Methodology and Performance Assessment of Zoom SPECT · 2015. 2. 26. · zoom (Z) SPECT to small organs and small animals, we explored the methodology and assessed its performance

Methodology and Performance Assessment of Zoom SPECT

Kypros Kouris, Abdelhamid Elgazzar, Raid Affana, Ezzat Higazi, Maha Awdeh, Abdelrahman Mahmoud and Hussein M. Abdel-Dayem

Department of Nuclear Medicine, Faculty of Medicine, Kuwait University and Ministry of Public Health, Safat, Kuwait

The methodology of zoom single-photon emission computed tomography (Z SPECT) imaging and its performance was assessed using two tomographic gamma cameras, systems A * and B •, fitted with high resolution collimators. The center of rotation (COR) varied linearly for zoom factors (ZF) from 1.0 to 4.0. It was more stable and reproducible in system B and its variation with ZF was significantly smaller. Z SPECT acquisitions were done using projection matrix sizes 64 x 64 and 128 x 128, 64 projections over 360 degrees and ZF equal to 1.0, 1.6 and 2.0. Point source acquisitions demonstrated the superior performance of system B. Two small phantoms were constructed (10 x 10 x 5 em each) containing cold rods and hot rods. The use of zoom improved both contrast and resolution. The Nowak reconstruction algorithm provided bet-ter contrast than the conventional filtered backprojection al-gorithm. The 2 mm hot rod was visible using system B with projection matrix 64 x 64 and ZF = 1.6. The use of zoom in SPECT leads to improved contrast and resolution; the appli-cation of Z SPECT to small organs is recommended.

Single-photon emission computed tomography (SPECT) with rotating gamma cameras is finding an increasing role in nuclear medicine (1, 2). Careful quality control is essential in order to avoid artifacts and ensure good quality studies (3). Compared to planar imaging, SPECT achieves improved localization of abnormalities and better contrast, but it ex-hibits inferior spatial resolution.

Commonly, SPECT studies are performed using a low-energy general-purpose collimator rather than a high resolu-tion collimator. It has been reported, however, that for reso-lution-limited imaging tasks, an improvement in resolution outweighs the increased noise due to the accompanying loss in sensitivity ( 4).

Motivated by the possible application of high resolution zoom (Z) SPECT to small organs and small animals, we explored the methodology and assessed its performance using phantoms. All studies were done using technetium-99m C9mTc) and a low-energy high resolution (LEHR) collimator. Two tomographic gamma camera systems were compared.

For reprints contact: Kypros Kouris, MD, Dept. of Nuclear Medicine, Faculty of Medicine, Kuwait University, PO Box 24923, Safat-13110, Kuwait.

198

The variation of planar resolution, measured at full width at half maximum (FWHM) with different zoom factors (ZF), was first studied. Then for SPECT systems, the variation of the center of rotation (COR) with ZF was assessed. Using a point source, a study was made investigating the relation between SPECT resolution FWHM, ZF and matrix size used. The relation between ZF, matrix size, spatial resolution and contrast was studied using cold and hot rod phantoms. Fi-nally, Z SPECT was compared with pinhole collimator (PH C) planar imaging for thyroid studies.

MATERIALS AND METHODS

Two tomographic gamma cameras were used for this study, system A and system B. Both gamma cameras were fitted with a LEHR collimator.

Technetium-99m point sources were prepared by drawing

2 7 3 rod diameters (mm)

50mm 100mm 12 6 8

5 10 4

FIG. 1. Design and dimensions of the cold and hot rod phantoms.

were placed vertically side by side so that the reconstructed transaxial slices would represent cross sections of the 3 x 3 rod arrangement.

Two SPECT rotation radii were used: a 17-cm radius corresponding to the gamma camera being close to the tom-ographic table in the lateral projections, and a 6-cm rotation radius using an extension centrally fixed on the edge of the tomographic table. Four acquisitions were taken for each rotation radius: acquisition matrix 64 x 64 and 128 x 128 with ZF = 1.0 and 1.6. The COR was determined for each matrix and ZF pair. Circular tomographic acquisition over 360" was used and the number of projections was 64. Using the cold and hot rod phantoms, acquisition time was such that about 70,000 counts per projection were obtained for each study.

Reconstruction for the point source and phantom studies was performed using a Ramp-Hanning filter with I cycle/em cut-off frequency (as commonly used clinically) without at-tenuation correction. In addition, point source reconstruc-tions were performed using a Ramp filter only. For system B, the Nowak reconstruction algorithm (5) was compared to the conventional filtered backprojection algorithm. For 64 X 64 acquisition studies, both the 64 x 64 and 128 x 128 recon-struction matrices were used.

