Comparison of SZ-HRX and LEHR collimators for reduced-duration 123I-ioflupane brain SPECT/CT: a phantom study

Abstract

Background

¹²³I-ioflupane Single Photon Emission Computed Tomography with Computed Tomography (SPECT-CT) imaging is widely used to assess dopaminergic denervation in parkinsonian syndromes, such as Parkinson’s disease and atypical variants. Standard imaging generally uses low-energy high-resolution (LEHR) parallel-hole collimators, which require long acquisition times and minimal source-to-detector distance to optimize spatial resolution. Recently, the Smart-Zoom High-Resolution and eXtended magnification volume (SZ-HRX) solution was designed specifically for neurological applications. It incorporates multifocal collimators and a dedicated reconstruction algorithm, promising a reduction in acquisition duration, and an improvement in patient comfort by allowing detectors to be positioned further away from the head. The aim of this study is to investigate the performance of SZ-HRX system compared to LEHR collimators.

Methods

A striatal phantom with 5 compartments (putamens, caudates, and background) was filled with different concentrations of ¹²³I-ioflupane to simulate various clinical situations. Tomographic acquisitions were performed on each LEHR and SZ-HRX system. The summation of dynamically acquired projections allowed testing different acquisition durations with the SZ-HRX collimator. Sensitivity and Spatial resolution were assessed and compared. The data were reconstructed according to EANM recommendations. Contrast-to-noise ratio (CNR), striatum-to-background ratio (SBR), coefficient of variation (CV), and normalized asymmetry index (NAI) were calculated for both systems and compared to the LEHR acquisition. To estimate the shortest SZ-HRX acquisition duration, a linear regression of all quantitative results were calculated between the two systems.

Results

SZ-HRX collimation had superior performance characteristics than LEHR, with relative changes in CNR, CV, and SBR of + 99%, -28% and + 42% respectively, without any decrease in spatial resolution or change in asymmetry index. SZ-HRX system seems to be at least as good as LEHR system, up to 40% of scan time reduction.

Conclusion

SZ-HRX collimation showed superior performance characteristics to LEHR collimation in the study of ¹²³I-ioflupane filled striatal phantom, enabling shorter acquisitions.

Background

Single Photon Emission Computed Tomography with Computed Tomography (SPECT-CT) imaging using ¹²³I-ioflupane is widely used in the in vivo assessment of parkinsonian syndromes [1–4]. ¹²³I-ioflupane binds selectively to striatal pre-synaptic dopaminergic receptors allowing analysis of nigrostriatal dopaminergic denervation. Striatal dopaminergic denervation can be found in idiopathic Parkinson’s disease, but also in atypical Parkinsonian syndromes, including Lewy body dementia, progressive supranuclear palsy, cortico-basal dementia and multi-system atrophy [1, 5]. It can therefore be used to differentiate parkinsonian disorders with and without dopaminergic denervation (essential tremor and iatrogenic parkinsonian syndromes), in patients with an unusual history or clinical picture, or a poor response to levodopa treatment [6–10].

This imaging test can be performed with several parallel collimators, such as Low-Energy, High- (or Ultra-High) Resolution parallel-hole collimators (LEHR or LEUHR), Low-Energy Low Penetration (LELP), medium energy (ME), or dedicated brain systems, such as fan beam or multi pinhole collimators [11]. Striatal SPECT requires at least 1.5 million counts, resulting in a long acquisition duration for LEHR systems (20 to 40 min according to the amount of administrated ¹²³I-ioflupane). Moreover, rotational radius must be as small as possible, because the spatial resolution depends on the source-detector distance. These two parameters (acquisition duration and distance between detectors and head) can be limiting, especially for patients with severe tremors or claustrophobia [12]. In contrast, because they focus on the patient’s head, fan-beam collimators can offer better comfort because they allow a larger radius and shorter acquisitions.

