Absolute Quantification of Bone Scintigraphy for the Longitudinal Monitoring of Vertebral Fractures With a High-speed Whole-body CZT-SPECT/CT System


 This study aimed to determine whether absolute quantification obtained with a high-speed CZT SPECT/CT system provides consistent results on phantom and in bone scintigraphy-based longitudinal monitoring of patients with vertebral fractures and/or fracture cascades.Materials and Methods. SPECT images were reconstructed as recommended for clinical routine. Both image quality and accuracy of measured activity were evaluated using an IEC body phantom. The optimized reconstruction process was applied to routine ≤ 15 min 99mTc-HDP SPECT spine recordings, which had been previously acquired from 25 patients (74±12 years old) at both early (1.3±1.1 months) and late (5.2±2.3 months) stage assessments of an acute vertebral fracture of a traumatic and/or osteoporotic origin.Results. A SPECT reconstruction with 32 equivalent iterations was used based on high levels of foci detectability in spheres as small as 0.6 mL in volume, and accuracy of measured activity, although the latter was affected by partial volume effect for spheres ≤ 5.8 mL. SUVmax from patients’ intact T1 vertebrae, used as a reference, remained stable between 1st and 2nd SPECT recordings (5.7±1.1 vs. 5.8±1.1, p=0.76). SUVmax from the initially fractured vertebrae were 3-fold higher on 1st SPECT (21.0±8.5, p<0.001 vs. T1 SUVmax) and markedly decreased on 2nd SPECT (11.2±4.2, p<0.001 vs. 1st SPECT). Inverse changes in SUVmax were documented for newly compacted fractures apparent on the 2nd SPECT (1st SPECT: 7.4±2.0 vs. 2nd SPECT: 21.8±10.3, p=0.002) together with an increased bone density on CT (in mean Hounsfield Units: 121±28 vs.190±39, p=0.003). Rate of SUVmax > 7.5 was 98% (56/57) for measurements obtained in vertebrae fractured in the preceding 7-months, whereas this rate was only 4% for the reference intact vertebrae (2/50).Conclusion. High-speed recordings with this CZT-SPECT/CT system provide reliable SUV measurements that may attest to longitudinal changes in vertebral bone metabolism and especially changes related to fracture healing or recurrence.


Introduction
Several recently published studies have given evidence of a signi cant contribution of absolute quanti cation in the interpretation of bone SPECT images, especially for the monitoring of longitudinal changes after treatment [1][2][3], and also for the characterization of metastasis [4][5][6], osteomyelitis, osteonecrosis [7], and certain degenerative bone lesions [8,9]. To the best of our knowledge, however, none of these previous quantitative bone studies were planned with new high-speed CZT cameras, nor conducted in patients with bone fractures.
Bone scintigraphy is increasingly used in the pre-therapeutic assessment and monitoring of patients with compression and/or traumatism of vertebrae at a possible multi-site level [10][11][12][13], with the main objective being to select the painful lesions for which a vertebral augmentation procedure, i.e., vertebroplasty or kyphoplasty, would be bene cial. Such fractures may occur in complex settings, within ankylosed thoracolumbar spinal segments [14] and especially, during a downward spiral of osteoporosisrelated recurrent fractures (known as the vertebral fracture cascade) that may be favored by a vertebral augmentation [15]. However, none of the previously documented bone scintigraphy studies of patients with compression and/or fracture was analyzed using an objective absolute method for quanti cation, nor conducted with a new high-speed CZT camera. Such cameras have the advantages of not only enhancing image quality, but also of reducing the recording time [16,17], a property that is particularly advantageous in patients with painful fractures.
The whole-body Veriton® camera (Spectrum Dynamics Medical, Caesarea, Israel) combines CZT detectors with an original 360° ring-con guration geometry that allows the detectors to be positioned very close to the patient, a con guration that is likely to maximize both count sensitivity and image quality [17][18][19]. With this camera, it has already been demonstrated that a whole-body 3-dimensional SPECT recording may be obtained in no more than 20 minutes for bone 99m Tc-scintigraphy [17], as well as for lutetium-177 therapeutic imaging [18]. Absolute quanti cation has also been developed with this camera and whilst the quanti cation of lutetium-177 has been investigated for dosimetry [18], as yet there has not been any assessment for routine bone scintigraphy.
This study aimed to determine whether absolute quanti cation obtained with this high-speed CZT SPECT/CT system provides consistent results on phantom and in bone scintigraphy-based longitudinal monitoring of patients with vertebral fractures and/or fracture cascades.