Assessment of the reconstructed images was made both visually and quantitatively. For the point source reconstruc-tions, the FWHM was calculated using Gaussian fits on row and column profiles. The fits excluded the tails of the point spread functions with values less than 10% of the maximum. It is known that for a reliable FWHM measurement, there must be at least three linear samples per FWHM. This con-dition was not met when M = 64 x 64 and ZF = 1.0 (corresponding to a spatial sampling interval of6.1 mm) and the fit was dependent on the position of the point source within the pixel. In such cases, the FWHM was obtained as the average of the FWHMs computed from the row and column fits for two acquisitions with slightly different point source locations. For the phantoms, contrast and resolution measurements were made. The hot rod phantom was used for contrast measurements. A horizontal profile was drawn along the lower row of rods as shown in Figure 2. Three peaks (P) and two valleys (V) resulted from each profile. The peak/

VOLUME 18, NUMBER 3, SEPTEMBER 1990

counts

P1

V1

t Diameter: 5mm

P2

t 10mm

V2

P3

t distance 4mm

Peak/Valley ratios: P1/V1,P2/V1,P2/V2,P3/V2

FIG. 2. Horizontal profile through the lower row of hot rods with peak/valley (P/V) ratios computed.

valley ratios PI/VI, P2/VI, P2/V2, and P3jV2 were calcu-lated for each acquisition and were used in contrast compar-isons.

Z SPECT was compared with PHC planar imaging for the investigation of nodular thyroid disease in 14 patients. Both studies were done in the same day, 15 min after i.v. injection of 5 mCi (185 MBq) [99mTc] pertechnetate. PHC images were anterior and both anterior obliques for 150,000 counts. Z SPECT acquisition was performed as follows: LEHR col-limator, 128 x 128 matrix, ZF = 1.6, 180" from left to right lateral anterior rotation, 32 frames, 30 sec/frame. Tomo-graphic slices were reconstructed using a Ramp-Hanning filter with a cut-off frequency between 0.8-2.0 cycles/em, depend-ing on thyroid uptake. No attenuation correction was applied. Data were interpreted in the coronal and transaxial sections.

RESULTS

The variation of planar resolution (FWHM) with different ZFs is shown in Table I. With the 64 x 64 acquisition matrix, there was a 3 mm improvement in the FWHM when going from ZF = 1.0 to ZF = 2.0 due to improved spatial sampling. There was no further improvement with the 128 x 128 matrix size.

The COR exhibited a close linear relationship with ZF, for ZF = 1.0 to 4.0, (Fig. 3). The linear correlation coefficient

TABLE 1. Variation of Planar Resolution FWHM (mm) with Zoom Factor (ZF) and Matrix Size (M) for

System B for a Point Source in Air at 10 em from the Gamma Camera

FWHM M ZF (mm)

64 1.0 10.4 1.3 9.1 1.6 8.2 2.0 7.5

128 1.0 7.5 1.6 7.5

1ft

330~----------------------------------------,

325

··················•····· ··•········· ···········•····· ···········• ··························

·····················•

320

315

LEGEND

e SYSTEMS

310+-------.------y------r-------.-----r---~-r------.----_, • SYSTEM A 5 tO 15 2.0 2.5 30 35 40 45

Zoom Factor (ZF)

was 0.999 for both systems A and B. The COR of system B was more stable and reproducible than that of system A and its variation with ZF was significantly smaller. In Figure 3, the corresponding slopes were -0.07 and -0.64 pixels per unit ZF, for system Band system A, respectively.

The point source SPECT studies with the 17-cm rotation radius indicated that there was a 1.5 mm improvement in the FWHM resolution when going from 64 X 64 matrix and ZF = 1.0 to 128 x 128 matrix and ZF = 1.6, as shown in Table 2. However, with the same change in acquisition parameters, an improvement of 3 mm FWHM was observed when the rotation radius was 6 em. For the four acquisitions with different combinations of matrix (M) and ZF (M = 64 x 64 and 128 x 128, ZF = 1.0 and 1.6), system B exhibited better spatial resolution than system A by about l mm FWHM. When the Ramp filter was used (rather than the Ramp-Hanning filter with l cycle/em cut-off frequency), the esti-mated FWHMs were -2 mm less than the values in Table 2 for system B with R = 17 em. For system B, 64 x 64 point source acquisitions, both reconstruction matrices achieved the same FWHM resolution, but the 128 x 128 matrix exhibited better contrast. The Nowak reconstruction algo-rithm and the conventional filtered backprojection algorithm also resulted in the same resolution, but the Nowak algorithm provided better contrast.

