Online program Overview Session: WS-04

Abstract

Cancer treatment has shifted from empirical, toxicity-based dosing toward precision oncology guided by the specific genetic and molecular profiles of individual tumors. Within this framework, physicians, medical physicists, engineers, and scientists share the responsibility of developing tools to deliver high-precision treatments, predict therapeutic responses, and track disease progression. Radiotheranostics contributes to this approach by leveraging advanced radiochemistry to design highly selective molecules that deliver diagnostic or therapeutic radionuclides to target cells. Although routine at major medical centers for metastatic disease, broader adoption requires advances in imaging performance and targeted radiopharmaceuticals. Conventional PET and SPECT systems remain constrained by physical limits in spatial resolution, sensitivity, and quantitative accuracy, all of which are critical for clinical translation. Addressing these gaps requires higher sensitivity and spatial resolution, refined reconstruction algorithms for improved image clarity, precise image-based dose calculations, and system standardization. These needs drive hardware development toward total-body and organ-specific PET and SPECT systems, as well as new gamma, x-ray, and hybrid imaging methods. For high linear-energy transfer approaches like targeted alpha therapy, ex vivo micro-scale imaging is increasingly important to measure dose heterogeneity at the suborgan/cellular level. Concurrently, artificial intelligence is improving image reconstruction and dose prediction. Finally, key translational challenges must be addressed: ensuring trial reproducibility, integrating new methodologies into clinical workflows, and demonstrating cost-effectiveness alongside tangible patient benefit. This workshop brings together experts across these areas to collaborate and advance next-generation radiotheranostic technologies.

Abstract

The development of radiotheranostics would greatly benefit from quantitative imaging techniques that can simultaneously map the distributions of diagnostic and therapeutic radionuclides across spatial, temporal, and functional domains. To address this need, we will present the design of a proposed preclinical PET/SPECT platform consisting of a long-axial-FOV PET scanner with embedded CZT pinhole camera modules. The system is designed to simultaneously image both PET tracers and typical SPECT isotopes, as well as therapeutic radionuclides, such as Lu-177, Ac-225, Ra-223, and At-211.

The PET subsystem is based on dual-ended-readout BGO detectors with a diameter of 16 cm and an axial length of 16 cm. The BGO detectors in the central portion are purposefully positioned to leave small gaps to accommodate micro-gamma-camera elements (MGCE’s) based on high-resolution CZT detectors with pinhole collimation for radionuclides. The MGCE’s are designed to cover a wide dynamic range of 20 keV to 700 keV, so the combined PET and SPECT system will be capable of imaging virtually all commonly used SPECT isotopes and therapeutic radionuclides. In addition, data acquisition for the PET and SPECT subsystems is synchronized via a unified clock-and-trigger architecture. It enables coincidence detection of both PET and single-photon signals. Correlations among detected photons in the spectral and temporal domains are used to reject down-scattering noise in SPECT acquisition and to identify cascade emission of gamma rays from Lu-177 and Ac-225.

In this presentation, we will discuss a comprehensive design study of an integrated PET/SPECT imaging system and its potential to characterize the complex pharmacokinetics of radionuclide therapies.

Abstract

Molecular radiotherapy requires personalized dosimetry for efficient treatment and reduced toxicity, but current clinical imaging systems and miniaturized gamma-cameras lack the required capabilities. We developed a mobile gamma-camera with a 10×10 cm² field of view tailored for quantitative assessment of ¹³¹I biokinetics during therapy of thyroid diseases. The system combines interchangeable 3D-printed tungsten collimators, a monolithic scintillator read out by a SiPM array, and commercial electronics. It achieves an intrinsic spatial resolution of 1.15 mm FWHM and negligible deadtime up to 150 kcps. Images are reconstructed in real time using an experimental data-trained convolutional neural network. The collimators and shielding were adjusted with Monte-Carlo simulations in order to reduce septal penetration and scattering. Preclinical studies demonstrated high quantitative accuracy, with activity bias below 5%, even for small sources. These performances stem from the camera’s compactness and mobile design, which enables shorter detector-to-target distances, optimized viewing angles, as well as from its high overall spatial resolution (4–6 mm at 365 keV). Its mobility also avoids transporting radioactive patients, reducing radiation exposure for staff and other patients. A first clinical feasibility study involving 20 patients with benign thyroid diseases was completed in March 2025, to assess patient acceptance and image quality during planning and post-treatment follow-up, compared with a standard gamma camera (Syngula). Results indicate unprecedented image quality around 400 keV with direct quantitative information, particularly for post-therapeutic imaging, enabling  improved assessment of heterogeneous ¹³¹I distribution, including for Graves' disease. This highlights the potential of the camera for personalizing administered activity and, more importantly, estimating the dose actually absorbed during treatment, directly linked to therapeutic success.