The use of non-conventional geometry collimators is already widely used in cardiology [13–15]. Fan-beam collimation geometry centered on the heart, offers 3 to 4 times higher sensitivity compared with LEHR collimators, with no significant impact on other detection performances [16–18]. This significant improvement in sensitivity enables about 50% shorter acquisition duration, or a reduction of around 37.5% in effective dose, while maintaining comparable performances [19, 20]. The use of non parallel-hole collimators has also been investigated in neurology [21–24]. A study carried out on striatal phantoms in 2017 indicated that the use of a cardiac dedicated Cadmium-Zinc-Telluride (CZT) gamma camera, equipped with multi pinhole collimators, provided 2 to 3 times higher quantification on striatal ¹²³I-ioflupane uptake, enabling shorter imaging times and/or a lower injected dose [25]. It should be noted that this improvement is probably due both to the difference in detection geometry and in detector type. Another study, designed to create a reference database for ultra-high-resolution fan-beam collimators, showed a significant improvement in Specific Binding Ratio (SBR) compared with parallel-hole collimators, with a particular 1.2 times higher SBR on striatal phantoms [26].

Smart-Zoom High Resolution and eXtended magnification volume (SZ-HRX) package has recently been released by Siemens Healthcare (Erlangen, Germany). SZ-HRX is a specific multifocal fanbeam collimator designed for neurology. SMARTZOOM HRX differs from the cardiac SMARTZOOM collimator by utilizing smaller 1.5-mm hexagonal hole diameter and adapted focal lengths to accommodate a larger and more homogenous area of interest. A dedicated reconstruction algorithm has been developed to account for the geometry of the collimator in the reconstruction [27]. It promises a 20% reduction in acquisition duration compared to LEHR for neuroimaging with ^99mTc and an improvement of patient comfort (the detectors being located further away), without reducing the examination quality [28].

The SZ-HRX collimation has recently been evaluated on phantoms, in both ^99mTechnetium (^99mTc) and ¹²³I-ioflupane [29, 30]. This system showed equivalent image quality for ^99mTc, and higher image quality for ¹²³I-ioflupane, compared with LEHR collimators. However, these two studies used a Hoffman phantom, which simulates cerebral perfusion (by simulating cerebral cortex and basal ganglia uptake), rather than a striatal phantom, which corresponds more closely to in vivo ¹²³I-ioflupane uptake.

The aim of our study is therefore to investigate the performance of SZ-HRX system compared to parallel-hole collimators for ¹²³I-ioflupane SPECT-CT, using a striatal phantom, then to consider the optimization opportunities in terms of acquisition duration for SZ-HRX system.

Materials and methods

Experimentations on phantoms

An anthropomorphic striatal phantom Alderson RSD (Radiological Support Devices, Long Beach, CA) was used to simulate clinical situations. This phantom (Fig. 1) contains symmetrical fillable inserts for each putamen, caudate nucleus and a fillable insert for the rest of the brain. These inserts are surrounded by structures that mimic the skull and peri-cranial soft tissues to reproduce cranial anatomy.

Fig. 1.

Alderson striatal phantom, and 3d-printed insert equipped with line-sources for spatial resolution measurement

The volumes of the various structures are as follows: Right putamen 6 mL, left putamen 6.1 mL, right caudate nucleus 5 mL, left caudate nucleus 4.9 mL, background 1250 mL. It was initially chosen to fill the striatum with 40 kBq/mL of ¹²³I-iofluplane (Curium^®), as already performed in several similar studies [25, 26, 31].

As normal striatal fixation progressively decreases with age, two striatal/background contrasts were made to mimic the striatal fixation of healthy younger and older subjects corresponding to the two extrema of the European Normal Control Database of DaTSCAN (ENC-DAT) (20–80 years) [32] : 10:1 (i.e. 4 kBq/mL background) and 5:1 (i.e. 8 kBq/mL background). To simulate parkinsonian patients, we have modified the concentrations in putamens and caudate nuclei inserts as shown in Table 1. All activity concentrations were calibrated at the acquisition time.

Table 1.

