Development of lesion and organ negative cast modelling technique for quality assurance and optimization of nuclear medicine images

Abstract

Background

Nuclear medicine imaging allows for a wide variety of data acquisition and image generation methods in the clinical setting. Imaging phantoms are routinely used to evaluate and optimize image quality and quantitative accuracy of features, but few phantoms realistically model the anatomy or heterogeneity of target regions within patient images, such as tumours that are commonly observed in oncology. We developed a negative cast modelling (NCM) technique which enables applications such as non-standard shape tumour phantoms, organ phantoms for radiation dosimetry, and quality control phantoms with small lesions.

Methods

Tumour templates were derived from segmented PET images of primary mediastinal B-cell lymphoma (PMBCL) patients. Lesion segmentations were saved and 3D-printed. Negatives were developed using silicone-based molding materials, and final models cast using a composition of liquid plastic, pigment, and PET radiotracer. Images of lesions were acquired using the GE DMI PET/CT scanner, and image features were quantified.

Results

Mean absolute error (MAE) for tumour volume between the original template and casted models is 13.8%, indicating that the method is reasonably accurate. The high viscosity of the liquid plastic used in the casting process establishes non-uniform tumour models, which is very useful in practice for evaluating image features related to heterogeneity. PET images using the NCM method is determined to be highly realistic by an experienced nuclear medicine physician, due to the non-standard shapes that can be established within the tumours.

Conclusions

The NCM method has potential to enable more realistic phantom studies within nuclear medicine imaging. The cost for the lymphoma tumour phantom study is less than $400 USD, making it feasible for large-scale studies.

Plain language summary

Radioactive substances are used to diagnose and treat disease. Special testing devices called phantoms are used to ensure that the scanners used to detect the radioactive substances work properly. These are usually simple shapes, such as spheres, which do not accurately represent the complexity of real human anatomy. In this study, we developed a technique that uses medical images from patients to create phantoms with more realistic shapes, such as irregular tumors, salivary glands, and very small lesions. These phantoms are made using a low-cost molding and casting process and can be reused. By improving how scanners are tested and optimized, this approach may lead to better diagnosis and treatment for conditions such as cancer.

Introduction

Nuclear medicine (NM) is a specialty that is used in more than 50 million medical procedures each year worldwide¹. NM involves the use of a radionuclide conjugated to a pharmaceutical (i.e., radiopharmaceutical) which can be used to obtain diagnostic information for a disease or provide therapy². Due to its versatility, NM is used to assess disease pathology in fields such as neurology, cardiology, and oncology, as well as infectious diseases and inflammatory disorders³. Positron emission tomography (PET) and single photon emission computed tomography (SPECT) are NM imaging techniques which visualize receptors or metabolic processes in vivo, allowing for three-dimensional visualization of the radiopharmaceutical distribution⁴.

PET and SPECT are critical diagnostic tools which have been shown to substantially influence clinical decision-making in oncology^{5, 6–7}. Image degradation occurs as the result of physical phenomena such as attenuation, scatter, and the partial volume effect^8,9. As further complication, images can be biased by respiration, cardiac motion, and random movements of the patient¹⁰. Due to broad application of PET and SPECT in a clinical setting, there is significant need to validate scanner performance for different malignancies, radiopharmaceuticals, and clinical tasks (e.g., detections vs. quantification).

In nuclear medicine, performance tests are required at installation, acceptance, and as part of a routine quality assurance (QA) programme^5,11. PET and SPECT performance tests are routinely performed using phantoms with precisely defined radiopharmaceutical concentrations^7,12,13. Such studies are also used to select optimal data acquisition and image generation protocols^14,15. Most commonly, these phantoms are prepared by injecting a radioisotope into fillable compartments to simulate distributions that are observed in patient scans^7,16,17. For instance, PET scanner performance is validated using the NEMA NU 2 methodology, in which tumours are modelled by injecting a radiotracer into hollow fillable spheres^7,16. These fillable spheres can create “cold-shells” surrounding the target source, which introduces image-biasing effects such as reducing measured concentration¹⁸ and increasing the observed volume of a target⁸.

Development of phantoms with three-dimensional (3D) printers represents the newest generation of phantom modelling^9,10,19,20. The radiopharmaceutical is directly mixed into the 3D printer resin, allowing phantoms of arbitrary size and shape to be manufactured. However, this requires modification of the device to print radioactive materials, which can be costly and a significant logistical challenge from the perspective of both radiation safety officers and physicists⁹. As such, there is a need to develop realistic phantoms for quality assurance and optimization of nuclear medicine imaging.

We define a “realistic” phantom in terms of being able to (i) cast non-standard shapes that better represent organs and tumours (i.e., not a sphere, cylinder, cube, or line source), (ii) have shell-less properties, and (iii) allow for re-usability. The requirement for non-standard shapes is important for clinical applications that are not adequately represented by the NEMA NU 2 phantom^7,16. The requirement for shell-less phantoms is important for reducing the “cold-shell” effect, which can create image-biasing effects that reduce the measured concentration¹⁸ and increase the observed volume of a target⁸. Moreover, we seek to develop phantoms that are re-usable and not readily dissolved in aqueous solutions. This property is critical for long half-life isotopes, such as sodium-22 (²²Na; T_1/2 = 2.6 y), which may be used to develop sources for periodic scanner quality control. We can identify three diverse tasks which may be achieved with our definition of a “realistic” phantom:

Task 1 is to develop non-standard shape tumour phantoms. Irregular tumour shapes (e.g., non-spherical) are frequently encountered in cancers with disseminated disease. Primary mediastinal B-cell lymphoma (PMBCL) is a rare form of lymphoma, which presents with bulky, non-standard shape tumours in the mediastinum^21,22. PMBCL tumours are composed of lymph node conglomerates, which present as lesions with heterogeneous radioactivity distribution^21,22. Accurate quantification of metabolic tumour volume (MTV) and total lesion glycolysis (TLG) can be used to better predict therapy response and overall survival in lymphoma patients^23,24. However, non-spherical geometries are not accounted for in conventional phantoms.

Task 2 is to develop organ phantoms for radiation dosimetry. Radiopharmaceutical therapy (RPT) has shown promise for treating patients with metastatic prostate cancer^25,26. Accurate quantification of organs can help predict treatment outcomes and adverse events such as xerostomia²⁵. In this task, we seek to create models of salivary and lacrimal glands (SLGs), which may be used to optimize imaging for better personalization of therapy. Physiologic uptake of salivary glands in PSMA PET images is considered to be heterogeneous²⁷.

