Monte Carlo simulation and experimental validation of SPECT systems via energy spectrum decomposition and image profile analysis

Abstract

Background

Accurate modelling and performance evaluation of gamma-ray scintillation detectors are required in single photon emission computed tomography (SPECT) systems to ensure reliable nuclear medicine diagnostics. Monte Carlo simulation is therefore a powerful computational technique used to accurately model the gamma-ray scintillation detectors and photon transport in medical imaging systems. Therefore, the objective of this study is to validate the SIMIND Monte Carlo simulation code using a self-developed peak-detection algorithm for clinical SPECT gamma cameras (Siemens E.Cam and GE Infinia), with a focus on energy-spectrum decomposition and image-profile analysis.

Methods

This study presents a dual-system evaluation of clinical SPECT gamma cameras—the Siemens E.Cam and GE Infinia systems using the SIMIND Monte Carlo simulation code. We assessed collimator and detector-specific SPECT performance metrics, including energy resolution, sensitivity, image evaluation, spatial resolutions, and spectral fidelity for a ^99mTc point source and spatial resolution with Co-57 flood source. A custom algorithm was then developed to automatically decompose energy spectra and identify key spectral components, including the photopeak, Compton edge, iodine escape, and lead X-ray peaks. Quantitative methods, such as 2D and 3D image profile analysis and FWHM comparisons, were integrated to validate the simulation against experimental data.

Results

Energy resolution differences ≤ 4% and sensitivity variations < 1.2% were observed between simulated and experimental measurements. Spatial resolution analysis revealed a range of 3.5–4.8 mm for Siemens E.Cam (bar and pie phantoms) and ~ 1 cm for GE Infinia, highlighting system-specific imaging capabilities. A custom algorithm successfully identified key spectral components, including photopeaks, Compton edges, and lead X-ray peaks, with deviations < 3%. Image profile analysis revealed FWHM values of 4–4.5 pixels for both systems.

Conclusion

This work shows that SIMIND accurately reproduces the fundamental photon interaction characteristics including energy spectra, energy resolution, and sensitivity for both Siemens E.Cam and GE Infinia SPECT systems. However, for image-profile evaluation, good agreement was obtained for Siemens E.Cam, whereas the GE Infinia results revealed a system-specific discrepancy associated with the actual clinical condition of the installed camera. Nevertheless, when used appropriately, SIMIND provides a robust framework for optimizing quantitative SPECT protocols and guiding system design improvements, particularly for well-calibrated systems.

Graphical abstract

Introduction

Single photon emission computed tomography (SPECT) is a nuclear medicine imaging technique that generates three-dimensional functional images by detecting gamma rays emitted from radiopharmaceuticals administered to patients. A gamma camera rotates around the patient during a SPECT scan, capturing projection data from multiple angles. These projections are then reconstructed into cross-sectional images, enabling physicians to assess physiological processes such as metabolic activity or blood flow. However, optimizing SPECT systems for clinical use requires rigorous performance evaluation of scintillation detectors. Therefore, a regular detector-specific characteristic performance is required to analyze the system's behavior.

Nuclear medicine imaging depends on precise and quantitative models to accurately represent the interactions between photons and biological tissues. These interactions are governed by the radiative transport equation (RTE), a mathematical framework that describes how radiation propagates and scatters within matter. Unfortunately, analytical solutions to the RTE are limited to basic domains, including infinite slabs, semi-infinite media, and infinite media; for complex or random media, numerical computations are necessary. Therefore, the Monte Carlo method solves the RTE by simulating the paths of many photons (or quanta of energy) as they propagate through the medium, experiencing scattering, absorption, and possibly emission. Each photon's path is determined by random sampling from probability distributions that represent the physical processes described by the RTE. Monte Carlo simulation is therefore a powerful computational technique used to model the behavior of photons in medical imaging systems. It is especially useful in scenarios where analytical solutions are difficult or impossible to obtain due to the complexity of the system [1].

Over the years, several Monte Carlo codes have been developed for the simulation of radiation transport, such as EGS [2, 3], EGSnrc [4], Geant4 [5], PENELOPE [6], SuperMC [7, 8]. These codes are capable of simulating particle transport and their interactions with matter under almost all circumstances. However, modeling nuclear medicine facilities such as SPECT and PET (Positron Emission Tomography) is more convenient in dedicated Monte Carlo codes, which include SIMIND [9], GATE [10], and SimSET [11]. SIMIND mainly focused on simulating standard clinical SPECT systems, with particular emphasis on accurate modeling of collimator-detector interactions and photon transport. It can be easily modified to handle different types of calculations and measurements related to SPECT imaging, such as studying the effects of patient motion, photon attenuation, and scatter on image quality.