Concentrations of ¹²³I-ioflupane used for each experimentation

Clinical situations	Background (kBq/mL)	Right caudate (kBq/mL)	Right putamen (kBq/mL)	Left caudate (kBq/mL)	Left putamen (kBq/mL)
Healthy (10:1)	4	40	40	40	40
Healthy (5:1)	8	40	40	40	40
Right striatum 3:1	8	24	24	40	40
Right putamen 1:1	8	40	8	40	40
Right putamen 3:1	8	40	24	40	40
Right putamen 4:1	8	40	32	40	40
Right caudate 1:1	8	8	40	40	40
Right caudate 3:1	8	24	40	40	40
Right caudate 4:1	8	32	40	40	40

Spatial resolution phantom

Assessing spatial resolution in clinical conditions was challenging because standard image quality phantoms are too large to be scanned at the appropriate radius for neurology SPECT with LEHR. To avoid this problem, we designed a custom-made brain insert for the striatal phantom was produced using 3D fused deposition modeling printer with polyethylene terephthalate glycol-modified filament. This phantom was a CT based copy of the original phantom, devoid of striatal structures, and equipped with two source lines of 2 mm internal diameter (Fig. 1). One in the craniocaudal axis, the other in the anteroposterior axis. The background insert was not filled with water, and the two-line sources with the same 40 kBq/mL concentration of ¹²³I-ioflupane as the striatal phantom.

¹²³I-ioflupane-derived radioactivity was measured using a dose calibrator (Capintec^® CRC15) to ensure that concentrations are accurate at the start of the first acquisition.

SPECT-CT acquisitions

SPECT/CT images were acquired in a gamma camera Symbia Intevo Bold equipped with LEHR collimation (Siemens Healthcare) (subsequently named “LEHR”), and a Symbia Pro.Specta equipped with SZ-HRX collimation (Siemens Healthcare) (subsequently named “SZ-HRX”). A total of 7 phases of 210s over a 360° scan arc and composed each by 120 projections acquired with an angular step of 3° in continuous detector motion were set to reach 1.5 million counts on LEHR system, as proposed by EANM, resulting in a 24.5-minute acquisition duration [1]. The same scan duration was set on SZ-HRX system. Data were acquired on both systems using energy windows as follows: photopeak window 159 keV ± 7.5%, lower scatter window 135 keV ± 7.5%, upper scatter window 183 keV ± 7.5%. The matrix size was 128 × 128 and the zoom was equal to 1.2, corresponding to a pixel size of 3.9 mm for both systems. The radius has been reduced as much as possible for the LEHR system (13 cm) to improve spatial resolution. For the SZ-HRX, radius was set to 26 cm which corresponds to the detection focal length. The SPECT-CT acquisition parameters are listed in Table 2.

Table 2.

SPECT-CT acquisition parameters for LEHR and SZ-HRX systems

	LEHR	SZ-HRX
Collimator hole length (mm)	24.05	40.25
Collimator hole diameters (mm)	1.11	1.5
Collimator septal thickness (mm)	0.16	0.26
Sampling angle	360°	360°
Projection (views)	120	120
Phases	7 × 210s	7 × 210s
Rotation mode	Continuous	Continuous
Acquisition duration (minutes)	24.5	24.5
Field of view (mm)	500	500
Matrix size	128 × 128	128 × 128
Photopeak energy window (keV)	159 ± 7.5%	159 ± 7.5%
Lower-Upper (keV) TEW scatter correction windows	135 ± 7.5% − 183 ± 7.5%	135 ± 7.5% − 183 ± 7.5%

On the Intevo Bold, CT acquisitions were performed immediately after the SPECT acquisition. The CT parameters were as follows: pitch 0.55, tube voltage 130 kV, effective mAs 57, FOV 500 mm, collimation 16 × 1.2 mm, and reconstruction filter h08. On the Pro.Specta, CT acquisitions were performed immediately before the SPECT acquisition. The CT parameters were as follows: pitch 0.55, tube voltage 130 kV with tin filter, effective mAs 150, FOV 500 mm, collimation 32 × 7 mm, reconstruction filter q40.