Abstract

We are developing a Personalized Remote Radiation Tracking Portable Organ Dosimetry Device (PRRT PODD) for at-home monitoring of patients receiving radiopharmaceutical therapy. The goal of this work is to provide a user and facility friendly device to enable organ at risk (OAR) dosimetry for patients undergoing Lu-177 DOTATATE treatment. The PRRT PODD consists of a portable frame with wings that encircle the patient's torso from the base of their kidneys to the top of their liver. In addition, there is a registration pad that enables precise positioning of the patient in the PODD for daily measurements. The patient lies down in a PODD template during their therapy qualifying PET/CT scan. This registers the PODD to their anatomy. Using software tools, a sensitivity matrix between the patient’s OAR and tumors is determined for a two-dimensional (2D) array of potential detector locations along the surface of the PODD. Using this matrix, 16-32 detector locations are selected based upon the patient’s specific anatomy to enable activity estimation for the patient’s OAR. After a quantitative SPECT/CT scan acquired 24-hours after treatment, a PODD measurement will be made in clinic to normalize with the SPECT/CT scan. The PODD will then be sent home with the patient for fourteen 5-minute daily measurements. The PODD utilizes a carbon fiber base plate, a vacuumable Styrofoam registration pad, 3D printed wings, compact pinhole SiPM-based radiation detectors, battery-operated acquisition electronics, and a cell phone app that supports wireless transmission from the PODD’s detector to a remote computer workstation. Our updated simulation software supports placement of a virtual 2D array of detectors on the polygonal surface defined by the PODD. An overview of the PODD development along with results from simulations based on human imaging examples will be provided at the workshop.

Abstract

In conventional ¹⁷⁷Lu imaging that relies solely on single photons or cascade gamma photons, the detection efficiency is fundamentally limited by the low photon emission probabilities. The decay of ¹⁷⁷Lu produces not only γ photons but also characteristic X-rays through the internal conversion process. In this work, we propose a multi-type time coincidence imaging strategy that exploits characteristic X-rays together with γ photons to improve the detection efficiency and image quality of ¹⁷⁷Lu imaging. The feasibility of X–X, X–γ, and γ–γ coincidence detection is analyzed based on the decay characteristics of ¹⁷⁷Lu, and a full-ring coincidence imaging system is subsequently employed for performance evaluation. Experimental results show that the integration of multiple coincidence-event types increases the system detection efficiency by 107.60\% and improves the contrast-to-noise ratio (CNR) by 8.39\% compared with conventional γ–γ coincidence imaging, while preserving nearly unchanged spatial resolution.

Abstract

Cascade gamma coincidence imaging is an alternative to conventional single photon imaging for therapeutic radionuclides such as ¹⁷⁷Lu by exploiting time-correlated cascade photon pairs. However,  high spatial resolution and reliable quantitative accuracy remain challenging because of photon attenuation, scatter, partial volume effects. We developed a cascade-coincidence list-mode maximum-likelihood expectation-maximization (CCLM-MLEM) framework for a full-ring ¹⁷⁷Lu cascade gamma coincidence imaging system. Attenuation correction (AC) and scatter correction (SC) were incorporated into reconstruction, followed by absolute quantification and post-reconstruction partial volume correction (PVC). The system response matrix was modeled according to the cascade-photon detection geometry. The framework was evaluated using seven-point-source and hot-rod phantoms for spatial-resolution assessment, the modified image-quality phantom for quantitative evaluation, and the Mouse Whole-Body (MOBY) digital phantom for small-animal imaging. Direct backprojection (DBP) was used for comparison. CCLM-MLEM reduced the mean point-source FWHM from 2.20±0.10 mm with DBP to 1.51±0.27 mm and resolved 1.0-mm rods, compared with 1.5-mm rods using DBP. In the modified image-quality phantom, AC and SC improved activity and contrast recovery at comparable noise levels. In the MOBY phantom, the relative bias for a 4-mm tumor decreased from -57.19% without correction to 1.83% after AC, SC, and PVC. These results support high-resolution quantitative ¹⁷⁷Lu cascade gamma coincidence imaging and its potential for preclinical radiopharmaceutical imaging and dosimetry.