A limited number of studies [12–21] have been done for the simulation of the SPECT system through the Monte Carlo code SIMIND [9]. Performance evaluation of the Siemens E.Cam system through SIMIND was done by Bahreyni Toossi et al. [12]. A similar analysis was performed by Rodrigues et al. [16], where GEANT4 has been used as a simulation toolkit. However, a gap exists in comprehensive energy spectrum analysis, including quantitative characterization of secondary peaks (e.g., Compton edge, iodine escape, lead X-rays), critical for spectral fidelity. This study introduces a custom peak identification algorithm (“Energy spectrum analysis and peak characterization algorithm” section) for detailed energy spectrum decomposition, enabling the automatic identification and classification of photon interaction peaks, including photopeaks, backscatter, Compton edges, and iodine escape peaks. This approach rigorously validates simulated spectra against experiments, addressing a key limitation.

In this work, we simulated the performance of Siemens’ E.Cam dual-head gamma camera using the SIMIND Monte Carlo program [9], installed in our nuclear medicine center, to evaluate the collimator and detector-specific performance in terms of energy resolution, sensitivity, image evaluation, and spatial resolution using ^99mTc point source. A new method of image evaluation is proposed by drawing image profiles and comparing the corresponding FWHM. Further, a comprehensive analysis of measured and simulated energy spectra is carried out to compare all peaks that appeared in the spectrum due to photon interactions with matter.

Methods and materials

The Siemens E.Cam dual-head gamma camera (Siemens Medical Solutions, Hoffman Estates, USA) and the GE Infinia gamma camera were modeled in SIMIND Monte Carlo code (v6.1.2), incorporating manufacturer-specified detector geometries and material properties. Experimental and simulated performance metrics, including energy resolution, sensitivity, spatial resolution, and image evaluation, were assessed for both Siemens E.Cam and GE Infinia systems using standardized phantoms (bar, pie, and cylindrical) and radioactive sources (^99mTc point source, ⁵⁷Co flood source) simulations.

Experimental setup

To evaluate energy resolution, sensitivity, and image quality, a point source was prepared in the hot lab using a small piece of cotton and a syringe cap (serving as a cylindrical phantom). The point source is an isotropic ^99mTc source emitting 140 keV photons. Two gamma camera systems, Siemens E.Cam and GE Infinia, were then used to carry out the performance analysis. The Siemens E.Cam dual-head gamma camera features two gamma-ray detector units, each integrating a removable Low Energy All Purpose (LEAP) collimator optimized for 140 keV photons, coupled to a thallium-activated sodium iodide (NaI(Tl)) scintillation crystal (9.5 mm thickness, 40,000 photons/MeV light yield) with peak emission at 415 nm. The optical system consists of a Pyrex light guide (n = 1.47) that efficiently channels scintillation photons to a 59-channel photomultiplier tube (PMT) array (Hamamatsu R6231-100), achieving 9.9% energy resolution at 140 keV. This detector configuration enables precise spatial and energetic characterization of gamma emissions through coordinated signal readout from both detector heads.

The same performance analysis tests have also been done on the GE Infinia dual-head camera to ensure the versatility of the SIMIND Monte Carlo code. The GE Infinia system also incorporates two detector assemblies, each equipped with a modular Low Energy General Purpose (LEGP) collimator. The LEGP collimator consists of parallel hexagonal channels (1.9 mm pore diameter, 35 mm length, 0.20 mm septal thickness), optimized for imaging low-energy γ-emitters like ^99mTc (photopeak 140 keV). This honeycomb structure provides a geometric efficiency of 1.2 × 10^–4 while maintaining 90% septal penetration rejection at 140 keV.

Monte Carlo simulation

The Monte Carlo code SIMIND was used to model both gamma camera systems to carry out their performance evaluation, as shown in Fig. 1. SIMIND employs a photon transport algorithm that tracks each gamma ray through successive interactions, including Compton scattering (modeled using the Klein–Nishina differential cross-section), photoelectric absorption (with characteristic X-ray emission), and coherent scattering. For NaI(Tl) detectors, SIMIND also simulates scintillation light generation, light transport through the crystal and light guide, and photomultiplier tube (PMT) signal formation.

Fig. 1.