Task 3 is to develop quality control phantoms with small lesions. Standardized phantoms do not currently model small lesions (lesions below 10 mm), relevant to tasks such as cardiac perfusion imaging²⁸, brain plaque measurements²⁹, and malignant tumour detection²⁹. In this task, we create quality control phantoms which may be used to monitor scanner performance over time. We use sodium-22 (²²Na), a positron emitter with 2.6-year half-life, to establish long-lasting sources with uniform radioactivity, which may be used for quality control protocols.

The purpose of this study was to develop and validate a method to produce realistic quality assurance phantoms for nuclear medicine. We develop negative cast modelling (NCM): a technique to manufacture low-cost, customized phantoms from a reverse injection molding process. NCM was used to create quality assurance phantoms for three tasks: tumour imaging in lymphoma, organ quantification for radiation dosimetry, and small-diameter sources for quality control of imaging systems. The phantoms were imaged with PET/CT scanners, though this technique could be easily applied to the SPECT modality or extended to a wide range of quality assurance protocols (e.g., detectability phantoms).

Using the NCM technique, we successfully fabricate low-cost phantoms with clinically relevant anatomical shapes and heterogeneous activity distributions. Volume measurements show reasonable agreement with the original models when quantified from PET images. The phantoms produce PET images that closely resemble patient data and replicate realistic uptake patterns in tumour models. Compared to conventional methods, NCM offers greater flexibility, reduced production costs and reusability, supporting its application in task-specific quality assurance and harmonization studies in nuclear medicine.

Methods

Delineation of template structures

To develop realistic geometries for the phantoms, regions-of-interest (ROIs) were defined from patient images obtained at our institution (Fig. 1). Target regions were identified and delineated by a trained nuclear medicine physician, using a gradient-based segmentation tool (PET Edge⁺, MIM Software Inc., USA). For the segmentation tool, boundaries were automatically defined by computing the second spatial derivative along concentration profiles (i.e., inflection points)^30,31. Segmented regions from each phantom template were saved in MIM using the RTStruct file format, and background regions were defined by moving the contour to adjacent anatomy. The radioactivity for each target and background ROI was measured, and median concentrations were used to select the injected activity for each compartment in the phantom. This methodology was applied to each of the following applications:

[See PDF for image]

Fig. 1

Overview of developed NCM technique.

a Segmented lesion from PET/CT image. b 3D printed tumour. c Negative of tumour casted from template. d Final tumour model for imaging, infused with visualization method (i.e., radiotracer).

Task 1: non-standard shape tumour phantoms

To address this aim, we considered a cohort of PMBCL patients imaged with ¹⁸F-FDG PET/CT scans. The largest tumour from each of five patients (2.7 ml–76.0 ml) was used to create tumour masks. Reference regions were positioned in soft-tissue of the mediastinum, directly adjacent to the tumour. Activity concentrations were determined from 22 tumours in 13 patient images. Target concentrations were 21.5 kBq/mL and 1.5 kBq/mL for the lesions and background, respectively. Patient imaging data used in this task were obtained from participants enroled in a retrospective study which was reviewed and approved by the University of British Columbia—BC Cancer Research Ethics Board (H19-0611). Since this was a retrospective study, the ethics committee of the hospital waived the requirement for obtaining informed consent from patients.

Task 2: organ phantoms for radiation dosimetry

A cohort of 40 metastatic prostate cancer (mPCa) patients that received a PSMA-targeting PET scan (¹⁸F-DCFPyL) were analyzed. Salivary glands (parotid, submandibular, sublingual, tubarial) and lacrimal glands from 4 patients were used to create template masks. Reference regions were positioned in the mandible. Activity concentrations were defined from all 40 patients, for each SLG and reference region. Target concentrations were 17.0 kBq/mL and 2.2 kBq/mL for the lesions and background, respectively. Patient imaging data used in this task were obtained from participants enroled in a clinical trial (NCT02899312) which was reviewed and approved by the University of British Columbia–BC Cancer Research Ethics Board. All participants provided written informed consent for the use of their imaging data for research purposes, including secondary analyses such as those conducted in the present study.

Task 3: quality control phantoms

To create realistic surrogates for PET performance tests (e.g., detectability phantoms), spheres ranging from 3 mm to 16 mm were defined. The volumes were selected based on an analysis of 37 lesions in 33 mPCa patients imaged with ¹⁸F-PSMA-targeting agents (median = 0.91 mL, quartile 1 = 0.61 mL, quartile 3 = 1.58 mL)³². Target concentrations were 57.6 kBq/mL and 1.5 kBq/mL for the lesions and background, respectively²⁸. Reference regions were positioned in soft-tissue anatomy adjacent to the lesions.

Template printing and technical considerations

A pipeline was built to create phantoms from the target geometries (see Delineation of Template Structures). A flow diagram illustrating the complete pipeline is shown in Fig. 2, indicating the material and/or data format used at each stage. To define template structures from each target region (Tasks 1–3), segmentations were exported from MIM as stereolithography (STL) files. STL files represent the surface of each target region, without any reference to the internal texture or colour, both of which are unnecessary within this pipeline. This file format was selected as it is commonly used in 3D printing software³³. As such, STL files are easily input and processed with a number of design programmes, such as Meshmixer³⁴, Blender³⁵, and Meshlab³⁶. For instance, surface smoothing may be performed to avoid irregular boundaries introduced by the resolution-limit of the imaging device.

[See PDF for image]

Fig. 2

Flowchart of manufacturing process.

Steps for target delineation, template printing, negative casting, final casting, and phantom measurements.

The STL files were imported to the 3D-printer interface (uPrint SE Plus, Stratysys Ltd., USA), and a semi-solid print setting was selected for each internal structure. Templates were printed by fused deposition modelling (FDM), which feeds a polymer filament through a heated nozzle (Fig. 1b)³⁷. The polymer is deposited in 2D layers on the build platform, thereby achieving a 0.254 mm isotropic print resolution³⁸. The molds were printed with acrylonitrile butadiene styrene (ABS)^39,40 and the soluble support material was subsequently re-dissolved in a detergent solution. Aluminium rods were fixed to the contours from Task 1, to establish injection holes for the reverse injection molding process (see Reverse Injection Molding Process).

Reverse injection molding process

Reverse molds of the phantom (i.e., negatives) were created using a tin cure silicone rubber (product number 82241, Smooth-On Ltd., USA), as shown in Fig. 1c. This material was selected as it offers a relatively short cure time (75 min), while allowing multiple molds to be prepared during its 15-min pot life⁴¹. Moreover, the hardened compound has a tensile strength of 1.66 MPa (240 psi)⁴¹, preventing breaks and thereby making the removal of templates as streamlined as possible (see Phantom Casting and Production). The negative mold was designed effectively in two halves. A schematic for the reverse injection process is shown in Fig. 3. Each 3D-printed object was positioned in a cylindrical container, with the aluminium rod oriented upwards (Fig. 3a; container not shown). The silicone compound was injected into the container until the phantom templates were partially submerged (Fig. 3b), and the molding material was allowed to cure for 2 h. To facilitate removal, an aerosol release agent was applied to the exposed surface of the mold (product number 70020 A, Smooth-On Ltd., USA). Next, the remaining silicone was injected into the container until the printed object was completely submerged (Fig. 3c). The release agent minimized adhesion between the two layers, thereby allowing the mold to separate into two halves. After curing, the printed template was removed to form a negative of the phantom (Fig. 3d).