Figure 4A presents horizontal profiles through the upper row of hot rods using transaxial reconstructions of the hot rod phantom with ZF = 1.0 (left profile) and ZF = 1.6 (right profile). An improvement in resolution and contrast using a ZF of 1.6 is evident by the better separation and relative magnitude of the three peaks corresponding to the 2, 7 and 3 mm diameter hot rods. Both acquisitions were done in the same setting with system B, using a LEHR collimator, 64 x 64 matrix size, same acquisition time and same reconstruction parameters. With ZF = 1.6, the 2 mm hot rod is well resolved. Compared to the hot rods, the cold rod reconstructions were of inferior quality. The 2, 3, and 4 mm cold rods were not visualized and the 5 mm cold rod was just resolved. Never-

200

FIG. 3. Variation of center of rotation (COR) with zoom factor (ZF) for system A and system B.

theless, an improvement in resolution and contrast was evi-dent with the use of zoom as shown in Figure 4B where horizontal proftles through the middle row of cold rods (12, 6 and 8 mm diameter) are presented for ZF = 1.0 and 1.6 with M = 64 X 64.

The contrast was assessed by using the hot rod reconstruc-tions and the P fV ratios. As shown in Figure 5, the P fV ratios improved significantly when going from ZF = 1.0 to ZF = 1.6 with both matrix sizes but especially with 64 x 64. Not much improvement was observed between matrix 64 x 64 with ZF = 1.6 and matrix 128 x 128 with ZF = 1.0. In both systems, the 6-cm rotation radius resulted in better resolution and contrast than the 17-cm rotation radius. The best reso-lution and contrast was observed with the 128 X 128 acqui-sition matrix and ZF = 1.6. In all the acquisitions, system B showed better resolution and contrast than system A.

Transaxial and coronal reconstructions from a thyroid Z SPECT study are shown in Figure 6. The comparison of Z SPECT with PHC in thyroid studies involved three physicians. Their consensus opinion was correlated with clinical findings. Z SPECT was useful in selected situations for detecting deep seated nodules not seen on the PHC images and in differen-tiating solitary from multiple nodules. However, there were

TABLE 2. Variation of SPECT Resolution FWHM (mm) with Zoom Factor (ZF) and Matrix Size (M) for Systems A and B Using Rotation Radii R = 17 em ·

and R = 6 em SPECTFWHM

(mm)

A B A M ZF R = 17cm R = 17cm R=&cm

64 1.0 15.6 14.6 14.9 1.6 14.3 13.7 12.6

128 1.0 14.5 13.5 13.0 1.6 14.2 13.1 11.9

.JOURNAL OF NUCLEAR MEDICINE TECHNOLOGY

FIG. 4. (A) Horizontal profiles through the upper row of hot rods from SPECT transaxial slices of the hot rods phantom obtained by system B. The three peaks in each profile correspond to the 2, 7 and 3 mm diameter hot rods. The left and right profiles (normalized to the same counts) correspond to zoom factors ZF = 1.0 and ZF = 1.6, respectively. (B) Horizontal profiles correspond to the middle row of cold rods (12, 6, and 8 mm diameter) from SPECT transaxial slices of the cold rod phantom.

some difficulties in the interpretation of hot and isthmus nodules.

DISCUSSION AND CONCLUSIONS

Two tomographic gamma cameras have been compared with respect to their SPECT performance, with and without the use of zoom. The aim was to establish the limits in spatial resolution and contrast detectability. Our motivation was the possible application of high resolution Z SPECT in the ex-amination of small organs such as the thyroid, heart, and knees, as well as in small animals.

VOLUME 18, NUMBER 3, SEPTEMBER 1990

The planar spatial resolution in the digital image improved as expected with the use of zoom, until the appropriate pixel size was reached for recovery of the gamma camera FWHM resolution (6). Thus, in the 64 x 64 acquisitions where the pixel size was larger than half the FWHM of the system, a decrease in pixel size by the use of zoom improved the resolution in the digital image. But in the 128 X 128 acquisi-tions, the pixel size was already less than half the FWHM and therefore a further decrease in pixel size did not result in any improvement.