Data reconstruction

All LEHR and SZ-HRX acquisitions were reconstructed using the same parameters, according to the latest EANM recommendations [1]. Reconstructions were obtained using OSEM-based algorithms with 10 iterations x 10 subsets, CT-based attenuation correction, triple energy window scatter correction, resolution recovery correction, and 8 mm Gaussian post-reconstruction filtering. LEHR reconstructions were performed with FLASH-3D algorithm, but SZ-HRX system mandates the use of a xSPECT dedicated algorithm. The xSPECT-EM reconstruction is carried out in CT coordinates rather than SPECT coordinates, i.e. the SPECT pixel size corresponds to a multiple of the CT pixel size. We used a 512 × 512 CT matrix associated to a 500 mm field of view (FOV) that enabled us to obtain equivalent pixel size of 3.9 × 3.9 mm for a 128 × 128 SPECT matrix. CT-based attenuation correction of the multi-focal projections is enabled, and resolution recovery is performed using a measured point spread function.

Sensitivity

Pro.Specta and Intevo share the same geometry in the detection channel after collimators. They have the same NaI(Tl) crystal (59.1 × 44.5 × 0.95 cm) and photomultiplier configuration. To ensure there is no significant difference in the detection channels of the two systems, we conducted a sensitivity measurement using the same collimator (LEHR). We acquired a 300-second static image of a point source with 6.5 MBq of ^99mTc. The total number of counts was measured, decay-corrected, and averaged between the two detectors for both systems to obtain the sensitivity in counts/s/MBq.

Data analysis

System spatial resolution

The data obtained on both systems were displayed in the sagittal section enabling the 2 linear sources to be visualized. Sections perpendicular to the linear sources were selected. We plotted averaged profiles over several slices in both image directions. A Gaussian fit was applied to these two profiles for both SPECT-CT systems. Spatial resolution is represented by the Full Width at Half Maximum (FWHM) of the Gaussian fit. This method was implemented using an in-house script developed in the Python programming language.

Semi-quantitative analysis

Striata, putamens and caudate nuclei were manually defined on a submillimeter reference CT so that the VOIs corresponds to the wall of the true structure. A 5 cm-diameter spherical VOI was defined in the occipital area to measure the background as recommended [33]. For the assessment of non-uniformity, we expanded the background VOI to cover the entire cortex, while avoiding the edges (i.e. the striatum and skull), to remain within a homogeneous region and to minimize partial volume effects. Figure 2 shows the different ROIs in axial, sagittal and frontal views. The CTs from the different experiments were then rigidly registered with our reference CT to propagate the same VOIs over each situation. For each side (i.e. right and left), and each structure of interest (i.e. putamen, caudate nucleus, and striatum), Contrast over Noise Ratio (CNR) and Specific Binding Ratio (SBR) were calculated as follows:

Fig. 2.

Regions of interests used for quantitative analysis (cortex VOI in green, occipital VOI in blue, putamens VOI in magenta and caudate nuclei in turquoise)

C_mean(S) the mean counts of the striatal structures (putamens, caudate nuclei or striata) Volume of Interest (VOI).

C_mean(BKG) the mean counts in the background VOI.

C_SD(BKG) the standard deviation in the background VOI.

In healthy clinical situations, only the characteristics of the 2 whole striata were considered. In pathological clinical situations, involving a different ratio of one of the structures, we considered the characteristics of this/these structure(s), and of the striatum concerned. For each right/left duo of VOI, the Normalized Asymmetry Index (NAI) was extracted.

C_mean(LS) the mean counts in the left striata.

C_mean(LR) the mean counts in the right striata.

In healthy clinical situations, and the clinical situation involving the entire right striatum (right striatum 3:1), only the striatal NAI was considered. In other clinical situations, involving asymmetry of one of the structures (i.e. putamen or caudate nucleus), the NAI involving that structure and the striatal NAI were considered. The coefficient of variation (CV) was calculated in the occipital VOI as follows:

Non-uniformity (NU) was calculated in the cortex VOI as follows:

C_max and C_min the maximum and minimum count value within the cortex VOI.