Materials And Methods
All images were acquired on the Veriton® hybrid system comprising the high-sensitivity 360° CZT-camera and 64-row detector CT. The present study involves the analysis of a routine protocol for bone SPECT/CT recording applied: (i) on a phantom for optimizing the image reconstruction process, with respect to image quality and absolute quanti cation measurement, and (ii) in patients with acutely fractured vertebrae to check for the consistency of SUV measurements, and especially for changes over time in fractured vs. non-fractured vertebrae.
The study was approved on January 22nd, 2021 by the Ethics Committee of the CHRU of Nancy (reference number 297). This research complied with the principles of the Declaration of Helsinki. Informed consent was obtained from all individuals included in the study.
Patients' bone SPECT/CT recordings Twenty-ve consecutive patients were retrospectively selected on the basis of prior imaging on the Veriton® hybrid system, comprising: a 1st SPECT/CT for pre-therapeutic assessment prior to a possible vertebral augmentation procedure after an acute vertebral fracture of traumatic and/or osteoporotic origin, and a 2nd SPECT/CT at a later stage, mainly due to suspicion of new vertebral fractures (vertebral fracture cascade). The fracture date was considered to be that of the acute onset or worsening of pain.
SPECT/CT recordings commenced 3 to 4 hours after the injection of approximately 550 MBq of 99m Tchydroxydiphosphonate (HDP). CT was recorded rst, with the following parameters: 120 kV, 150 mAs modulation, pitch of 0.8, slice thickness of 1.25 mm, increment 1.25 mm, iterative reconstruction, bone lter. SPECT was recorded thereafter with an energy window of 140 keV ± 7.5% during a total recording time of 10 to 15 mins (2 to 3 bed positions of 5 mins each).

Phantom experiments
The CZT-camera was previously calibrated with the clinical routine reconstruction parameters and by using a cylinder lled-in with a homogeneous solution of 99m Tc, as previously described [20].