Construction of a gamma camera in the SIMIND Monte Carlo code

A computer equipped with AMD Ryzen 7 4700U 2.00 GHz (8 CPUs) with Radeon(TM) Graphics and 8 GB memory is used for the simulation of the SPECT system. The SIMIND Monte Carlo simulation code employs a dual-program architecture for radiation detector modeling. The CHANGE module provides a graphical interface for configuring the gamma camera's physical parameters, including source geometry, collimator specifications (hole pattern, septal thickness), and detector properties (scintillator thickness, material composition). These user-defined parameters are stored in standardized input files. The SIMIND then processes these inputs using a photon transport algorithm that accounts for Compton scattering, photoelectric absorption, and characteristic X-ray emission within the detector system.

As documented in Table 1, all simulations incorporated manufacturer-specified parameters for both the Siemens E.Cam and GE Infinia systems, ensuring physical accuracy in modeling the radiation detection from photon interaction to signal generation.

Table 1.

Parameters of Siemens E.Cam and GE Infinia defined in the SIMIND Monte Carlo code.

Parameters		Specifications
		Siemens E.Cam	GE Infinia
Source	Sources	(i) Tc-99m Point source (ii)  57Co Flood source (5 mCi) for spatial resolution
	Phantom	Cylindrical Phantom for point source, Bar phantom, Pie phantom
	Source activity	51.06 MBq	24.79 MBq
	Number of counts	128 million counts
	Gamma-ray energy	140 keV
Detector	Crystal	NaI(Tl)
Collimator	Collimator type	Low Energy All Purpose (LEAP)	Low Energy General Purpose (LEGP)
	Hole shape	Hex
	No. of holes	90000	56560
	Hole length	24.1 mm	35 mm
	Septal thickness	0.2 mm	0.2 mm
	Hole diameter	1.45 mm	1.9 mm
Image	Pixel size	0.39 mm 128x128
	Matrix size
Camera Parameters	Source to detector height	25 cm	25 cm
	Back-scattering (BK) material	Pyrex	Pyrex
	Thickness of BK material	6 cm	6 cm
	Intrinsic spatial resolution	3.9 mm	3.9 mm
	Energy resolution	9.90%	9.80%
	Energy window	15%	10%

Performance evaluation of gamma-ray scintillation detectors

The performance of gamma-ray scintillation detectors was evaluated through four critical metrics: energy spectrum analysis (characterizing photopeaks and secondary features), sensitivity quantification, spatial resolution assessment using specialized phantoms, and image evaluation via 2D and 3D profile comparisons. Both experimental measurements and SIMIND Monte Carlo simulations were conducted in parallel for the Siemens E.Cam and GE Infinia systems, enabling direct validation of detector models against clinical performance data under identical conditions.

Energy spectrum analysis and peak characterization algorithm

For energy resolution measurements, a lower activity of 3.7 MBq was used to minimize pulse pile-up effects, while higher activities (51.06 MBq) were employed for sensitivity and image quality assessments to achieve adequate count statistics within clinical acquisition times. Energy resolution was measured using a 3.7 MBq ^99mTc point source centered in the detector field of view at 25 cm from the NaI(Tl) crystal surface. For the Siemens E.Cam system, energy spectra were acquired under two configurations: (1) with the LEAP collimator (1.45 mm hexagonal holes, 24.1 mm length) to assess extrinsic resolution, and (2) without collimation to determine intrinsic detector performance. Each measurement collected 1 million photopeak events (140 keV ± 10%) to ensure statistical significance (< 2% Poisson noise). The GE Infinia system was similarly evaluated using its LEGP collimator (1.9 mm holes, 35 mm length), maintaining identical source geometry for comparative analysis. Monte Carlo simulations in SIMIND replicated these experimental conditions, including exact collimator geometries and the 25 cm source-to-detector distance.

For a ^99mTc source (140 keV gamma rays) interacting with a NaI(Tl) scintillation crystal, the photopeak represents full energy deposition, the Compton edge arises from inelastic scattering, the iodine escape peak (112 keV) results from characteristic X-ray emission following photoelectric absorption in iodine, and lead X-ray peaks (~ 72–75 keV) originate from collimator interactions. To quantitatively analyze the energy spectra acquired through experimental measurements and SIMIND Monte Carlo simulations, a custom algorithm was developed to identify and characterize key photopeaks and secondary spectral features (Compton edge, back-scatter peak, characteristic Lead X-ray peak, and Iodine escape peak). Accurate energy spectrum analysis is critical for validating Monte Carlo detector models because different photon interaction mechanisms produce distinct spectral features. This analysis was critical for determining the energy resolution, full width at half maximum (FWHM), and spectral fidelity of both gamma camera systems.

(a) Calculation of energy resolution and FWHM

The energy resolution (R) is given as the ratio of the full width at half maximum (FWHM) of the photopeak to its peak energy E_peak, expressed as a percentage.