These extracted indices were then compared between LEHR and SZ-HRX, for each clinical situation, except for NU, where we considered only the 8 kBq/ml background situations.

The relative differences in CNR, SBR, CV, and NU were compared between the two systems for each clinical situation. NAI was compared between the two systems in absolute values and using a Bland-Altman test (mean bias, 95% limits of agreement: mean difference +/- 1.96 SD), the presence of any systematic bias or proportional error was evaluated visually.

Acquisition duration optimization

SZ-HRX data were then reconstructed gradually decreasing the number of phases used, to simulate 21-min, 17.5-min, 14-min, 10.5-min and 7-min acquisitions. These reconstructions were then compared with the 24.5-min LEHR acquisition as a reference, using the indices mentioned above (SBR, NAI, CV, CNR). The minimum acquisition duration with SZ-HRX that would guarantee image quality comparable to that obtained with LEHR was assessed by comparing semi-quantitative indices.

Statistical analysis

SZ-HRX data for different acquisition durations were compared to the reference measurement on the LEHR system. For SBR and CNR the measurements taken in the different VOIs and for all clinical situations on the SZ-HRX system were plotted against those measured on the LEHR system (Striatum, caudate and putamen, left and right, eight contrasts, n = 48). The linear regression coefficient was then extracted (intercept set to zero) to compare the two systems, p-values were calculated (regression was considered significant for p < 0.05). A coefficient greater than one implies a better result for the SZ-HRX system. Median, minimal, maximal CV, and SBR measured with the 8 kBq/ml background were averaged for each number of frame and compare to the LEHR reference. All statistical analyses were performed using Excel (Microsoft©, Redmond, Washington, USA) software.

Results

Sensitivity

The relative difference between the two systems in terms of sensitivity was 5%. We measured a sensitivity of 223 cts/s/MBq on the Pro.Specta and 235 cts/s/MBq on Intevo.

Systems’ spatial resolution

Spatial resolutions in X (anteroposterior) and Y (craniocaudal) were measured. LEHR system showed spatial resolutions of 9.9 mm in X and 10.9 mm in Y. SZ-HRX system showed spatial resolutions of 10.2 mm in X and 10.2 mm in Y respectively. These values are roughly equivalent for the two systems.

Features comparison between the two systems

The results obtained after extraction of CNR, SBR CV and NU from LEHR and SZ-HRX acquisitions using the longest acquisition duration (24.5 min), reconstructed using EANM parameters, are presented in Table 3. CNR and SBR are consistently improved with the SZ-HRX system, showing average increases of 98% and 41%, respectively. Similar enhancements are observed in NU (with relative differences ranging from − 14% to -30%) and CV, especially in low concentration conditions such as the 10:1 contrast scenario with a background concentration of 4 kBq/mL, where CV decreases by 41%. These results were consistent with the visual comparison of images extracted from these acquisitions. Some of these images are shown in Fig. 3. Visually, the background appeared less noisy, and the striata’s shape was more clearly defined for the SZ-HRX system, particularly for the clinical situations “right striatum 3:1” and “right putamen 3:1”.

Table 3.

Features extracted for LEHR and SZ-HRX systems. Reldif: relative difference, RS: right striatum, LS: left striatum, RC: right caudate, RP: right caudate. ‘/’ indicates results that are not applicable, such as relative differences when the contrast value is zero or when the whole striatal contrast cannot be reported due to differing concentrations in the caudate and putamen