Image reconstructions and analysis
SPECT images of the IEC phantom were reconstructed using two OSEM iterative methods both being recommended by the manufacturer for optimizing the image quality of bone SPECT. The rst one, which provides the best spatial resolution but requires high level of count statistics, involves 8 subsets, 8 iterations and a kernel inter-iteration lter with a factor 0.125. The second one, which favors contrast-tonoise ratio and is more adapted to lower count statistics, involves 8 subsets, 4 iterations and a factor 0.2 for a kernel inter-iteration lter. The remaining parameters were xed as follows: post-reconstruction median lter (3 x 3 x 3 voxels); and corrections for scatter, attenuation with the CT attenuation map, resolution recovery with a point-spread function, and partial volume with a dedicated vendor-proprietary algorithm (PVC) that uses the anatomical information provided by the CT scan. For the phantom images, this PVC algorithm was applied on a CT scan for which the spheres' density was xed at 700 Houns eld Units (HU), i.e., an intermediate level between the HU from spongious and cortical bone. All SPECT images were displayed through cubic voxels of 2.46 x 2.46 x 2.46 mm 3 .
MIM software (MIM Software Inc, Cleveland, OH, USA) was used to provide paired display of SPECT and CT images. Volumes-of-interest (VOIs) were drawn manually, encompassing: 1) each of the 6 spheres of the IEC phantom, and 2) the bodies of targeted fractured vertebrae and the reference T1 vertebra on each patient's SPECT/CT exam. The targeted fractured vertebrae were de ned as those showing the highest SUV values on SPECT recordings, together with typical CT signs of bone compression and/or fracture. In addition, the T1 vertebra was used as a control reference since this vertebra is at very low risk of osteoporotic fractures (21), and no fracture or severe arthrosis lesion was detected on T1 vertebra on any of our SPECT/CT recordings. Additionally, cement volume was excluded from the VOIs of all vertebrae treated by cementoplasty.
Mean HU were obtained from the CT vertebral VOIs; maximal SUV (SUVmax) from the SPECT vertebral VOIs, and a maximal activity concentration expressed in % of the actual concentrations was computed with the SPECT phantom VOIs. SUV were determined with the following formula: A VOI is the activity concentration within a VOI in kBq/ml, A INJ is the decay-corrected injected activity in kBq, and BW is the patient body weight in g.
A contrast-to-noise ratio (C/N) was computed on each SPECT-image reconstruction of the IEC phantom, with a spherical VOI of 37 mm diameter corresponding to the diameter of the largest sphere. This VOI was placed at the center of the largest sphere to measure mean sphere activity ("Mean sphere" in counts per second) and corresponding standard deviation ("SD sphere"). A 5 cm translation of the VOI was subsequently applied to measure mean background activity ("Mean BKG" in counts per second) and corresponding standard deviation ("SD BKG"

Phantom experiments
Mean background activity ranged between 98 and 99% of the actual activity concentration when measured on images provided by each of the two reconstructed methods.