[See PDF for image]

Fig. 3

Schematic of negative casting process.

a 3D printed tumour model. b Tumour model partially submerged with silicone material. c Tumour model fully submerged with silicone material. d Tumour model cavity in hardened silicone mold. e Silicone mold cavity filled with radioactive liquid plastic. f Final radioactive tumour model.

Phantom casting and production

A semi-rigid casting resin (product number 64921, Smooth-On Ltd., USA) was used to create the final tumour models (Fig. 3e). This material was selected as it allows for sufficient mixing time (pot life = 15 min), but cures relatively fast to minimize decay of the positron emitter (cure time = 75 min). The liquid plastic (200 mL) was combined with 5 mL of an aqueous solution containing the radiopharmaceutical. ¹⁸F was used for the tumour and organ phantoms, as described in Tasks 1 and 2, respectively. ¹⁸F has a half-life of 109.77 min, and as such, it is not practical to use a standard source in quality control phantoms (Task 3). Instead, we used ²²Na (T_1/2 = 2.6 y) as a long-lasting surrogate for ¹⁸F. The ²²Na and ¹⁸F have similar positron energies (220.3 keV vs. 252 keV), resulting in ranges of 0.53 mm and 0.6 mm in soft-tissue^42,43. For ¹⁸F, the complete ¹⁸F-FDG molecule was selected rather than ¹⁸F, as the chemical properties of ¹⁸F make it adhesive which is not desirable from a radiation safety perspective. The ¹⁸F activity concentration was selected based on the median concentration of segmented ROIs from the patient analysis.

To visualize mixing of the radiopharmaceutical, 1 mL of a liquid plastic-compatible fluorescent red pigment was added to the prepared aliquot (product number 90411, Smooth-On Ltd., USA). As a secondary purpose, the pigment also improves visibility for spills and/or contamination. The mixture was stirred for 20 seconds using a disposable mixing rod, and a Luer lock syringe was used to dispense the mixture into each mold. Prior to injecting the mixture, the mold was coated with the aerosol release agent. To ensure efficient dispensing, the liquid plastic was dispensed using a 20 mL Luer lock syringe with a 20 Gauge stainless steel extension. While casting smaller phantoms (e.g., QC spheres; Task 3), a 16 Gauge needle was used to improve precision at the entry point. To minimize contamination, absorbent pads were covered on all surfaces and multiple layers of gloves were worn. To reduce radiation exposure, the dispensing syringe was covered by a lead syringe shield. The mixture was allowed to harden for 2 h. Adhering to “as low as reasonably achievable” (ALARA) principles⁴⁴, tweezers were used to maximize hand distance while removing radioactive sources from each mold (Fig. 1d). The above steps were performed in a shielded fume hood.

PET/CT imaging and analysis

The phantoms were inserted into the Probe-IQ shell (Radiology Support Devices, Inc., USA) and pelvis phantom (Data Spectrum Corp., USA)^{45, 46–47}, as described in further detail by Fedrigo et al. ³² The thorax included a fillable liver and lungs in order to simulate a 92 kg patient⁴⁸. The pelvis contained a 440-mL compartment and clinical-grade tubing to simulate the bladder and ureters, respectively. Polyurethane filter foam^45,49 was used to mount the NCM sources within Probe-IQ. Two scans were performed using a GE Discovery MI PET/CT scanner (General Electric, USA):

No activity in the background

This acquisition was used to obtain the ground truth. To accurately determine the radioactivity of each casted phantom, a long-duration (10 min/bed), background-free acquisition was performed using a Discovery PET/CT scanner (5-ring DMI model for Tasks 1 and 2, D690 model for Task 3). Since the background compartment did not contain activity in this scan, an expanded ROI was used to capture the total activity that was emitted by each source. The volume of the template was defined via the stereolithography file. To obtain a ground truth for the casted phantom volume, the mass was measured with an analytical balance and converted to volume using the known density of the casting resin (ρ = 1.15 g/mL). The volume was further defined with two imaging modalities: (i) measuring the volume from PET images, and (ii) measuring the volume from computed tomography images, which were segmented either by $\ge$−25 Hounsfield Unit (HU) thresholding or 5% of maximum thresholding. Two-sided paired t tests were used to compare the volume obtained with imaging and the original template volume.

With activity in the background

This acquisition was used to simulate a patient scan.

Task 1: non-standard shape tumour phantoms

For added realism, the Probe-IQ liver and background compartments were injected with ¹⁸F to achieve target concentrations determined from the clinical analysis of PMBCL PET images. One frame (30 min/bed) was acquired in list-mode.

Task 2: organ phantoms for radiation dosimetry

The NCM models were inserted into the Probe-IQ pelvis cylinder (without any inserts). The main compartment was injected with ¹⁸F to simulate realistic background radioactivity concentrations for PSMA PET. Five frames (3 min/bed) were acquired in list-mode.

Task 3: quality control phantoms with small lesions

The NCM models were inserted into the Probe-IQ thorax and pelvis. The liver and background compartments were injected with ¹⁸F to achieve target concentrations determined from a patient analysis of ¹⁸F-PSMA PET images³². The Probe-IQ pelvis, composed of a bladder, ureters, and background compartment, were also injected with ¹⁸F. A fifteen-frame dynamic acquisition was performed, adjusting the scan duration to obtain similar counting statistics from ¹⁸F in each frame. A 2.5 min frame was acquired when the background concentration approximated soft-tissue concentrations observed in PSMA scans (1.8 kBq/mL)³².

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Results

Development of phantoms for quality assurance

Using the negative cast modelling (NCM) technique, we manufactured phantoms for three quality assurance tasks in nuclear medicine. The lymphoma phantoms (Task 1) were segmented, printed, and cast in about 40 h, or one work week, by a student trainee in medical physics (Fig. 4; top). Production cost was about $400 USD, with material descriptions and production quantities shown in Table 1. Reverse injection molding was the most time- and cost-intensive step, with the silicone casting material accounting for 50% of the cost for the entire experiment. However, due to the resilience and high tear strength of the selected material (7001 N/m)⁴¹, zero instances of mold breakage occurred during the removal process. Since the silicone molds can be re-used, the cost of subsequent experiments are reduced by 50 percent. The remaining costs were attributed to the ABS 3D printing material ($100 USD), casting resin ($30 USD), fluorescent dye ($25 USD), mold release agent ($15 USD), and supplies such as syringes, utensils, and mixing containers ($25 USD).

