabelled with 68Ga-mesoporous silica nanoparticles, achieving specific activities of 30–110 Bq per cell, lung trapping within 2–3 s post tail-vein injection, and average cell velocity ~50 mm s−1. However, living cell tracers are limited by radiation-induced damage and low achievable activity, reducing detectability at high speeds and precluding real-time PEPT. The literature also supports chelator-free labelling of oxophilic radiometals (e.g., 68Ga) onto silica substrates, and highlights PEPT algorithm developments and potential biases or background issues in PET systems using LYSO crystals.

- Particle synthesis and characterization: Sub-micrometre silica particles (smSiP) were synthesized via a modified Stöber method using TEOS in ethanol with ammonia catalyst and KCl electrolyte; controlled dropwise addition at 50 °C yielded highly monodisperse spheres (0.95 ± 0.05 µm). Particles were purified by centrifugation and ethanol washes, then dried. Characterization included SEM (size/shape), EDS (SiO2 composition), FT-IR (Si–O and Si–OH vibrations at 1090, 950, 795 cm−1), and ζ-potential (−41.1 ± 3.3 mV; PEGylated smSiP-PEG shifted to −28.4 ± 3.1 mV). - 68Ga production and concentration: 68Ga was eluted as [68Ga]GaCl3 (4 ml, 0.1 M HCl) from a 68Ge/68Ga generator. A cation-exchange concentration protocol (Strata-X-C cartridge; acetone/HCl washes and elution) reduced volume to 50 µl HEPES buffer (pH 4.9) with 86.2 ± 8.5% recovery in ~20 min. - Radiolabelling (bulk suspensions): smSiP suspensions (0.002–1 mg ml−1 in 0.5 M HEPES, pH 4.9) were reacted with concentrated 68Ga at 90 °C for 30 min. Radiolabelling yield (RLY) assessed by radio-TLC was 90–100% at 0.125–1 mg ml−1 and 75.0 ± 9.2% at 2 µg ml−1; radiochemical purity (RCP) after purification was 98.2 ± 1.6%. Radiochemical stability in human serum at 37 °C remained 98.5 ± 1.0% at 3 h. - Particle quantification and selection: Flow cytometry with CountBright beads was evaluated for sorting fixed numbers (500–2,000 particles), which showed high deviations; instead, it provided accurate concentration estimates (208 ± 8 particles µl−1 at 0.05 mg ml−1; n=10), enabling preparation of defined small-particle-number batches. - Radiolabelling of 500 particles: Batches containing 500 smSiP were radiolabelled (90 °C, 30 min) with polysorbate 80 present. A multi-step EDTA-based purification (10 mM, then 1 mM, then 0.1 mM EDTA with polysorbate 80 in PBS, multiple washes) removed unbound and colloidal 68Ga. Control reactions without particles had RLY <0.1%; with 500 smSiP, RLY was 1.9 ± 1.3%. - Single-particle isolation (fractionation): The labelled mixture at ~1 particle µl−1 was aliquoted into four serial fractions (final 12.5 µl each) using a defined splitting scheme. Gamma counting identified samples with most activity in a single fraction; after sonication, a second fractionation removed residual activity from non-particle fractions. Autoradiography of a TLC strip confirmed a single hotspot for particle-containing fractions. The isolated single particle specific activity was 2.1 ± 1.4 kBq. - PEGylation: smSiP were grafted with mPEG-silane (PEG5k) in 98% ethanol with ammonia; yield ~43% (estimated by unbound PEG mass). PEGylated particles preserved morphology; ζ-potential shifted to −28.4 ± 3.1 mV. Radiolabelling of 500 smSiP-PEG used identical conditions (RLY 3.3 ± 3.1%); three fractionation steps yielded RCP >99% and specific activity 3.1 ± 3.2 kBq per particle. - In vitro PET/CT and cannula test: A single 68Ga-smSiP (~2.1 kBq) in a tube was imaged over 2 h with 5-min frames to confirm detectability. Cannula delivery tests showed particle retention in the cannula in 6/10 attempts; a 50 µl PBS flush ensured 10/10 successful deliveries. - In vivo PET/CT: Healthy BALB/c mice were anesthetized, cannulated, and injected with a single 68Ga-smSiP (0.4–1.9 kBq; n=4) or 68Ga-smSiP-PEG (0.95–2.9 kBq; n=2) in 100 µl PBS, followed by 50 µl PBS flush. PET list-mode data were acquired for 2 h (400–600 keV, Tera-Tomo 3D reconstruction; voxel 0.4 mm3) with CT coregistration. ROI quantification tracked activity over time. - Ex vivo biodistribution and autoradiography: After imaging, organs were excised for gamma counting (%IA g−1). Lungs were dissected; 20 µm cryosections were scanned to locate the single radioactive slice; autoradiography quantified Bq using calibration with 68Ga standards. Activity at each protocol step was decay-corrected to injection time and compared across measurement modalities. - PEPT reconstruction and tracking: PET list-mode data were processed using the Birmingham PEPT method. LoRs were converted to spatial coordinates; iterative minimum distance point (MDP) estimation with outlier LoR rejection was applied. Adaptive LoR sample sizes were used: early phase <60 s, 100–200 LoRs with f=0.1 (1–5 s intervals); later 1,000–2,000 LoRs (30–60 s intervals). Particle speed was computed from consecutive positions; positioning error decreased at longer sampling intervals. Energy window lower discriminator was set to 400 keV to minimize intrinsic LYSO background; randoms were rejected prior to PEPT analysis.