In addition, as detailed in Fig. 1, the maximal activity concentrations measured on phantom spheres exhibited an increase according to the sphere size and was clearly underestimated for the smallest spheres with volume ≤ 5.6 mL. This underestimation was somewhat more marked for the reconstruction favoring signal-to-noise ratio at the expense of spatial resolution, but this latter had the advantage of improving image quality and facilitating image analysis on both phantom and patient bone SPECT images (Fig. 1). Thereby, the reconstruction method favoring contrast-to-noise ratio was considered to provide the best compromise and robustness for routine image interpretation and it was selected for the additional analyses (Fig. 1).

Characteristics of the study population
At the time of the 1st SPECT/CT, mean age of the study patients was 73 ± 12 years (51 to 90 years), 11 (44%) were women, and 15 (60%) had a previous history of vertebral compaction fracture and could thus be considered to be affected by a vertebral fracture cascade. An at least partially osteoporotic mechanism was suspected for all these acute fractures except in 4 patients for whom this mechanism was presumed purely traumatic.
The 1st SPECT/CT was performed at a mean of 1.3 ± 1.1 months after the acute fracture date. It was prescribed as part of a pre-therapeutic evaluation based on which 12 patients were referred for cementoplasty and 13 referred for conservative treatment involving a back brace.
The 2nd SPECT/CT was performed at a mean of 5.2 ± 2.3 months after the initial acute fracture episode and it was prescribed for the persistence or resurgence of back pain in all patients. A new episode of acute vertebral fracture was nally diagnosed in 12 patients for whom the mean delay-time between symptom resurgence or aggravation and the 2nd SPECT was 4.9 ± 4.1 weeks.
Overall a total of 37 vertebrae were considered as acutely fractured, based on evocative symptoms and analysis of both the 1st and 2nd SPECT/CT. In order of decreasing frequency, these vertebral fractures were located on L2 in 7 cases, L1 and T12 in 5 each, T11 and L4 in 4 each, T7 in 3 cases, T4, T10, and L3 in 2 each, and T5, T6 and L5 in one case each.
Broad differences in SUVmax were also documented on the 12 vertebrae showing a new acute fracture between the 1st and 2nd SPECT/CT (1st SPECT: 7.4 ± 2.0 vs. 2nd SPECT: 21.8 ± 10.3, p = 0.002). These differences were associated with an increase in mean HU and thus, in bone density, between the 1st and the 2nd SPECT/CT (121 ± 28 HU vs.190 ± 39 HU, p = 0.003). By contrast, mean HU were stable between the 1st and 2nd CT for T1 as well as for the initially fractured vertebrae (Table 1).
All SUVmax measured on T1 and on the acutely fractured vertebrae are displayed according to the delaytime from the fracture date in Fig. 2. This Figure shows a decline over time in the SUVmax from fractured vertebrae, but not for the SUVmax from T1. It may additionally be observed that SUVmax was > 7.5 in as many as 98% (56/57) of measurements obtained on vertebrae fractured in the preceding 7-months, whereas rate of SUVmax > 7.5 was only 4% for reference intact vertebrae (2/50).
These results are illustrated in Fig. 3 by images from the same patient with color and grey scales similarly adjusted to a maximal SUV of 13 on the 1st and 2nd SPECT. This likely leads to facilitate comparison between the two SPECTs and to favor the identi cation of bone structures reaching a ≥ 7.5 SUV level (i.e., orange for the color scale and dark grey for the gray scale). In particular, the initially fractured vertebra remains easily detectable on the 2nd SPECT, despite a marked decreased uptake from the 1st SPECT.
Opposingly, an increased uptake may be easily detected between the 1st and 2nd SPECT on several vertebrae showing CT signs of new compressions on the 2nd SPECT/CT, in a typical setting of vertebral cascade fracture.