[See PDF for image]

Fig. 4

Applications of non-standard tumour phantoms, organ phantoms for radiation dosimetry, and quality control phantoms with small lesions.

Template phantoms (left) and radioactive models (right) for a lymphoma tumours, b salivary and lacrimal glands, and c quality control (QC) sources.

Table 1. Materials associated with the negative cast modelling (NCM) technique. Cost is listed in US dollars

Application	Procedural step	Product quantity	Cost
		[mL]	[US dollars]
Phantom template	3D printing	400	$100
Negative casting	Silicone	7600	$200
Model casting	Liquid plastic	748	$30
Visualization of mixing process	Fluorescent dye	90	$25
Spray used to release casting compounds	General release agent	414	$15
Miscellaneous	Utensils, containers, syringes	n/a	$25

Table 2 shows the volume difference of each volume measurement as compared to the original template contour. In the lymphoma tumour models, percent difference ranged from 3.6% (20.74 mL tumour) to 37.3% (15.39 mL tumour). The PET mean absolute error (MAE) was 13.84% (P = 0.10). The volume of each phantom was also measured with computed tomography (CT) imaging (Table 2, column 3–4). The percent difference between −25 HU thresholding and template volume ranged from 40.7% to 57.5% (P = 0.12). In comparison, the percent difference for 5% of maximum thresholding ranged from 49.6% to 68.3% (P = 0.12). Overall, determining the phantom volume with an analytical balance appears to be more reliable than CT, as it had lower percent difference and error measurements.

Table 2. Template and measured volume of lymphoma tumour phantoms, as compared to imaging measurements with positron emission tomography (PET) and computed tomography (CT)

Template volume [mL]	Measured volume [mL]	PET percent difference [%]	CT percent difference [%]
			−25 HU threshold	5% threshold
2.68	2.96	10.6	57.5	68.3
7.56	8.36	10.6	40.7	49.6
15.39	21.13	37.3	43.2	57.8
20.74	21.49	3.6	56.0	62.5
75.99	81.39	7.1	43.5	51.9

The template volume is quantified by summing voxel volumes from the stereolithography file. The percent difference of each volume measurement is compared to the measured volume, as defined with an analytical balance.

Nuclear medicine procedures and applications

Figure 5 shows PET slices of NCM phantoms scanned with radioactivity present in the background, as compared to patient images. The models in Task 1 (Fig. 5; top-right) were compared to ¹⁸F-FDG PET images of patients with lymphoma tumours, while Task 2 mimicked salivary and lacrimal glands observed in patients imaged with PSMA-targeting radiopharmaceuticals (Fig. 5; middle). Lastly, the spheres in Fig. 5 (bottom-right) were compared to the focal, high-uptake metastases observed in PSMA PET images of patients with prostate cancer. The casted spheres shown in Fig. 5 have diameters of 16 mm and 12 mm, respectively (sphere concentration = 57.6 kBq/mL, background concentration = 1.5 kBq/mL). Qualitatively, the bulky tumours and salivary glands (Task 1 and 2, respectively) had high heterogeneity, while the smaller QC spheres appeared more uniform and exhibited higher contrast. As shown in the transaxial PET slices, the high viscosity of the liquid plastic (800 cps) appeared to reduce the diffusion rate of the radiopharmaceutical, resulting in heterogeneous ¹⁸F texture in the tumour models. The lymphoma tumours had multiple patches of differing activity concentration, which helped to mimic the heterogeneous pathology observed in lymphoma patients (Fig. 5; top-left)^50,51.

[See PDF for image]

Fig. 5

Visual comparison of positron emission tomography (PET) images.

Human subjects and NCM phantoms compared (left and right, respectively) for a bulky lymphoma tumours, b salivary glands, and c small-diameter prostate cancer metastases. The prostate cancer metastases have diameters of 16 mm and 12 mm, respectively (sphere concentration = 57.6 kBq/mL, background concentration = 1.5 kBq/mL).

Discussion

In nuclear medicine, quality assurance tests are required to validate scanner performance for different modalities, radiopharmaceuticals, and clinical tasks. Phantom studies are also used to optimize data acquisition and image generation protocols. However, there remains a lack of phantoms to represent the diverse range of patient physiologies observed in clinical practice. In this study, we developed a technique to produce customized phantoms in nuclear medicine imaging. Using this negative cast modelling (NCM) technique, arbitrary phantom geometries can be generated from existing targets identified in patient images (Fig. 1). Given a template model, the target geometry can be converted to a stereolithography file format that is readable with a wide range of CAD software programs (Fig. 1a, b). From this, a silicone compound is used to create an inverse injection mold for the desired phantom geometry (Fig. 1c). By mixing a liquid plastic resin with the visualization agent (i.e., radiotracer), radioactive phantoms of arbitrary size and shape can be established, as shown in Tasks 1–3. Although this methodology was developed using ¹⁸F, the same techniques may be adapted to short-lived PET tracers, such as ⁶⁸Ga (T_1/2 = 68 min), as well as gamma-emitters relevant to SPECT imaging (¹²³I; T_1/2 = 13.2 h). By mixing the resin with a long half-life positron emitter, the QC sources can be used over a period of many months or years (Task 3). The overall objective of this study was to establish an inexpensive approach to manufacture phantoms with non-standard shape contours. By controlling the resin mixing time, the NCM technique provides the ability to cast structures with uniform (Task 3) or heterogeneous (Task 1, Task 2) radioactivity distribution.

The NCM technique provides key improvements over conventional phantoms, which rely on the use of fillable spheres to simulate lesions (injectable spheres, Table 3)^7,16,52,53. As the plastic walls displace background activity, “cold shell” artifacts are created, thereby reducing the measured concentration¹⁸ and increasing the observed volume of a target region⁸. Simple casting methods have been used to establish target regions without cold shells (casted spheres, Table 3)^18,47,49. However, developing these molds often requires access to specialized machining equipment (e.g., CNC lathe and milling machine). Meanwhile, resin-infused printing methods are effective for simulating complex target regions, but requires a dedicated 3D-printer and substantial radiation safety considerations (3D printing, Table 3)^9,10,19,20.