1

For ^99mTc, the nominal peak energy is 140 keV. To compute FWHM reliably from discrete energy spectra, the following sequential procedure was applied identically to both experimental and simulated data.

(1) Smoothing: The raw spectrum often contains statistical fluctuations due to Poisson noise. Therefore, prior to peak identification, smoothing filters are applied to reduce statistical fluctuations in measured spectra [23–25]. To enhance peak visibility and stability of calculations, a two-sided moving average filter was applied. For each bin i, the smoothed count value S_smooth(i) was calculated by averaging the counts in the interval from i−w to i+w, where w is the half-window size of the filter and S(j) denotes the original spectrum count in bin j:

2

(2) Peak detection: After smoothing, a peak was identified as a local maximum satisfying the condition given in equation (3).

3

Only peaks with heights exceeding a certain threshold (≥ 5% of the maximum spectrum intensity) were selected for further analysis. The energy corresponding to a detected peak at index p is assigned as the peak energy, E_peak = E(p), where E(p) is the calibrated energy value corresponding to bin p. The 140 keV photopeak was selected for FWHM and energy-resolution calculation.

(3) Half-maximum positions: In order to calculate FWHM, the two energy values E_L (left half-maximum energy) and E_R (the right half-maximum energy) are calculated where the peak amplitude S_smooth(p) of the smoothed spectrum falls to half its maximum height. Let H = S_smooth(p) be the peak height, then:

4

Because the spectrum is discrete, linear interpolation between adjacent bins gives:

5

where is the nearest bin to the peak on the left side for which , and . Thus E_L is obtained by interpolation between bins and . Similarly, the right half-maximum energy E_R is

6

where is the nearest bin to the peak on the right side for which , and . Thus E_R is obtained by interpolation between bins and .

(4) FWHM and energy resolution: The full width at half maximum, as shown in Fig. 2, is then calculated as

7

Fig. 2.

Schematic illustration of the interpolation method used to determine the left and right half-maximum energies, E_L and E_R, from a smoothed energy spectrum

Finally, the energy resolution R of the 140 keV photopeak was calculated using Eq. (1). This value provides a quantitative assessment of the detector’s ability to resolve monoenergetic gamma rays. The interpolation procedure used to determine the left and right half-maximum energies, E_L and E_R, corresponding to Eqs. (5) and (6), is illustrated schematically in Fig. 2.

(b) Peak classification

Each peak was classified based on its central energy E_peak using known interaction signatures for ^99mTc. Table 2 summarizes the characteristic spectral features observed when 140 keV ^99mTc γ-rays interact with NaI(Tl), including the photopeak, Compton edge, backscatter peak, iodine escape peak, and lead X-ray peaks, which collectively enable detector performance evaluation.

Table 2.

Energy calculation of different peaks in the energy spectrum of ^99mTc

Peaks in energy spectrum	Calculation
Photo-peak	140 keV
Compton edge
Back-scatter peak	= 90.442 keV
Characteristic Lead X-ray peak	72 keV
Iodine escape peak	= 112 keV

The classification was determined by evaluating:

8

This algorithm was implemented in C++ and applied identically to both experimental and simulated spectra, ensuring uniformity in performance validation across both modalities. It enabled high-fidelity peak characterization critical for detector benchmarking and image quality analysis in SPECT systems. The pseudocode of this algorithm is given in “Pseudocode for energy spectrum analysis” Appendix section.

Sensitivity

The sensitivity of the gamma-ray detector was determined as

9

Where N_det is the total counts in the photopeak window (126–154 keV), t is acquisition time (in seconds), and A is the source activity (MBq).

Spatial resolution

Spatial resolution of the Siemens E.Cam dual-head gamma camera and GE Infinia system was measured by placing a flood source of ⁵⁷Co (5 mCi) on the top of the bar phantom. The bar phantom is placed on the top of the collimator, while simulations employed a QC Pie Phantom to qualitatively assess resolution limits.

Image evaluation

The acquisition parameters for experiments were a 128 × 128 matrix, 0.39 mm pixel size, and approximately 1 million counts/projection. Image and corresponding 2-D and 3-D profiles were obtained with collimators through both gamma camera systems and compared with their simulated image profiles and corresponding FWHM.

Results

Energy resolution

The simulated and measured energy spectra produced by Siemens E.Cam dual-head gamma camera with and without LEAP collimator for ^99mTc point source (γ -ray energy of 140 keV) are shown in Fig. 3. An excellent agreement is obtained between the simulated and the measured energy spectra for both cases.