- Synthesized highly monodisperse, nonporous silica particles (smSiP) with diameter 0.95 ± 0.05 µm; ζ-potential −41.1 ± 3.3 mV (smSiP) shifting to −28.4 ± 3.1 mV after PEGylation. - Chelator-free 68Ga radiolabelling of smSiP achieved high radiolabelling yields in bulk: 90–100% at 0.125–1 mg ml−1 and 75.0 ± 9.2% at 2 µg ml−1; radiochemical purity 98.2 ± 1.6% after purification; serum stability 98.5 ± 1.0% at 3 h (37 °C). - Accurate particle concentration by flow cytometry (208 ± 8 particles µl−1 at 0.05 mg ml−1); sorting fixed numbers showed large deviations, favoring quantification over direct sorting for precise small-number batches. - Radiolabelling of 500-particle batches yielded RLY 1.9 ± 1.3% (control without particles <0.1%); single-particle isolation by fractionation and autoradiography confirmed one hotspot; specific activity 2.1 ± 1.4 kBq per particle (PEGylated: RLY 3.3 ± 3.1%, specific activity 3.1 ± 3.2 kBq). - PET sensitivity: a single ~2.1 kBq particle was detectable in vitro with 5-min frames over 2 h. - Cannula delivery: Particle was retained in the cannula 6/10 times without flush; a 50 µl PBS flush ensured 100% delivery (10/10). - In vivo PET/CT (BALB/c mice): Single 68Ga-smSiP (0.4–1.9 kBq) produced a clear single hotspot in lungs within 5 min. In one case, a subtle relocation within the lung occurred between 5–10 min, then remained static. ROI quantification yielded 97 ± 3% of injected activity localized to the lungs across frames; ex vivo biodistribution showed exclusive lung uptake; a single 20 µm lung slice contained all radioactivity. Cross-method quantification deviation averaged 8.5 ± 2.2% (autoradiography vs gamma counting). - PEPT tracking: Early trajectory captured from injection site through lower abdomen to heart, then into lungs; initial particle speed ~48 mm s−1. Later anterior–posterior oscillations ± ~2 mm consistent with breathing motion. Positioning error ~±1.9 mm early, ~0.9 mm later with larger LoR samples. - PEGylated particles (0.95–2.9 kBq) behaved similarly in vivo: single stationary lung hotspot; mean 96.9 ± 2.8% of injected activity in lungs; exclusive lung uptake ex vivo; single-slice autoradiography hotspot; cross-method deviation 14.8 ± 6.6%. - Demonstrated feasibility of real-time single-particle tracking using PEPT on a preclinical PET/CT system at kBq activities—around three orders of magnitude lower activity than typical PET radiotracers.