Table 3. Comparison of tumour modelling capabilities between standardized phantom (NEMA IQ), spherical tumour models, 3D-printed tumours, and the NCM technique

Methodology characteristics	Injectable spheres7,16,52,53	Casted spheres18,47,49	3D Printing9,10,19,20	NCM Technique
Time-efficient	✓	✓	✓	✓
Mitigates cold-shell effect	✕	✓	✓	✓
Simulates complex target regions	✕	✕	✓	✓
Cost-effective	✕	✕	✕	✓

In this work, we showed that the NCM technique can be used to facilitate a variety of clinical applications including realistic modelling of tumours, organs, and QC sources for applications in nuclear medicine. By defining models from 3D-printed templates, this phantom design and manufacturing method provides a technical means to create irregular shaped phantoms and small lesion sizes below 10 mm. In the context of radiopharmaceutical therapies, irregular phantoms may be used to validate measurements of activity in organs-at-risk, an important step in the dosimetry workflow⁵⁴. On the other hand, irregular phantoms can also be used to validate measurements of tumour volume, an important prognostic factor of survival in patients with lymphoma^23,55. As such, this manufacturing technique supports the development of phantoms that can be used to validate clinically-relevant measurements of activity and volume in nuclear medicine images.

For this initial study, we used a casting compound that cures within 1–2 h. However, depending on the phantom study, compounds with shorter cure times may be desirable depending on the study parameters (e.g., filling time, total activity injected in phantom). To secure the NCM models within the Probe-IQ shell, polyurethane filter foam was cut to mount sources with irregular sizes and shapes. This method was selected as it allowed us to use phantom shells that were already available at our institution, thereby eliminating the need to design or purchase a new phantom. By scanning the NCM models for a long duration in a background-less phantom, the ground truth radioactivity was defined using expanded contours. In the future, this can potentially allow for definition of more complex imaging features, which rely on measuring spatial distributions of the radiopharmaceutical (e.g., tumour shape or texture)⁵⁶.

Overall, the NCM protocol was relatively inexpensive ($<$$400 USD), with one half of the cost required to manufacture customized molds which may be re-used in subsequent experiments (Table 1). The NCM approach utilizes silicone molds, which have an estimated shrinkage of 1% (i.e., 0.1 mm tolerance for a 10 mm sphere)⁵⁷. The reusability of the silicone mold depends on its storage conditions, as well as the type of casting material and application of a release agent⁵⁸. It is estimated that the silicone molds may be reused 10–30 times with a polyurethane casting resin⁵⁸. To increase phantom lifespan and reduce manufacturing costs, isotopes with long half-life may be implemented, such as germanium-68 (T_1/2 = 271 d) or sodium-22 (T_1/2 = 2.6 y). This may be particularly beneficial for clinical trials which need to implement non-conventional phantoms in repeated QC protocols. As a benchmark, the standardized NEMA NU 2 phantom is marketed for $3400 USD⁵⁹, an 8.5x greater upfront cost than the NCM method. The NEMA phantom is manufactured from an optically transparent acrylic, with an estimated build tolerance of $\le$0.127 mm⁶⁰. As such, the NEMA phantom and NCM method are expected to have a comparable tolerance for 10 mm spheres.

We also recognize some limitations in our study. The casting resin has a viscosity of 800 cps, which requires 1–2 min of continuous mixing to ensure a uniform radioactivity distribution. As the total activity within the syringe can be measured, this allows for ground truth determination of first-order metrics (volume, total activity, mean activity). In certain applications, a heterogeneous radioactivity may be desired to more accurately reflect imaging conditions, such as lesions observed in PMBCL. Heterogeneous radioactivity distributions were achieved by reducing the mixing time to 30 s, as shown in Fig. 5 (Task 2). However, establishing a ground truth for second- and third-order radiomics features, such as texture, is not within the scope of this work. In a previous study, the ground truth was estimated using a long-duration (1-h) scan with a high-resolution preclinical PET scanner⁶¹, but this was not the aim of the current work. Additionally, the viscosity of the casting resin makes it difficult to cast anatomical structures which have thin appendages, such as tubarial glands. This may be overcome by heating the resin during mixing, as performed by Kao et al. ⁶² Future studies may investigate alternative materials such as low-viscosity silica⁶³ or gelatin^64,65. However, a low-viscosity resin may come at the trade-off of having increased spill risk. As such, proper radiation safety precautions are needed to mitigate the risk of contamination.

The NCM method utilizes a casting material intended to approximate the density of soft tissue ($\rho =$1.15 g/cc), in order to establish realistic attenuation encountered in PET imaging. As the urethane casting material has a similar density as water, CT scans of the phantom produce images with low contrast relative to the water-filled background. As shown in Table 2, this results in poor quantification of phantom volume measurements in CT images. As a result, the NCM method should not be applied to CT image validation with usage of its current materials. Future studies may investigate alternative materials, such as incorporating silica gel⁶⁵ or a gadolinium-based contrast agent⁶⁶. However, care should be taken when introducing alternative resins or contrast agents, as this can influence the PET attenuation correction.

The NCM protocol required 1–2 h of hardening time, and due to the half-life of the radiopharmaceutical, the operator received greater radiation exposure than in standardized phantoms³². While preparing the phantom with ¹⁸F, the operator received a radiation dose of 24 $\mu$Sv to the torso. As reference, the typical dose to the torso during our NEMA phantom experiments is approximately 9 $\mu$Sv. As such, the NCM procedure required additional care and handling as compared to conventional filling procedures. Before casting the NCM models, multiple “cold” runs (i.e., without radioactivity) were performed to ensure that radioactive contamination was minimized. Disposable materials, such as syringe tips, were carefully transported to sharps containers, and absorbent pads were used to prevent the spread of contamination. To ensure that the NCM models were safe to use, wipe tests were periodically performed on phantom surfaces and measured for a long duration using a gamma camera. Due to the long half-life of the ²²Na (T_1/2 = 2.6 y), the sources will be stored behind lead shielding for at least 10 half-lives (~26 years). Injection techniques and radiation safety procedures are described in further detail by Fedrigo et al. ³²