Fig. 3.

Comparison of the measured and simulated energy spectra for a ^99mTc point source from Siemens E.Cam a with LEAP collimator, and b without collimator

For the 140 keV photopeak, the experimental FWHM values were 14.58 keV (both with and without LEAP collimator), corresponding to relative energy resolutions of 10.42%. Monte Carlo simulations yielded comparable results, with FWHM values of 14.05 keV (10.03%) and 14.11 keV (10.08%) for collimated and uncollimated configurations, respectively. This ≤ 0.4% deviation between simulated and experimental resolutions validates the accuracy of our detector modeling approach. Notably, both results align well with the manufacturer's specified resolution [25] of 9.9% at 140 keV, confirming proper system calibration and simulation parameterization.

The simulated and experimentally determined values of FWHM, energy resolution, and energies of other photon interactions of a ^99mTc point source acquired through Siemens E.Cam are given in Table 3. Percentage differences are calculated as

10

Table 3.

Results from the energy spectrum of a ^99mTc point source acquired through Siemens E.Cam

Parameters	Simulated		Measured		% Difference
	With LEAP Collimator (Extrinsic)	Without LEAP Collimator (Intrinsic)	With LEAP Collimator (Extrinsic)	Without LEAP Collimator (Intrinsic)	Extrinsic	Intrinsic
FWHM (keV)	14.05	14.11	14.58	14.58	3.6	3.2
Energy resolution (%)	10.03	10.08	10.42	10.42	3.7	3.3
Compton edge (keV)	47.2	47	47	47	0.4	0
Pb X-Ray peak (keV)	74.9	–	75	–	0.1	–
Back-scatter peak (keV)	88.8	92	90.3	90	1.7	2.2
Iodine escape peak (keV)	113	111.9	112	111.5	0.9	0.4
Photo-peak (keV)	140.95	140.8	140	140	0.68	0.57

The GE Infinia system also demonstrated strong agreement between measured and simulated performance metrics, as shown in Fig. 4. Experimental measurements yielded an FWHM of 13.44 keV for the 140 keV photopeak, corresponding to an energy resolution of 9.80%. Monte Carlo simulations predicted slightly broader spectral resolution, with FWHM = 13.84 keV (9.60% resolution). This close correspondence validates the accuracy of our Monte Carlo model in replicating the LEGP collimator's photon detection characteristics.

Fig. 4.

Comparison of the measured and simulated energy spectra for a ^99mTc point source from GE Infinia with LEGP collimator

The simulated and experimentally determined values of FWHM, energy resolution, and energies of other photon interactions of a ^99mTc point source acquired through GE Infinia are given in Table 4.

Table 4.

Results from the energy spectrum of ^99mTc point source acquired through GE Infinia system

	FWHM (keV)	Energy resolution (%)	Compton edge (keV)	Pb X-Ray peak (keV)	Back-scatter peak (keV)	Iodine escape peak (keV)	Photo-peak (keV)
Simulated	13.84	9.6	47.02	75	89	112.3	141.1
Measured	13.44	9.8	47	74.02	90.1	112	140
% Difference	2.97	2.04	0.04	1.3	1.22	0.27	0.79

Sensitivity

Sensitivity measurements demonstrated close agreement between experimental and simulated values for both detector systems. The Siemens E.Cam showed experimental sensitivity of 136.83 cps/MBq compared to simulated values of 138.35 cps/MBq, representing a 1.1% difference. Similarly, the GE Infinia exhibited experimental sensitivity of 107.67 cps/MBq versus simulated 108.91 cps/MBq, representing a 1.15% difference.

Image evaluation

Image evaluation of siemens E.Cam

The acquired images from the Siemens E.Cam system and their corresponding spatial profiles are presented in Figs. 5, 6 and 7.

Fig. 5.

Images of ^99mTc point source obtained by the a simulation of Siemens E.Cam and through the b experiment

Fig. 6.

a Horizontal image profiles for simulated and actual images, b vertical image profiles for simulated and actual (measured) images

Fig. 7.

a simulated 3D profiles of the image of ^99mTc point source, b measured 3D profile of the image of the ^99mTc point source.

Figure 6 demonstrates strong agreement between the simulated and experimental image profiles, with all profiles exhibiting a full width at half maximum (FWHM) within the range of 4–4.5 pixels. This close correspondence is further supported by the 3D image profiles shown in Fig. 7, where the simulated and measured data show minimal divergence in spatial response.

Image evaluation of GE Infinia

Figures 8, 9 and 10 present the acquired images from the GE Infinia system and their corresponding spatial profiles.

Fig. 8.