Abstract

Background: ²²⁵Ac targeted alpha therapy requires quantitative post-therapy SPECT for verification and dosimetry. Multiple emissions, scatter, and inter-window crosstalk complicate reconstruction.
Methods: We developed a ¹⁷⁷Lu-informed deep support-driven kernel (DSDK). A support generator was pretrained on 63 clinical [¹⁷⁷Lu]Lu-DOTATATE SPECT scans using simulated projections at multiple count levels and a projection-consistency objective. For ²²⁵Ac reconstruction, the learned support was used only to construct an image-domain kernel prior; activity remains governed by measured multi-energy projections and a Poisson model. Scatter and crosstalk are explicit projection-space physical models using MC-optimized DEW and self-consistent crosstalk correction for 218-keV, 440-keV, and joint-window reconstruction. Evaluation included MC calibration, single-sphere NEMA IQ scatter-window validation, dual-sphere IQ CNR-COV assessment at 15/30/60 s, and 10-patient [²²⁵Ac]Ac-DOTATATE SPECT/CT evaluation. Baselines included OSEM, DDIP-TTA, and PET-guided methods; PET was not a DSDK input.
Results: MC selected a narrow 5% lower scatter window beside ±10% photopeak windows, with projection MSEs of 0.16/0.10 at 218/440 keV. Single-sphere joint-window CNR/SBR rose from 3.032/8.390 without scatter correction to 3.104/9.081. Dual-sphere DSDK with crosstalk improved the CNR-COV trade-off, especially at 15/30 s; at a representative joint window, CNR/COV was 22.2/0.87 versus 17.1/1.19 for DDIP-TTA. Clinically, joint-window DSDK achieved TBR 2.43 ± 1.27 and the highest lesion CNR among non-PET-prior methods. DSDK also achieved the highest joint-window lesion SSIM, 0.88 ± 0.05, without a PET-derived prior.
Conclusion: DSDK integrates support-level transfer, projection-consistent statistical reconstruction, and crosstalk physical modeling for quantitative multi-window ²²⁵Ac SPECT and may support patient-specific dosimetry.

Abstract

Immediate visualization of therapeutic radionuclide distributions could provide useful spatial feedback during or shortly after interventional radionuclide therapy, but this use case requires a portable camera to acquire images quickly without close contact with the target region. We evaluated a Telephoto-Lens Magic-Cube portable gamma camera for rapid ¹⁷⁷Lu imaging using hybrid front-end long-hole geometric encoding and detector-side self-collimation encoding. A measured ¹⁷⁷Lu system matrix covered an 80 × 80 mm² field of view and source-to-camera distances of 8–22 cm. Across 54 point-source positions from 10–20 cm, mean absolute localization errors were 0.26, 0.32, and 2.58 mm along the x, y, and z axes, respectively. Representative dual-source reconstructions showed visually distinguishable hotspots separated by 5–7 mm at distances of 10–18 cm. At approximately 16 cm, all principal hotspots remained visible in 1 s reconstructions of three phantoms and an in vivo rat liver, with NCC values of 0.942–0.992 and transverse hotspot displacements of 0.27–1.12 mm relative to 300 s references. An excised rat liver containing 72.2 μCi of residual ¹⁷⁷Lu also showed a localized activity region. These results demonstrate the feasibility of rapid ¹⁷⁷Lu distribution imaging at extended working distances for procedure-room theranostic assessment.

Abstract

Theragnostic imaging is an emerging approach in precision oncology that combines molecular imaging with targeted radionuclide therapy. Its safe and effective implementation requires not only advanced radiopharmaceuticals, but also reliable radiation detection systems, standardized quality control (QC), and comprehensive quality assurance (QA) throughout the imaging and treatment workflow. This work focuses on the role of radiation detection technologies in theragnostics, including detector performance, calibration and verification of QC equipment, and radiopharmaceutical quality assessment.

Particular attention is given to radionuclide and radiochemical purity, activity calibrator verification, measurement traceability, and harmonized QA protocols. The work also addresses challenges in conventional gamma spectrometry, particularly spectral overlap in composite and low-activity samples, highlighting maximum-likelihood estimation (MLE) approaches for improved activity quantification of theragnostic radionuclides using NaI(Tl) detectors. In addition, emerging detector materials such as graphene quantum dots are considered for their potential to enhance radiation detection technologies.