	VOI	CNR			SBR				CV			NU
		LEHR	HRX	Reldif	LEHR	HRX	Reldif	True	LEHR	HRX	Reldif	LEHR	HRX	Reldif
Healthy (10:1)	RS	14.2	29.0	104.5%	2.7	3.8	40.1%	9	21	13	-41%	69	48	-30%
	LS	13.7	28.0	105.0%	2.6	3.7	40.1%	9
Healthy (5:1)	RS	8.1	14.3	76.5%	1.2	1.6	33.9%	4	17	12	-27%	62	44	-28%
	LS	8.4	13.9	65.8%	1.3	1.6	25.4%	4
Right striatum 3:1	RS	3.5	11.5	228.8%	0.5	0.9	80.8%	2	16	11	-32%	55	47	-14%
	RC	4.1	12.4	203.9%	0.6	1.0	65.6%	2
	RP	3.1	10.8	253.9%	0.5	0.9	91.3%	2
Right putamen 1:1	RS	3.2	3.9	21.6%	0.4	0.5	18.8%	/	15	11	-25%	50	39	-22%
	RP	0.2	0.0	-105.7%	0.0	0.0	/	0
Right putamen 3:1	RS	3.1	7.2	129.9%	0.6	0.8	33.3%	/	19	15	-18%	64	53	-17%
	RP	2.3	5.6	146.9%	0.4	0.6	43.9%	2
Right putamen 4:1	RS	7.1	8.9	26.3%	1.1	1.3	15.3%	/	15	12	-22%	52	44	-16%
	RP	6.4	9.1	41.0%	1.0	1.3	28.7%	3
Right caudate 1:1	RS	4.2	9.7	129.4%	0.7	1.1	63.2%	/	15	11	-26%	57	43	-23%
	RC	1.3	3.1	135.6%	0.2	0.4	/	0
Right caudate 3:1	RS	7.4	9.7	32.5%	1.0	1.0	6.2%	/	16	12	-26%	56	41	-28%
	RC	5.8	8.4	44.2%	0.8	0.9	17.1%	2
Right caudate 4:1	RS	8.3	18.8	126.3%	1.2	1.7	48.7%	/	19	12	-35%	65	45	-30%
	RC	6.8	14.2	108.5%	1.0	1.3	37.5%	3

Fig. 3.

Axial slices obtained on the LEHR and SZ-HRX systems for different situations. A: “Healthy 10:1”. B: “Right striatum 3:1”. C: “Right putamen 3:1”

The NAI features extracted from the two systems are presented in Table 4 and in Fig. 4.

Table 4.

NAI features for LEHR and SZ-HRX systems

	VOI	LEHR	SZ-HRX
Healthy (10:1)	Striata	-0.03	-0.03
Healthy (5:1)	Striata	0.02	-0.02
Right striatum 3:1	Striata	0.38	0.32
Right putamen 1:1	Striata	0.33	0.45
	Putamens	0.65	0.83
Right putamen 3:1	Striata	0.24	0.26
	Putamens	0.33	0.34
Right putamen 4:1	Striata	0.03	0.10
	Putamens	0.08	0.10
Right caudate 1:1	Striata	0.21	0.19
	Caudates	0.55	0.62
Right caudate 3:1	Striata	0.09	0.16
	Caudates	0.22	0.29
Right caudate 4:1	Striata	0.03	0.00
	Caudates	0.12	0.19

Fig. 4.

Bland-Altman plot of NAI absolute difference

The mean value of the differences (bias) was calculated to be 0.04. The two systems appeared to give comparable results in terms of NAI with a limit of agreement (95%) between − 0.09 and + 0.16. The linear regression coefficient was estimated at 1.16, supporting this hypothesis.

Acquisition duration optimization

An example of two linear regressions is shown in Fig. 5. Figure 6 shows the regression coefficients of the extracted CNR and SBR, as well as the mean values [min; max] of NU and CV, depending on acquisition time.

Fig. 5.

Graphical representation of SZ-HRX CNR as a function of LEHR CNR, and the corresponding linear regression, for 24.5-min and 7-min SZ-HRX acquisition durations. p-values < 0.0001

Fig. 6.