Images of ^99mTc point source obtained by the a simulation of GE Infinia and through the b experiment

Fig. 9.

a Horizontal image profiles for simulated and actual images, b vertical image profiles for simulated and measured images

Fig. 10.

a simulated 3D profiles of the image of ^99mTc point source, b 3D profile of the measured image of ^99mTc point source

As evident in Fig. 8b, the experimental data exhibit sub-optimal uniformity and linearity, resulting in broadened point-source profiles without a distinct sharp peak. While the simulated profiles (Figs. 9, 10) align well with theoretical expectations, they replicate the observed “star effect” artifact due to the inherent limitations of the clinical GE Infinia system, as confirmed by the measured image quality in Fig. 8b. The differences in image profiles, as shown in Figs. 9, 10, acquired through simulation and experiment, are because of the poor image quality of the GE Infinia installed in our nuclear medicine center, and this can be verified from Fig. 8b.

Spatial resolution

Spatial resolution of siemens E.Cam

The spatial resolution of the Siemens E.Cam dual-head gamma camera was measured by placing a 5 mCi ⁵⁷Co flood source on top of a bar phantom, which was positioned directly on the collimator. In the simulation, the spatial resolution was assessed using a QC Pie Phantom. The resulting images from both the experiment and simulation are shown in Fig. 11.

Fig. 11.

a Actual bar phantom image obtained from Siemens E.Cam, b Simulated pie phantom image

Notably, in the Pie Phantom, all rods are distinctly visible, with the smallest rod diameter being 4.8 mm, suggesting that the spatial resolution of the Siemens E.Cam could be as fine as 4.8 mm. The bar phantom image further supports this, indicating a resolution of approximately 3.5 mm or slightly higher. Based on these findings, it can be concluded that the spatial resolution of the Siemens E.Cam installed at our nuclear medicine facility ranges between 3.5 and 4.8 mm.

Spatial resolution of GE Infinia

The spatial resolution of the GE Infinia dual-head gamma camera was measured using the same method as that of the Siemens E.Cam. A bar phantom was placed on the collimator for the experimental measurement, while the spatial resolution in the simulation was qualitatively assessed using a QC Pie Phantom. The resulting images from both the experiment and simulation are presented in Fig. 12.

Fig. 12.

a Actual bar phantom image obtained from GE Infinia, b simulated pie phantom image

In the Pie Phantom, all rods are clearly visible, with the smallest rod diameter being 4.8 mm, suggesting that the spatial resolution of the GE Infinia could be as fine as 4.8 mm. However, the bar phantom image yielded disappointing results, indicating a resolution of approximately 1 cm, which corresponds to the separation between the four quadrants. Thus, the spatial resolution of the GE Infinia installed at our nuclear medicine facility appears to be around 1 cm.

Discussion

The findings validate that the SIMIND Monte Carlo algorithm can accurately model the energy-spectral and sensitivity characteristics of clinical gamma camera systems (Siemens E.Cam and GE Infinia) under standardized testing conditions. However, the image profile discrepancies observed for the GE Infinia system are attributed to detector non-uniformity and electronic noise in the clinical installation, demonstrate that SIMIND models an idealized system.

Energy resolution: The measured and simulated energy resolutions for both systems exhibited close agreement, with discrepancies of ≤ 3.7% for Siemens E.Cam with LEAP collimator, and ≤ 3.3% for Siemens E.Cam without collimator. For the GE Infinia system with LEGP collimator, the difference between simulation and experiment was approximately 2.0%. This close agreement demonstrates that the SIMIND Monte Carlo model accurately reproduces both intrinsic and extrinsic detector energy resolution characteristics, including scintillation photon generation, photon transport in the NaI(Tl) crystal, and photomultiplier signal response.

To further evaluate the accuracy of the intrinsic detector modeling, the obtained energy resolution values were compared with previously published Monte Carlo studies of Siemens gamma cameras [12, 13]. Table 5 summarizes the intrinsic energy resolution values reported in the literature [12, 13] for Tc-99 m (140 keV), together with the present study results and manufacturer’s specified data [25], demonstrating that all values fall within the expected 9.5–10.5% range for clinical SPECT systems. The simulated and measured intrinsic energy resolution obtained in the present study is 10.08 and 10.42%, respectively, which is in good agreement with previously reported values 10.01% and 9.68% by Bahreyni Toossi et al. [12] and 9.68% and 9.54% by Morphis et al. [13] for Siemens gamma camera systems. The slight differences between studies may arise from variations in detector configuration (Siemens E.Cam vs Siemens Symbia T16 [12, 13]), source size, and the specific energy-resolution modeling approach implemented in the SIMIND code.