The study discusses the challenges of implementing theragnostic programs in emerging healthcare systems, emphasizing the need for international standards, specialized training, appropriate infrastructure, and adaptable regulatory frameworks. These elements are essential to achieving accurate personalized dosimetry, reliable treatment planning, and safe, standardized theragnostic practices.

Abstract

Clinical and preclinical demand for higher-precision molecular imaging across an extended energy range is pushing gamma-ray detectors beyond the limits of conventional SPECT. CZT detectors offer excellent energy resolution and good sensitivity for high-energy isotopes, but position resolution and spectral distortions, including charge sharing and depth dependence, remain limited by pixel size, detector thickness, and cost of scaling readout electronics.

A novel approach is presented, using machine learning models trained on high-fidelity simulations to extract sub-pixel and depth-of-interaction (DOI) information from anode signals. Implemented in a prototype CZT gamma camera, the method achieves 10×10 sub-pixelisation of 2 mm pixels, DOI resolution of ~500-800 µm, and gamma-camera-level energy resolution with spatial precision approaching mid-range x-ray detectors, pointing toward a new class of high-resolution SPECT instrumentation.

A key challenge is the increased demand on readout speed and bandwidth. Accurate positioning requires acquisition of positive and negative signals across pixel clusters, increasing data volume and processing requirements while limiting real-time operation and scalability.

To address this, we explore neural network inference integrated directly into the detector front-end, enabling photon-by-photon signal processing. A dedicated ASIC will digitise and buffer electrode signals, which are then characterised by a neural-state machine that outputs high-level interaction data. A pulse-level simulation framework is being developed to train and validate front-end neural network models.

Abstract

Theragnostic interventions commonly require both fluoroscopic and nuclear image guidance. In liver tumor radioembolization, for example, ⁹⁰Y microspheres are injected into the liver bloodstream under fluoroscopy. First, ^99mTc microspheres are administered, and a SPECT scan is made. Therapeutic ⁹⁰Y microsphere delivery is typically performed two weeks later, based on the results of the pre-treatment procedure. If a SPECT scan of the ^99mTc microsphere distribution can be made inside the intervention room, the pre-treatment and treatment procedures can be merged into a single session, enabling in-situ adaptation of the catheter position and near-real time, personalized dosimetric treatment planning. Moreover, the burden for the patient as well as the overall cost of treatment are reduced. To this end, a first prototype of an interventional X-ray and scintigraphy imaging (IXSI) system was built at UMC Utrecht. It consists of a C-arm equipped with an X-ray tube and a detector stack consisting of an adapted X-ray flat panel detector positioned in front of a gamma camera fitted with a cone-beam collimator. Here, we report on two recent developments. First, a clinical trial is ongoing in which the first patients with liver cancer are receiving single-session radioembolization using the IXSI prototype. Second, X-ray imaging is currently limited to 70 kVp and the X-ray tube is translated off-focus from the focal point of the cone-beam collimator by 5 cm, to avoid blinding of the gamma camera by X-rays penetrating through the flat-panel detector. To enable concurrent, intrinsically co-registered X-ray and nuclear imaging using X-ray spectra up to 120 kVp, a novel detector stack is under development, comprising a gamma camera based on a low-afterglow, pixelated NaI(Tl) scintillator read out by SiPMs. An update of these developments will be presented at the workshop.

Abstract

An increasing number of isotopes have seen use in theranostic applications but have generally been limited by poor image quality or low quantification accuracy. These limitations are caused by factors like low emission intensities, clustered photopeaks, or lower system sensitivity. These factors drive the use of easier to image surrogate isotopes for biodistribution studies. Theranostic SPECT with CZT semiconductors reduces the need for traditional cut-and-count measurements and can produce quality reconstructions for isotopes that are otherwise difficult to image. The Alpha-SPECT^TM Mini preclinical device utilizes CZT semiconductors as the detecting medium, providing the highest energy resolution with a full-body field-of-view for mice. This SPECT imager has a 2.5% single-pixel energy resolution at 140 keV, and it has been used in the imaging of popular isotopes such as ²²⁵Ac, ²¹³Bi, ²¹²Pb, ²¹¹At, ¹⁷⁷Lu, ¹¹¹In, and ^99mTc.
The detectors’ superior energy resolution allows for discrimination of the clustered photopeaks in the X-ray energy regions of ²²⁵Ac and ²¹²Pb. Projection windows with sufficient separation from the tungsten characteristic X-ray at 60 keV allow for removal of excess scatter in reconstructed images, one of the major image degradation factors in low-emission isotope imaging. The system’s high sensitivity enables the dynamic imaging of fast decaying isotopes over the whole body of a mouse, such as ²¹³Bi with its 45-minute half-life. The dynamic imaging function has been utilized for binning as small as 2 seconds for ^99mTc, 1 minute for ²¹¹At, 15 minutes for ²¹³Bi, and 30 minutes for ²²⁵Ac. This functionality unveils important information on theranostic isotope distribution shortly post-injection.