First row: Linear regression coefficients of CNR (A) and SBR (B) depending on acquisition time. All p-values < 0.0001 for regressions; n = 48. Second row: Mean values [min, max] of NU (C) and CV (D), depending on acquisition time. Yellow lines correspond to the 24.5 min LEHR reference value

We noted a gradual decline in the linear regression coefficient for CNR, but it consistently remained above 1, even at the shortest acquisition duration of 7 min. For 14-min acquisitions the regression coefficient reached a value of 1.52. Finally, the improvement of SBR with SZ-HRX was stable, whatever the acquisition duration, with a regression coefficient constantly close to 1.3. The mean values of CV and NU are better for all acquisitions with a duration of 14 min or more. These results were consistent with the visual assessment of images extracted from these acquisitions. Some of these images are shown in Fig. 7.

Fig. 7.

Axial slices obtained on the SZ-HRX system by progressively decreasing the acquisition duration, in comparison with LEHR axial slices, for different clinical situations. A: “Right striatum 3:1”. B: “Right caudate 1:1”. C. “Right putamen 1:1”

Discussion

Although several previous studies have evaluated the performance of the SZ-HRX collimator [34] or assessed image quality in brain SPECT using 99mTc and Hoffman phantoms [29, 35], to the best of our knowledge, no study has investigated this system in the context of 123I-ioflupane SPECT using a striatal phantom. The results of this study showed that SZ-HRX system outperformed LEHR for the same acquisition duration. SZ-HRX offered an improvement in image quality. The SZ-HRX system is more sensitive, thus CV and non-uniformity are lower (-28% and − 22% respectively). The contrast is also higher thanks to a lower septal penetration (SZ-HRX septal thickness: 0.26 mm vs. LEHR septal thickness: 0.16 mm), as shown by the SBR (+ 42%). Combined, the higher sensitivity and the higher contrast leads to a significantly higher CNR (+ 99%). For clinical situation “right putamen 1:1”, the contrast is supposed to be equal to 0, so the relative difference is of no interest and has not been included in the results. For the right caudate 1:1 situation, SBR and CNR were slightly above zero, probably due to the partial volume effect and putamen spill-out. This performance improvement was achieved while maintaining the same spatial resolution (~ 10 mm), thanks to the collimator specifically designed for this purpose [27] and without any significant impact on NAI. The reduction in examination duration is greater than expected. Siemens reported a decrease of about 20% for neuro acquisition using ^99mTc [28]. This significant difference could be attributed to a reduction in scatter fraction and septal penetration with the SZ-HRX collimator for ¹²³I SPECT-CT. The ¹²³I has a complex decay scheme with a photopeak at 159 keV (83%) associated with gamma photons of energies higher than 300 keV (≤ 3%). Due to these high-energy gamma photons, image quality may be affected depending on choice of the collimators [36, 37]. Striatal SPECT imaging, as described in the SNM and EANM recommendations, is usually performed with LEHR or LEUHR collimators [1, 2]. These collimators are also adapted to the 159 keV photopeak and offer sufficient spatial resolution, ~ 7.4 mm with 10 cm radius [37], to image small striata structures, despite the contribution of scatter from the septal penetration of high-energy gammas [36, 37]. The use of Medium Energy General Purpose (MEGP) parallel-hole collimators could be of interest to reduce this septal penetration with equivalent or slightly improved sensitivity [37]. However, the loss in resolution compared with LEHR [37, 38] particularly for imaging small structures such as the tail of caudate nuclei (~ 1 cm diameter) could affect diagnosis [26]. Consequently, clinical databases built to assist physicians in their diagnosis are usually acquired with low energy collimators [32, 39, 40].