Table 5.

Comparison of intrinsic energy resolution of Siemens gamma cameras for Tc-99m (140 keV)

Study		Manufacturer	This Study	Bahreyni Toossi et al. [12]	Morphis et al. [13]
System		Siemens E.Cam	Siemens E.Cam	Siemens E.Cam	Siemens Symbia T16
Energy Resolution (%)	Simulated	–	10.08	10.01	9.68
	Measured	≤ 9.9	10.42	9.86	9.54

In addition to intrinsic detector performance, extrinsic energy resolution was also compared with previously reported results obtained using different collimator configurations. Table 6 summarizes the simulated and measured extrinsic energy resolutions for Tc-99 m reported in the present study and in earlier SIMIND-based gamma camera simulations [12, 13]. The comparison includes different collimator types, including LEGP, LEAP, and LEHR (Low Energy High Resolution), which influence photon penetration and scatter within the detector system. Despite these differences, the results remain consistent with manufacturer-specified extrinsic detector performance of approximately 9–10% at 140 keV.

Table 6.

Comparison of extrinsic energy resolution of GE Infinia and Siemens gamma cameras for Tc-99m (140 keV)

Study		This Study		Bahreyni Toossi et al. [12]	Morphis et al. [13]
System		GE Infinia LEGP	Siemens E.Cam LEAP	Siemens E.Cam LEHR	Siemens Symbia T16 LEHR
Energy Resolution (%)	Simulated	9.6	10.03	9.94	9.71
	Measured	9.8	10.42	9.61	9.36
	Manufacturer [25, 26]	≤ 9.8	≤ 9.9	≤ 9.9	≤ 9.9

Figure 13 shows the comparison between simulated and measured energy resolution values at 140 keV (⁹⁹ᵐTc) for different SPECT systems and collimator configurations reported in this study and previous Monte Carlo validation studies, as shown in Tables 5 and 6. The proximity of the data points to the reference line indicates good agreement between simulated and experimentally measured detector energy resolutions across different studies.

Fig. 13.

Comparison between simulated and measured energy resolution values at 140 keV (⁹⁹ᵐTc) for different SPECT systems and collimator configurations reported in this study and previous Monte Carlo validation studies

Energy spectra: The comparison of simulated and measured energy spectra also confirms the accuracy of the detector model. The proposed peak detection algorithm successfully identified all principal spectral features produced by 99mTc photon interactions in NaI(Tl), including the photopeak (140 keV), Compton edge (~ 47 keV), Pb characteristic X-ray peak (~ 72–75 keV), backscatter peak (~ 90 keV), and iodine escape peak (~ 112 keV). The differences between simulated and measured peak positions were below 3% for both systems, demonstrating good spectral fidelity.

A comparison with previously published simulated spectra of the Siemens E.Cam system reveals that the primary emphasis has been placed on photopeak modeling and detector performance metrics such as sensitivity and spatial resolution [12, 13]. In particular, the backscatter peak (~ 90 keV) and iodine escape peak (~ 112 keV) are not clearly separated in the simulated spectra reported in previous SIMIND-based analyses [12]. In contrast, the present study explicitly identifies and quantitatively characterizes these secondary peaks using the proposed automated peak identification algorithm.

Sensitivity: Sensitivity comparison analyses indicated discrepancies of around 1.1% for Siemens E.Cam and 1.15% for GE Infinia, both of which fall within the bounds of acceptable experimental errors. The sensitivity measure is significantly affected by parameters like collimator geometry and detector material, which were precisely incorporated into the simulation according to manufacturer specifications. The < 2% overall variation between measured and simulated sensitivities validates the accuracy of our detector modeling through the Monte Carlo model while highlighting areas for future refinement in signal processing simulation.

Image evaluation: The analysis of 2D and 3D image profiles for Siemens E.Cam demonstrated a strong agreement between simulation and experimental results, with FWHM values ranging from 4 to 4.5 pixels. This signifies that spatial response attributes, encompassing intrinsic resolution and collimator blur, are accurately represented in the simulation.

However, in the GE Infinia, wider experimental profiles and diminished peak sharpness were noted, indicating inherent constraints of the particular system implemented at the site, including compromised detector uniformity and electronic noise.

Spatial resolution: The Siemens E.Cam system exhibited spatial resolution between 3.5 mm (bar phantom) and 4.8 mm (pie phantom), aligning with the anticipated performance of an adequately operating LEAP-collimated system. In contrast, the GE Infinia system demonstrated markedly worse experimental resolution (~ 1 cm), which may be attributable to detector age or inadequate calibration.