Due to the intensive nature of phantom preparation, the NCM method is not a practical approach to model individual patients within clinical practice. This is further complicated by the fact that treatments can rapidly change characteristics of the source tissue⁶⁷, particularly in applications such as theranostics⁶⁸. As such, computational phantoms likely remain the most feasible approach. For instance, digital twins have been proposed as a conceptual framework to adapt management plans in nuclear medicine, using in silico representations of patients to predict imaging conditions and treatment outcomes^69,70. At the same time, we would like to recommend techniques to streamline the NCM process for routine clinical practice. To bypass the mold production steps in the NCM process, a set of standardized polyurethane molds may be produced centrally and distributed by partners. This approach would greatly reduce the time investment for individual hospitals, thereby increasing the feasibility of applying NCM to clinical practice. Additionally, isotope analogs with longer half-life may be used, allowing hospitals to replace phantoms at defined intervals. For instance, the half-life of ²²Na (T_1/2 = 2.6 y) may allow the source lifetime to be extended to as long as 2–3 years, as currently implemented within a prostate cancer harmonization study⁶¹. In this work, we showed that NCM can be used to manufacture a wide range of phantoms for quality assurance and optimization in nuclear medicine. In next steps, surface interpolation will be used to avoid the irregular edges introduced by the resolution-limit of the PET scanner. Future work may investigate different applications for the technique, such as the use of gamma-emitting radiopharmaceuticals in SPECT imaging. While the NCM methodology provides the advantage of manufacturing custom phantoms in-house, we do not envision that it will be used to replace established quality assurance techniques. As an example, the NEMA phantoms already provide a standardized approach to obtain metrics such as uniformity, spatial resolution, and contrast¹⁶. As the NCM methodology does not require a machine shop for production, they may help establish a low-barrier approach to develop phantoms for novel clinical applications. As an example, the NCM method was recently used to cast ²²Na spheres for the Q3P phantom, as part of a multi-centre harmonization study for ¹⁸F-PSMA PET lesion quantification. Additionally, the NCM phantoms are being used to automate segmentation of salivary glands in ¹⁸F-PSMA PET images. Ideally, this will lead to better quantification of organs prior to radiopharmaceutical therapy, thereby leading to improved dosimetry and clinical management of patients with metastatic prostate cancer.

Conclusions

Nuclear medicine scanners are validated using phantoms with precisely defined radiopharmaceutical concentrations, and can also be used to optimize imaging protocols. However, standardized phantoms do not represent the diverse range of patient physiologies observed in clinical practice. In this work, we developed negative cast modelling (NCM); a cost-efficient technique to manufacture customized phantoms for applications in nuclear medicine. First, contours are segmented from patient images and files are exported to a 3D printer. The models are printed, thereby establishing templates to create negatives from a reverse injection molding process. Finally, the phantom is cast by injecting a radiopharmaceutical-infused liquid plastic into the mold, thereby creating contours of any desired size or shape. In the future, this technique will be used to manufacture quality control as well as protocol optimization and harmonization sources for clinical trials, ideally leading to better performance, reproducibility, and comparison of scanners in nuclear medicine.

Acknowledgements

This project was in part supported by Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant RGPIN-2019-06467 and the Canadian Institutes of Health Research (CIHR) Project Grant PJT-173231. The authors gratefully acknowledge Jugoslav Kitanovic and Scott Young from the BC Cancer Medical Physics Machine Shop, and Nicolas Fedrigo from the University of British Columbia, for their support during the development process.

Author contribution

R.F. was responsible for development and refinement of the technique, phantom preparation, scanning, and performing data analysis and interpretation. R.C. provided design insights for the negative casting process, selection of materials, and drilling holes to insert the stabilization rods. G.C. provided segmentations for salivary glands in PSMA patient scans. I.B. and C.G. provided segmentations for lymphoma tumours in ¹⁸F-FDG patient scans. A.R. helped with project design and interpreting results. C.U. helped with project design, phantom preparation, scanning, and interpreting results. All authors contributed to the drafting of the manuscript, and all authors read and approved the final manuscript.

Peer review

Peer review information

: Communications Medicine thanks Vanessa Nadig, Jazmin Schwartz and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Data availability

The datasets used in this study consist of phantom PET/CT images generated for system calibration and method validation. While the data do not involve human subjects, they are maintained by the corresponding author to ensure appropriate use and interpretation. As such, the datasets are not publicly available at the time of publication but can be shared upon reasonable request for research purposes.

Competing interests

The authors declare no competing interests.

Ethics

This work depicts compounds pertaining to patent WO 2017/117687 A1, which entitles F.B. to royalties upon licensing.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s43856-025-01009-z.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. World Nuclear Association. Radioisotopes in Medicine. January 10, Accessed May 29, 2025. https://world-nuclear.org/information-library/non-power-nuclear-applications/radioisotopes-research/radioisotopes-in-medicine (2025).

2. Sgouros, G. Radiopharmaceutical therapy. Health Phys.; 2019; 116, pp. 175-178.1:CAS:528:DC%2BC1cXisF2iurjO [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/30585960][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6310043]

3. Weber, WA et al. The future of nuclear medicine, molecular imaging, and theranostics. J. Nucl. Med.; 2020; 61, pp. 263S-272S. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33293447]

4. Walker M. Quantitative Dynamic 3D PET Scanning of the Body and Brain Using LSO Tomographs. University of Manchester; 2009.

5. Zanzonico, P. Routine quality control of clinical nuclear medicine instrumentation: A brief review. J. Nucl. Med.; 2008; 49, pp. 1114-1131. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/18587088]

6. Busemann Sokole, E et al. Routine quality control recommendations for nuclear medicine instrumentation. Eur. J. Nucl. Med Mol. Imaging; 2010; 37, pp. 662-671. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/20130859]

7. Boellaard, R et al. Quality control for quantitative multicenter whole-body PET/MR studies: A NEMA image quality phantom study with three current PET/MR systems. Med Phys.; 2015; 42, pp. 5961-5969. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26429271]

8. Hofheinz, F; Dittrich, S; Pötzsch, C; Hoff, J vanden. Effects of cold sphere walls in PET phantom measurements on the volume reproducing threshold. Phys. Med Biol.; 2010; 55, pp. 1099-1113.1:STN:280:DC%2BC3c%2Fns1GqsA%3D%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/20107246]

9. Gear, J. I. et al. Radioactive 3D printing for the production of molecular imaging phantoms. Phys. Med. Biol. 65, https://doi.org/10.1088/1361-6560/aba40e (2020).

10. Bieniosek, MF; Lee, BJ; Levin, CS. Technical Note: Characterization of custom 3D printed multimodality imaging phantoms. Med. Phys.; 2015; 42, pp. 5913-5918. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26429265][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4575317]

11. Busemann Sokole, E; Płachcínska, A; Britten, A. Acceptance testing for nuclear medicine instrumentation. Eur. J. Nucl. Med. Mol. Imaging; 2010; 37, pp. 672-681. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/20135120]

12. National Electrical Manufacturers Association. Performance Measurements of Positron Emission Tomographs (PET).; 2018.

13. Boellaard, R et al. FDG PET/CT: EANM procedure guidelines for tumour imaging: version 2.0. Eur. J. Nucl. Med Mol. Imaging; 2015; 42, pp. 328-354.1:CAS:528:DC%2BC2cXitVWhu7rL [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/25452219]

14. Kupitz, D et al. Optimization of SPECT/CT imaging protocols for quantitative and qualitative 99mTc SPECT. EJNMMI Phys.; 2021; 8, [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34328565][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8324619]57.