Abstract

Time-of-flight (TOF) improves effective sensitivity and image signal-to-noise ratio in positron emission tomography (PET). Other medical applications that benefit from high time resolution include ion range verification in hadron therapy via prompt gamma imaging and dose verification in targeted alpha therapy via cascade gamma imaging. All these applications would benefit from intrinsically radiation-free scintillators, motivating the search for alternatives to LYSO. GAGG is a strong candidate due its high light yield, which enables good energy resolution, and to its fast decay time, which makes GAGG's time resolution very competitive. At high gamma energies — between 1–7 MeV, the range relevant to prompt gammas in hadron therapy and to characteristic gammas emitted in cascade by alpha-emitting isotopes — BGO also becomes particularly attractive, as its prompt Cherenkov photon yield increases significantly in this range.

Here, we present a proof-of-concept measurement of the time resolution of BGO and GAGG at ~1 MeV. Pairs of 3×3×15 mm³ BGO and GAGG crystals, coupled to FBK-NUV-MT SiPMs, were read out using state-of-the-art high-frequency electronics. Each crystal pair detected in coincidence the gamma cascade from a ⁶⁰Co source (1173 keV and 1333 keV gamma rays emitted with a sub-picosecond half-life). In our preliminary studies, we achieved a coincidence time resolution (CTR) FWHM of 120 ps for GAGG and 188 ps for BGO, while for 511 keV the CTRs are 179 ps and 245 ps, respectively. Additionally, simulations are used to investigate Cherenkov photon detection yield and time resolution as a function of gamma-ray energy. These results highlight the potential of GAGG and BGO both for TOF-PET and other emerging applications involving high-gamma energies.

Abstract

The growth of radiopharmaceutical therapy (RPT) has created an increasing need to characterize RPT distributions at microscopic spatial scales, where heterogeneity can strongly influence absorbed dose and biological response. This is particularly important for alpha-particle therapies, where information of the activity distributions is needed at spatial scales comparable to alpha-particle ranges in tissue (40–100 µm). The ionizing-radiation Quantum Imaging Detector (iQID) is a high-resolution imaging system developed for quantitative digital autoradiography of alpha- and beta-particle emitters at these spatial scales. iQID combines scintillation-based detection with CMOS-camera imaging and real-time processing to localize individual particle interactions and record their spatial, temporal, and intensity information. Unlike phosphor imaging plate autoradiography, which integrates the cumulative signal from all radioactive decay events over an exposure, iQID detects and localizes individual scintillation events in real time. Event size, intensity, position, and time are retained as list-mode data, enabling high-resolution quantitative imaging of radioactive sources and radiolabeled tissue sections. Event characteristics can also distinguish alpha- and beta-particle interactions, allowing both emission types to be imaged simultaneously. The spatial and temporal information retained by iQID provides additional capabilities for characterizing radioactive sources. Changes in activity and decay behavior can be followed over time, and measured decay signatures can help distinguish radionuclides within mixed sources. This presentation will describe the iQID imaging approach and highlight applications in quantitative alpha/beta autoradiography, source characterization, isotope identification, radiopharmaceutical biodistribution, and micro-scale dosimetry.

Abstract

Introduction: Theranostic cancer treatments rely on ligands which can be tagged with two different isotopes. One for imaging and one for treatment. The imaging ligand is used as a proxy for monitoring the ligand used to treat the target tissue. Key to using the imaging agent to measure the dosimetry of the treatment ligand is knowing the bio-equivalence of the ligand pair. Bio-equivalence is defined as equal biological interaction of two ligands within tissue.  This can be achieved with the Quantitative Particle Identification Autoradiograph (QPID).