A noticeable difference was observed between the simulated and experimentally measured image profiles for the GE Infinia system. While the SIMIND Monte Carlo simulation produced well-defined point-source profiles consistent with theoretical expectations, the experimentally acquired images exhibited broadened peaks and reduced uniformity.

The discrepancy between the simulated and measured image profiles of GE Infinia is mainly attributed to the actual operating condition of the GE Infinia system installed at our facility, rather than to the SIMIND Monte Carlo model. SIMIND models an ideal and properly calibrated SPECT system based on the defined physical and geometrical parameters. Therefore, if the clinical detector has local hardware, calibration, or uniformity-related problems, these effects may not be fully reproduced unless they are specifically characterized and introduced into the simulation model. In clinical gamma cameras, factors such as detector non-uniformity, photomultiplier tube (PMT) gain variations, electronic noise, and imperfect energy or uniformity calibration can degrade the spatial response function and introduce artifacts such as star patterns or profile broadening. Importantly, despite these spatial differences, the energy spectrum characteristics, energy resolution, and sensitivity measurements for the GE Infinia system remained in close agreement with the Monte Carlo simulations. This confirms that the SIMIND-based detector model accurately reproduces the fundamental photon interaction processes in the detector while also demonstrating the capability of the proposed methodology to identify performance deviations in clinical imaging systems.

Energy spectrum: The self-developed peak detection algorithm effectively recognized and categorized all principal spectral components (photopeak, Compton edge, iodine escape, Pb X-rays, backscatter peaks), exhibiting less than 3% difference between simulated and experimental peak positions for both systems. Furthermore, this also validates accurate modeling of photon interactions in NaI(Tl) crystals by the SIMIND Monte Carlo code.

Conclusion

This study presents a comprehensive Monte Carlo validation of two clinical gamma camera systems (Siemens E.Cam and GE Infinia) using the SIMIND Monte Carlo code. The results demonstrate that SIMIND can accurately reproduce the energy-spectrum characteristics, energy resolution, characteristic peak positions, and sensitivity of both systems under idealized detector-modeling conditions. For image-profile evaluation, good agreement was obtained for Siemens E.Cam, whereas the GE Infinia results revealed a system-specific discrepancy associated with the actual clinical condition of the installed camera.

Furthermore, the integration of the peak detection and classification algorithm for energy-spectrum analysis, together with sensitivity, spatial resolution, and image-profile evaluations, provides a useful methodology for SPECT performance characterization. Importantly, this framework not only validates the agreement between simulation and measurement for well-modeled detector responses but also helps identify deviations between ideal Monte Carlo predictions and the actual performance of clinical gamma camera systems. In addition, a comparison with previously published Monte Carlo validation studies demonstrates that the obtained energy resolution and spectral characteristics are consistent with reported values for clinical NaI(Tl)-based SPECT systems, further confirming the reliability of the proposed simulation and analysis framework. These findings confirm that the simulated system is suitable for further research, encompassing modifications, planning, and the development of a SPECT system to enhance image quality. This validation represents an essential step before the simulation can be used in more complicated studies of optimizing performance characteristics.

Abbreviations

MC

Monte Carlo

SPECT

Single photon emission computed tomography

FWHM

Full width at half maximum

PMT

Photomultiplier tube

NaI(Tl)

Thallium-activated sodium iodide

RTE

Radiative transport equation

cps/MBq

Counts per second per megabecquerel

LEAP

Low energy all purpose

LEGP

Low energy general purpose

LEHR

Low energy high resolution

Appendix

A1: Pseudocode for energy spectrum analysis

Author contributions

ZP and MAW designed the study, performed experiments and simulations for the ^99mTc source; IS and FW conducted experimental studies for the ⁵⁷Co flood source and validated the results; MAW and MS developed the peak detection algorithm and analyzed the spectral features. All authors wrote and reviewed the manuscript.

Funding

This work was supported by the Hunan Provincial Natural Science Foundation Project(No.2025JJ70160), Scientific Fund of Hunan Provincial Education Department (No.23A0629), Hunan Institute of Technology Provincial Applied Characteristic Discipline Open Project (Grant No. KF24014), Hunan Provincial Department of Education Key Teaching Reform Project (202401001561), and the College Students’ Innovation and Entrepreneurship Plan Projects of Hunan Province (Grant No. S202411528012, S202411528018S, S202411528108S, S202411528170, S202411528032X).

Data availability

The original contributions presented in this study are included in the article/supplementary material. Further inquiries can be directed to the corresponding author(s).

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.