15. Fragoso Costa, P et al. Phantom-based acquisition time and image reconstruction parameter optimisation for oncologic FDG PET/CT examinations using a digital system. BMC Cancer; 2022; 22, 1:CAS:528:DC%2BB38Xit1ClsL3N [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35978274][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9387080]899.

16. NEMA Standards Publication NU 2-2012. NEMA NU 2: Performance measurements of positron emission tomographs. National Electrical Manufacturers Association. Published online:11 (2013).

17. Nichols, KJ; DiFilippo, FP; Palestro, CJ. Texture analysis for automated evaluation of Jaszczak phantom SPECT system tests. Med Phys.; 2019; 46, pp. 262-272. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/30418674]

18. Turkington, TG; DeGrado, TR; Sampson, WH. Small spheres for lesion detection phantoms. IEEE Nucl. Sci. Symp. Med. Imaging Conf.; 2002; 4, pp. 2234-2237.

19. Gillett, D et al. 3D printing 18F radioactive phantoms for PET imaging. EJNMMI Phys.; 2021; 8, [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33909154][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8081805]38.

20. Wollenweber, SD; Alessio, AM; Kinahan, PE. A phantom design for assessment of detectability in PET imaging. Med Phys.; 2016; 43, pp. 5051-5062. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/27587035]

21. Lucey, BC; Stuhlfaut, JW; Soto, JA. Mesenteric lymph nodes seen at imaging: Causes and significance. Radiographics; 2005; 25, pp. 351-365. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/15798054]

22. Zhou H. et al. Prognostic factors, therapeutic approaches, and distinct immunobiologic features in patients with primary mediastinal large B-cell lymphoma on long-term follow-up. Blood Cancer J. 10, https://doi.org/10.1038/s41408-020-0312-7 (2020).

23. Boellaard, R et al. International benchmark for total metabolic tumor volume measurement in baseline 18F-FDG PET/CT of lymphoma patients: A milestone toward clinical implementation. J. Nucl. Med.; 2024; 65, pp. 1343-1348.1:CAS:528:DC%2BB2cXis1Gqsb%2FF [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/39089812][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11372260]

24. Ceriani, L et al. PET-based risk stratification in primary mediastinal B-cell lymphoma: A comparative analysis of different segmentation methods in the IELSG37 trial patient cohort. J. Nucl. Med.; 2025; 66, pp. 209-214.1:CAS:528:DC%2BB2MXntVaku7k%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/39819690]

25. Hofman, MS et al. 177 Lu]-PSMA-617 radionuclide treatment in patients with metastatic castration-resistant prostate cancer (LuPSMA trial): a single-centre, single-arm, phase 2 study. Lancet Oncol.; 2018; 19, pp. 825-833.1:CAS:528:DC%2BC1cXptFCltL4%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/29752180]

26. Sgouros, G; Bodei, L; McDevitt, MR; Nedrow, JR. Radiopharmaceutical therapy in cancer: clinical advances and challenges. Nat. Rev. Drug Discov.; 2020; 19, pp. 589-608.1:CAS:528:DC%2BB3cXhsVygtLrF [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32728208][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7390460]

27. Sample, CM et al. Heterogeneous PSMA ligand uptake inside parotid glands. Phys. Med.; 2024; 121, 103366. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38657425]

28. Green, S; Grice, J. Technical note: 3D-printed phantom for dedicated cardiac protocols and geometries in nuclear medicine. Med Phys.; 2022; 49, pp. 943-951. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34910308]

29. Rowe, CC; Villemagne, VL. Brain amyloid imaging. J. Nucl. Med Technol.; 2013; 41, pp. 11-18. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/23396994]

30. Werner-Wasik, M et al. What is the best way to contour lung tumors on PET scans? Multiobserver validation of a gradient-based method using a NSCLC digital PET phantom. Int J. Radiat. Oncol. Biol. Phys.; 2012; 82, pp. 1164-1171. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/21531085]

31. Sridhar, P et al. FDG PET metabolic tumor volume segmentation and pathologic volume of primary human solid tumors. Am. J. Roentgenol.; 2014; 202, pp. 1114-1119.

32. Fedrigo, R et al. Quantitative evaluation of PSMA PET imaging using a realistic anthropomorphic phantom and shell-less radioactive epoxy lesions. EJNMMI Phys.; 2022; 9, [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35032234][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8761183]2.

33. Kai, CC; Jacob, GGK; Mei, T. Interface between CAD and rapid prototyping systems. Part 1: A study of existing interfaces. Int. J. Adv. Manuf. Technol.; 1997; 13, pp. 566-570.

34. Schmidt R., Singh K. Meshmixer: An interface for rapid mesh composition. ACM SIGGRAPH 2010 Talks, SIGGRAPH ’10. Published online 2006, https://doi.org/10.1145/1837026.1837034 (2010).

35. Markova, KT; Dovramadjiev, TA; Jecheva, GV. Computer parametric designing in Blender software for creating 3D paper models. Annu. J. Tech. Univ. Varna, Bulgaria; 2017; 1, pp. 77-84.

36. Cignoni, P. et al. MeshLab: An open-source mesh processing tool. 6th Eurographics Italian Chapter Conference 2008 - Proceedings. Published online:129-136 (2008).

37. Mazzanti, V; Malagutti, L; Mollica, F. FDM 3D printing of polymers containing natural fillers: A review of their mechanical properties. Polym. (Basel); 2019; 11, 1094.

38. Hong, AL et al. Feasibility of fabricating personalized 3D-printed bone grafts guided by high-resolution imaging. Med. Imaging 2017: Imaging Inform. Healthc., Res., Appl.; 2017; 10138, 101380O.

39. Osman, MA; Atia, MRA. Investigation of ABS-rice straw composite feedstock filament for FDM. Rapid Prototyp. J.; 2018; 24, pp. 1067-1075.

40. Harris, BD; Nilsson, S; Poole, CM. A feasibility study for using ABS plastic and a low-cost 3D printer for patient-specific brachytherapy mould design. Australas. Phys. Eng. Sci. Med; 2015; 38, pp. 399-412. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26108891]

41. Smooth-On. OOMOO Series Product Overview. 2023. Accessed April 5, 2023. chrome-extension://efaidnbmnnnibpcajpcglclefindmkaj, https://www.smooth-on.com/tb/files/OOMOO_25_30_TB.pdf

42. Jodal, L; Le Loirec, C; Champion, C. Positron range in PET imaging: An alternative approach for assessing and correcting the blurring. Phys. Med Biol.; 2012; 57, pp. 3931-3943.1:STN:280:DC%2BC38nntFWjtA%3D%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/22643300]

43. Jødal, L; Le Loirec, C; Champion, C. Positron range in PET imaging: Non-conventional isotopes. Phys. Med Biol.; 2014; 59, pp. 7419-7434. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/25386999]

44. Agency IAE. Regulations for the Safe Transport of Radioactive Material. IAEA Safety Standards for protecting people and the environment. 2018;6.