Materials and Methods: The next generation of spectral autoradiograph has been developed which mates a Timepix3 sensor developed by the CERN Medipix collaboration with a CeBr3 scintillation crystal. This allows one to use the gamma emission energy in forming autoradiograph images of tissue samples harvested in bio-distribution studies. The QPID time resolution is on the order of 1.6 ns allowing for the detection of single decays, thus achieving alpha and beta identification through ionization topology and energy deposition in the Timepix3 sensor, and spectral gamma emissions from the imaged sample.

Results: The QPID can segregate alpha and positron emissions, generating separate quantitative alpha and positron autoradiograph images from a sample which contains both an alpha and positron emitting ligand. This provides a mechanism to analyze co-injected theranostic imaging and treatment ligands. Using these separate images, one can measure the uptake and distribution in the same tissue sample. Furthermore, one can directly measure the pharmacokinetic properties of both ligands and measure their bio-equivalence.

Conclusion: Measuring any bio-equivalent deviation of the two ligands can be used to adjust the imaging agent activity and pharmacokinetic properties to more accurately be used as a proxy to measure the treatment effects and dosimetry results of the treatment ligand.

Abstract

PET and SPECT are often used even when radionuclides emit complementary radiations. Gamma-positron imaging exploits nuclides that emit a prompt gamma with a positron: annihilation photons define a PET line of response and prompt-gamma Compton interactions define a cone. Their intersection can estimate the radionuclide decay location more directly than PET alone.

We are developing a brain-dedicated gamma-positron imager using crossed-strip CZT detectors, event-correlated radiation transport, and GPU-accelerated list-mode reconstruction. The system records PET coincidences, prompt-gamma Compton events, and triple events in which both signatures arise from one decay. We evaluated event-class-specific LM-MLEM, conventional joint-Poisson reconstruction, and an event-type-conditioned method designed to preserve the localization value of triple events.

UAMonteCarlo preserves the lineage of prompt photons, positrons, annihilation photons, and secondaries from emission through detector interaction, allowing detector response, event classification, sensitivity, and reconstruction to be evaluated in one framework. In simulated 44Sc studies, all event classes localized point sources correctly. In a hot-rod phantom, rod-to-background contrast was 1.36 for PET-only, 1.14 for prompt-only, 4.93 for triple-only, 3.24 for conventional combined, and 3.86 for event-type-conditioned combined reconstruction, showing that triple events are uncommon but spatially informative.

The project also applies complete radiation-transport histories to emerging theranostic imaging. For 149Tb, gamma-positron events from the positron branch could image the same radiopharmaceutical responsible for alpha therapy. We are also applying the framework to 203Pb surrogate imaging for 212Pb-targeted alpha therapy. Together, these studies show how preserving decay and particle histories can inform detector design, reconstruction, quantitative imaging, and dosimetry.

Abstract

Multiplexed PET (mPET) enables simultaneous imaging of multiple radiotracers in a single acquisition by exploiting radionuclide-specific signatures beyond conventional 511-keV annihilation photons. Positron–prompt gamma emitters such as ¹²⁴I, ⁸⁶Y, or ⁴⁴Sc can generate triple coincidences that help distinguish them from pure positron emitters. This opens new opportunities for theranostic imaging, where complementary tracers could simultaneously characterize therapeutic targets and tumor biology, investigate intra- and inter-lesion heterogeneity, assess radiopharmaceutical biodistribution and dosimetry, or monitor early biological response to treatment. 

We present a modular mPET framework applicable to different radionuclides and scanners. It incorporates all the corrections and reconstructions required for quantitative PET while adapting to different scanner capabilities. When available, we exploit energy and time-of-flight information, while coincidence multiplicity alone provides radionuclide-specific information in more limited systems. The method also includes radionuclide-specific positron-range correction for each separated image. 

We quantitatively evaluated the approach with combinations such as ¹⁸F+¹²⁴I, ¹⁸F+⁵²Mn, ⁶⁸Ga+⁸⁶Y, and ⁸⁹Zr+⁸⁶Y across several preclinical and clinical PET scanners. Studies ranging 
from simultaneous tracer imaging in mice to multi-radionuclide phantom experiments demonstrated recovery of the individual radionuclide distributions across systems with different detector characteristics.

These results demonstrate the translation of mPET across PET scanners and radionuclide combinations. mPET could extend PET from a predominantly single-tracer technique toward a multidimensional molecular imaging platform in which complementary aspects of patient selection, treatment delivery, and response are interrogated within the same imaging session.