Discussion
This study shows that in conditions of image reconstruction targeting a high level of image quality, highspeed recording from this whole-body 360° CZT-SPECT/CT system provides reliable and potentially helpful SUV measurements in patients with vertebral fractures and/or compaction.
However, the poorer spatial resolution of SPECT compared with PET, and resultant higher partial volume effect, is considered as a disadvantage for SUV measurement. In analysis of the present study, only one SPECT reconstruction series was utilized with parameters favoring contrast-to-noise ratio -i.e., cubic voxels of only 2.46 mm and partial volume correction -and also, with ltering -i.e., median and kernel lters and only 32 equivalent OSEM iterations. These reconstruction parameters were those considered to provide the best compromise between image quality and the accuracy of activity measurements -i.e. high foci detectability, even for a 0.6 mL sphere volume, and with signi cant SUVmax underestimation only for spheres ≤ 5.8 mL (Fig. 1).
SUVmax was preferred here to other SUV parameters such as SUV mean and SUV peak. SUV mean was di cult to use in the absence of any precise knowledge regarding limits of the diseased bone volumes and when applied to the entire bodies of vertebrae, its ability to separate fractured from intact vertebrae was lower than that of SUVmax (see supplemental Fig. 4). In addition, a 1 cm 3 volume within which the SUV peak is commonly determined was too large, exceeding the thickness of several very attened vertebrae and the volume of the smallest phantom sphere. Finally, the well-known confusing in uence of noise level on the determination of SUVmax [23,24] was minimized here, due to our choice of reconstruction parameters favoring the contrast-to-noise ratio of the SPECT images.
Overall, our phantom results agree with the conclusions achieved by previous comparable studies planned on more conventional gamma-cameras, regarding the acceptable accuracy of SPECT SUV measurements [25,26], especially for volumes ≥ 10 mL [27] and when associated with CT-based methods enabling reduction in the partial volume effect [28][29][30]. It is of note however, that underestimation of SUVmax for spheres < 10 mL was no longer observed when utilizing alternative reconstruction parameters that favored spatial resolution to a greater extent (≥ 20 OSEM equivalent iterations, less image ltering), albeit at the cost of an increase in the noise level (results not shown).
However, the remaining partial volume effects, observed here for volumes smaller than 5 to 10 mL, constitute a limitation, considering that the diseased parts of certain compacted vertebrae may represent a smaller than 5 to 10 mL volume. Despite this limitation, however, consistent results were observed in our longitudinal study of patients with vertebral fractures and/or compactions, thereby reinforcing the potential usefulness of our method of SUV measurement. These results include the stability over time of the SUVmax from intact control reference vertebrae. The T1 vertebrae were chosen for this stability analysis because of their very low risk of osteoporotic fractures [21], and because they were visually intact on the SPECT/CT images of our all patients.
The results also demonstrate a dramatic decrease over time for the SUVmax from fractured vertebrae, as would be expected, although elevated SUVmax levels were still documented during a 7-month period (Fig. 3). Such persistence of elevated SUVmax over the longer-term is not surprising considering that a great proportion of bone scintigraphy remains visually abnormal even at one year after a vertebral compression fracture [31].
SUV were already proven helpful for documenting longitudinal changes with PET imaging and more recently with bone-SPECT [1][2][3]. However, to the best of our knowledge, the present study is the rst to use SUV for analyzing longitudinal changes in the metabolism of bone fractures. The potential usefulness of SUV measurement in this setting is best illustrated in Fig. 3 by the easy identi cation of any increase in bone metabolism, corresponding to new compaction fractures involved in a vertebral fracture cascade, when serial SPECT images are displayed with the same SUV-based scaling. Furthermore, this assessment was obtained with recording times of only 12 to 15 minutes for the entire spine. This latter property is likely advantageous in patients for whom a prolonged supine position on the camera bed is di cult to endure due to their painful fractures.
An additional observation was that as many as 98% of the fractured vertebrae exhibited SUVmax > 7.5 in the 7-month period following the acute episode, whereas this was highly unusual in the reference intact T1 vertebrae with only 4% exhibiting SUVmax > 7.5. Therefore, it may be considered that an image scaling favoring the identi cation of bone structures reaching a > 7.5 SUV level is required to easily detect most bone fractures, and especially the oldest ones.
Until now, this image scaling has routinely been done in an empirical and subjective visual way, mainly based on the % of maximal voxel activity, with high reader variability being a di culty for uniformity of SPECT diagnoses. Standardized SUV-based scaling would likely enhance bone SPECT analysis reproducibility by optimally standardizing readers' diagnostic review. It might also alleviate the di culties encountered when scaling images where bone lesions are very diffuse (i.e., a super scan) or have very high maximal activity levels. This point is illustrated in a supplemental Fig. 5 by the SPECT images from an actual clinical case where the diagnosis could be corrected with a secondary analysis, thanks to the application of an SUV-based scaling method.
In conclusion, the high-speed whole-body CZT-SPECT/CT system presented here, in association with an adapted SPECT reconstruction for bone scintigraphy images, provides high image-quality together with reliable SUV measurements, at least for structures with ≥ 5 to 10 mL volume. These SUV measurements may attest to the longitudinal changes in vertebral bone metabolism -i.e., SUV decrease during fracture healing and iSUV increase after a fracture recurrence and within a possible framework of cascade fractures. And further, more generally, these SUV measurements might also help to enhance the reproducibility and robustness of bone scintigraphy analyses.

Declarations
Funding None Ethics approval and consent to participate All procedures performed in studies involving human participants were in accordance with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. The study was approved on January 22 nd , 2021 by the Ethics Committee of the CHRU of Nancy (reference number 297). Informed consent was obtained from all individuals included in the study.

Consent for publication
Not applicable

Availability of data and material
The data that support the ndings of this study are available on request from the corresponding author (LI). Figure 1 Comparison of results provided by the 2 recommended OSEM reconstruction processes, the rst one favoring spatial resolution but requiring high counts statistics (red lines) and the second one favoring contrast-to-noise ratio (blue lines), with (A) the analyses of the evolutions of the maximal voxel activity concentration measured on the spheres of the IEC phantom and expressed relative to the actual activity concentration, (B) the contrast/noise ratio determined on the IEC phantom, and representative tomographic SPECT slices from the IEC phantom and (D) a patient bone sagittal slice passing through a fractured T4 vertebra.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.