45. Fedrigo, R. et al. Optimization of Quantitative 18F-DCFPyLPET using a Realistic Anthropomorphic Phantom with Shell-less Radioactive Epoxy Lesions. In: EANM Annual Meeting. European Association of Nuclear Medicine; 2020.

46. Fedrigo, R; Kadrmas, D; Edem, P; Benard, F; Rahmim, A; Uribe, C. Phantom-guided optimization of BSREM reconstruction parameters using shell-less radioactive epoxy lesions for [18F]F-DCFPyL prostate cancer imaging. Med. Phys.; 2020; 47, pp. e255-e2720.

47. Morey, AM; Noo, F; Kadrmas, DJ. Effect of using 2 mm voxels on observer performance for PET lesion detection. IEEE Trans. Nucl. Sci.; 2016; 63, pp. 1359-1366.1:CAS:528:DC%2BC1cXkvFCjt70%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/27499550][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4970864]

48. Morey, A; Kadrmas, D. Effect of varying number of OSEM subsets on PET lesion detectability. J. Nucl. Med Technol.; 2013; 41, pp. 268-273. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/24221921]

49. Kadrmas, DJ; Christian, PE. Comparative evaluation of lesion detectability for 6 PET imaging platforms using a highly reproducible whole-body phantom with 22Na lesions and localization ROC analysis. J. Nucl. Med.; 2002; 43, pp. 1545-1554. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/12411558]

50. Fedrigo, R. et al. Development of scalable lymphatic system in the 4D XCAT phantom: Application to quantitative evaluation of lymphoma PET segmentations. Med. Phys. Published online, https://doi.org/10.1002/mp.15963 (2022).

51. Martelli, M et al. 18F]fluorodeoxyglucose positron emission tomography predicts survival after chemoimmunotherapy for primary mediastinal large B-cell lymphoma: Results of the International Extranodal Lymphoma Study Group IELSG-26 study. J. Clin. Oncol.; 2014; 32, pp. 1769-1775. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/24799481]

52. Hirtl, A et al. Technical Note: Fully-automated analysis of Jaszczak phantom measurements as part of routine SPECT quality control. Med Phys.; 2017; 44, pp. 1638-1645.1:CAS:528:DC%2BC2sXhtFCktbvK [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/28186647]

53. Lasnon, C et al. Contrast-enhanced small-animal PET/CT in cancer research: Strong improvement of diagnostic accuracy without significant alteration of quantitative accuracy and NEMA NU 4-2008 image quality parameters. EJNMMI Res.; 2013; 3, pp. 1-11.

54. Wahl, RL; Sunderland, J. Radiopharmaceutical dosimetry for cancer therapy: From theory to practice. J. Nucl. Med.; 2021; 62, pp. 2-3.

55. Ceriani, L et al. Utility of baseline 18FDG-PET/CT functional parameters in defining prognosis of primary mediastinal (thymic) large B-cell lymphoma. Blood; 2015; 126, pp. 950-956.1:CAS:528:DC%2BC2MXhslSks7rF [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26089397]

56. Zwanenburg, A et al. The image biomarker standardization initiative: Standardized quantitative radiomics for high-throughput image-based phenotyping. Radiology; 2020; 295, pp. 328-338. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32154773]

57. Gatto, A; Iuliano, L. Evaluation of inserts for modular thermoplastic injection moulds produced by spin casting. J. Mater. Process Technol.; 2001; 118, pp. 411-416.

58. Shetty, B., Naveen G. J., Rao S. Silicone mold making: A versatile and cost-effective solution for diverse applications. In:20006, https://doi.org/10.1063/5.0235868 (2024).

59. Store BG. Pro-NM NEMA NU2. Accessed, https://www.bcgroupstore.com/Pro-NM-NEMA-NU2-P3251.aspx (2025).

60. Visser, EP et al. Image quality phantom and parameters for high spatial resolution small-animal SPECT. Nucl. Instrum. Methods Phys. Res. A; 2011; 654, pp. 539-545.1:CAS:528:DC%2BC3MXhtFyntb%2FI

61. Fedrigo, R., Coope R., Rahmim A., Bénard F., Uribe C. F. Development of the quantitative PET prostate phantom (Q3P) for improved quality assurance of 18 F‐PSMA PET imaging in metastatic prostate cancer. Med. Phys. Published online February 13, https://doi.org/10.1002/mp.16977 (2024).

62. Kao, YH; Luddington, OS; Culleton, SR; Francis, RJ; Boucek, JA. A Gelatin liver phantom of suspended 90Y resin microspheres to simulate the physiologic microsphere biodistribution of a postradioembolization liver. J. Nucl. Med Technol.; 2014; 42, pp. 265-268. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/25472514]

63. In, E; Walker, E; Naguib, HE. Novel development of 3D printable UV-curable silicone for multimodal imaging phantom. Bioprinting; 2017; 7, pp. 19-26.

64. Rai, R et al. 3D printed phantoms mimicking cortical bone for the assessment of ultrashort echo time magnetic resonance imaging. Med Phys.; 2018; 45, pp. 758-766. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/29237232]

65. Shen, S. et al. Freeform fabrication of tissue-simulating phantom for potential use of surgical planning in conjoined twins separation surgery. Sci. Rep. 7, https://doi.org/10.1038/s41598-017-08579-6 (2017).

66. Niebuhr, NI et al. Technical Note: Radiological properties of tissue surrogates used in a multimodality deformable pelvic phantom for MR-guided radiotherapy. Med Phys.; 2016; 43, pp. 908-916. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26843251]

67. Lodge, OJH; Wahl, MA. RL. Practical PERCIST: A simplified guide to PET response criteria in solid tumors 1.0. Radiology; 2016; 280, pp. 576-584. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26909647]

68. Sgouros, G et al. Tumor response to radiopharmaceutical therapies: The knowns and the unknowns. J. Nucl. Med.; 2021; 62, pp. 12S-22S.1:CAS:528:DC%2BB38XitFCjtb0%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34857617][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12079724]

69. Rahmim, A et al. Theranostic digital twins for personalized radiopharmaceutical therapies: Reimagining theranostics via computational nuclear oncology. Front Oncol.; 2022; 12, pp. 1-7.

70. Brosch-lenz, J., Uribe C., Rahmim A., Saboury B. Theranostic Digital Twins: An indispensable prerequisite for personalized cancer care. Published online:1-4, https://doi.org/10.2967/jnumed.122.264929 (2022).