Our results (SZ-HRX higher sensitivity and contrast to noise ratio) are concordant with those already published concerning fan-beam collimation in nuclear cardiology compared to parallel collimators [17, 18]. Nuclear cardiology and dopaminergic denervation SPECT have in common the study of a restricted area (left ventricle for cardiology and striata for neurology). The use of a fan-beam collimation seems therefore more appropriate than parallel-beam collimation, providing a clear gain in sensitivity for the restricted area studied [23, 24]. These results are also consistent with the studies carried out on by Shibutani et al. investigating the technical performance of SZ-HRX on a cerebral perfusion phantom filled with ^99mTc and ¹²³I-ioflupane [29, 30]. Our results are also in line with those of the recently published study on the creation of a reference database with fan-beam collimators, finding a significant improvement in SBR with this type of collimator, compared with parallel-hole collimator [26]. We note that the measured SBR was 2 to 4 times lower than true SBR. These results are well in line with the literature. Bailly et al. found SBR values between 2.66 and 2.91 for 8:1 true ratios [41], Yamamoto et al. found also SBR values between 1.8 and 2.9 for the same true ratio [25]. These results can be explained by the significant partial volume effect. This effect will substantially reduce the measured signal in small structures such as putamen and caudate nuclei. On the other hand, scatter and septal penetration are responsible for an artificial increase in the background activity [42, 43]. Finally, contrast, and thus SBR, are higher for SZ-HRX when compared to LEHR, thanks to the lower septal penetration.

The SZ-HRX CV remained at least equivalent to LEHR CV up to 14 min of acquisition, while maintaining a better CNR. Moreover, the spatial resolutions were equivalent between the two systems. Nevertheless, the rotational radius was shorter for LEHR (13 cm) than for SZ-HRX (26 cm). So rotational radius of SZ-HRX seems to be more appropriate for patients with short-neck or over weighted.

Comparing SZ-HRX to LEHR, from 24.5 min to 14 min, CV and Non uniformity are at least better or equivalent. This suggests that acquisition durations could be reduced by up to 40% in clinical practice, improving patient comfort and minimizing the risk of movements during the procedure. This reduction also enables an increased number of examinations.

This clear superior image quality of the SZ-RHX collimation system also allows to consider a reduction in injected doses (without any modification of acquisition duration), although this was not investigated in our study.

Our study has several limitations. Acquisitions were performed on anthropomorphic phantoms which, although controlled and complying with EANM recommendations, does not necessarily reflect real clinical conditions. In clinical situations, variables such as inhomogeneous radiotracer distribution or interpatient variations could influence the results. The absence of clinical hazards, such as patient movement, is another limitation. In practice, such movements can degrade image quality, and their impact has not been assessed on the SZ-HRX system. Another limitation is comparing two collimators on two different systems (Intevo vs. Pro.Specta), Intevo gamma camera being 8 years older. Nevertheless, no significant difference in terms of sensitivity was measured between the two cameras (223 cts/s/MBq for Pro.Specta vs. 235 cts/s/MBq for intevo). Acquisitions on both systems were carried out immediately one after the other without any preferential order, resulting in an inter-acquisition delay of 30 min, corresponding to only 2.5% activity decay. Furthermore, the study was conducted exclusively using the EANM-recommended reconstruction parameters, which may not reflect the optimal capabilities of either system. Additional investigations are warranted to determine the best trade-off between spatial resolution and image noise across varying signal levels.

Finally, it should be noted that the final images obtained with the SZ-HRX system, which differ from that obtained with the LEHR system, may initially confuse clinicians. A learning phase is therefore required. Thus, it appears necessary to complete this study with an investigation based on actual clinical acquisitions. An interesting approach would be to virtually reduce acquisition duration for patients’ data. This would enable us to evaluate the performance of SZ-HRX system under conditions closer to daily clinical practice, and to estimate more precisely the impact of these parameters on image quality and calculated indices.

Conclusion

Our study demonstrated that the performance of SZ-HRX collimation is superior to LEHR system on phantom study. These performances remained at least equivalent to LEHR collimation, up to 40% reduction time. Clinical studies are needed to confirm these results, particularly concerning the reduction of acquisition time.

Author contributions

SG produced the 3d printed insert. MP, DB, AR, CR and KK performed the phantom manipulations, acquisitions and reconstructions. MP, CR, KK, AR and DB extracted features from the acquired data. MP, DB, KK and SQ were a major contributor in writing the manuscript. All authors read and approved the final manuscript.

Funding

No funding was provided for this study.

Data availability

The datasets used during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.