The work directly addresses the key barriers to biomedical PEPT by creating a tracer and workflow that provide very high specific activity per single particle and reliable isolation/manipulation of that particle. High radiochemical purity and serum stability enabled clear PET imaging at kBq activities within minutes post-injection. Implementing the Birmingham PEPT algorithm on preclinical PET list-mode data provided unique quantitative measures not accessible by standard PET, including initial particle velocity (~48 mm s−1) and respiratory-driven motion within the lungs, offering a pathway to assess haemodynamics and organ motion in vivo. Rapid lung sequestration was observed for both uncoated and PEGylated ~950 nm rigid silica particles. The authors suggest this may result from size/rigidity relative to small (~2 µm) pulmonary capillaries and from phagocytic uptake by lung-marginated neutrophil pools or other phagocytes. Consequently, although the single-particle approach achieves high detectability and precise tracking, systemic circulation beyond the lungs was not observed. Future strategies should evaluate smaller and more flexible particles and transient modulation of lung neutrophil phagocytosis to enable broader vascular tracking. Methodologically, PEPT performance depends on LoR sample size: smaller samples capture rapid motion at the cost of higher positional uncertainty, while larger samples reduce error but lower temporal resolution. Potential biases include positron range effects near heterogeneous lung–tissue interfaces (more pronounced for higher-energy emitters like 68Ga), and intrinsic background from LYSO scintillators. The energy window lower discriminator (400 keV) and random rejection were used to mitigate background. Algorithmic constraints and the low-activity/fast-motion regime may limit path curvature resolution (e.g., through the IVC to heart), but did not affect feasibility. Overall, results demonstrate the capability and repeatability of in vivo single-particle PEPT using standard preclinical PET hardware.

The study establishes, for the first time, a single radiolabelled particle platform for in vivo nuclear imaging and real-time PEPT tracking. Homogeneous ~950 nm silica particles were synthesized and radiolabelled chelator-free with 68Ga to achieve unprecedented specific activities per particle (~2–3 kBq) with high purity and stability. A practical fractionation method isolated a single particle, verified by autoradiography. PET/CT at very low activities (0.4–2.9 kBq) yielded high-quality images within 5 min and, combined with the Birmingham PEPT algorithm, enabled tracking of particle trajectories, velocities, and respiratory motion in mice. Although particles accumulated in the lungs, the proof-of-concept demonstrates strong potential for quantitative assessment of haemodynamics and organ motion. Future work should focus on optimizing particle size, flexibility, and biointeractions to promote systemic circulation and on refining PEPT algorithms for small-animal regimes. Clinically, the approach could synergize with total-body PET to study blood flow complexities, vessel pathology, and tumor motion with minimal administered activity and material, supporting applications in diagnosis, therapy guidance, and surgical planning.

- Lung sequestration of ~950 nm rigid silica particles prevented systemic circulation beyond the pulmonary capillary bed, limiting current tracking to the lungs. - Potential positional bias due to positron range and tissue heterogeneity at lung–tissue interfaces, especially with higher-energy emitters like 68Ga, requires further evaluation. - Intrinsic LYSO background in PET scanners can introduce random coincidences; mitigations were applied (400 keV lower discriminator), but residual effects and generalizability across systems remain considerations. - Algorithmic and sampling constraints (limited LoR counts at early times) reduce trajectory detail and increase positional error during rapid motion, potentially simplifying curved paths to apparent linear segments. - Small sample sizes for PEGylated particle in vivo studies (n=2) limit generalizability; overall in vivo cohort sizes remain modest. - Radiolabelling yield for 500-particle batches, though notable for so few particles, is low in absolute terms (1.9 ± 1.3%), which may affect throughput and reproducibility.