The o_pserved linear variation of the COR with ZF agrees with the findings of Saw et al. ( 7). System B was shown to be

201

A

c

SYSTEM A 1 R = 17 em

22.,--------------------,

20

18

16

12

10

:~---

:':'~/ ,.•- .......... ' /

:>/

P2/V1

SYSTEMA 1 R=6cm P2/V2 P3/V2

22.,.--------------------,

20

18

16

14

12

I()

Pl/VI

.•- ..............•

'· \

P2/V1 P21V2 P3/V2

B

1/)

~ > ii:

~ ~

e M•128.ZF•16

e M•128.ZF•10

e M=64.ZF•16

•M.o64.ZF•10

D

i > ii:

~ ~

e M•128.ZF•16

e M•128.ZF•10

e M=64.ZF•16

e M=64.ZF•10

22

20

18

16

14

12

10

SYSTEM B 1 R = 17 em

• M'"128.ZF•16

e M•128.ZF•l0

• M-64ZF•16

o\------~---~----.---~~-----l eMz64.ZF•10

22

20

18

16

12

10

Pl{Vl P2}V1

SYSTEMB 1 R=6em

,"'··. ··--...

,//> .. -·-·-.

'' ',/ ' '

P2/V2

·· ..

-......... , '-·

P3fV2

\\ '"''.

\:._ e M•128.ZF•16

• M•128.ZF•10

e M-64.ZF•16

'1----~---~----.---~~----j eM .. 64.ZF•10 P1/V1 P2/V1 P21V2 P3/V2

FIG. 5. Variation of contrast expressed as hot rod peakfvalley (PfV) ratios with acquisition matrix size (M) and zoom factor (ZF) for the two SPECT systems A and B. The Z SPECT rotation radius was 17 em in (A) and (B) and 6 em in (C) and (D).

FIG. 6. An example of Z SPECT thyroid study with M = 128 x 128, ZF = 1.6 and 32 projections. Top: transaxial slices, bottom: coronal slices.

more stable and reproducible than system A exhibiting an almost insignificant variation over a wide range of ZF.

In agreement with Mueller et al. (4), we have found that a high-resolution collimator and a smoother reconstruction til-

202

ter yielded better quality images than a more sensitive lower-resolution collimator and a sharper filter. The Nowak recon-struction algorithm, known as distance weighted filtered back-projection, gives greater weight in the reconstruction process to the nearest of the 180 degrees opposed views (5 ). It led to improved contrast resolution and we, therefore, recommend its use.

The use of ZF and larger matrix size resulted in improved resolution and contrast in both systems, but more so in system B. As expected, this was especially seen when the rotation radius was 6 em, in agreement with the principle that the smaller the camera-to-object distance, the better the resolution (8). The FWHM resolution improves by 1.5 mm with the 17-cm rotation radius and by 3 mm with the 6-cm rotation radius. With the use of 64 x 64 acquisition matrix, LEHR collimator and ZF = 1.6, the 2-mm hot rod was just visible. It is well known that detection of a cold lesion in a hot background is a more difficult task. Under the same condi-tions ( 64 x 64 matrix and ZF = 1.6) the 5-mm diameter cold rod was just visible. The COR reproducibility and general stability of system B resulted in its superior contrast and spatial resolution (8). Although there were some difficulties in the interpretation of hot and isthmus nodules, Z SPECT was useful in selected situations.

Spatial resolution together with contrast and statistical noise govern lesion detectability. As resolution improves, contrast also improves but the improvement in resolution is often at the cost of sensitivity. Thus, unless acquisition time is in-

JOURNAL OF NUCLEAR MEDICINE TECHNOLOGY

creased, the total number of gamma rays detected will be less, statistical noise will be greater and lesion detectability may suffer. In practice, a compromise is necessary. Nevertheless, in reaching such a compromise for SPECT, we suggest, in agreement with others ( 4, 9) that resolution improvement should be given a greater weight than the accompanying loss of sensitivity.

In conclusion, our studies have shown that for small organs or animals, Z SPECT with high resolution collimator can both improve contrast and spatial resolution. We recommend that acquisition be performed with 64 x 64 or 128 x 128 acquisition matrix size (the latter is preferred for a stable tomographic system) with a LEHR collimator and with a ZF = 1.6. For reconstruction, the Nowak algorithm (5) with 128 x 128 reconstruction matrix size is recommended. If the size of the organ to be studied prohibits the use of zoom, the 128 x 128 acquisition matrix size should be preferred.

NOTES

* GE 400 AT with Star computer, GE Medical Systems, Milwaukee, WI t GE 400 ACT with Starcam computer, GE Medical Systems, Milwaukee, WI

VOLUME 18, NUMBER 3, SEPTEMBER 1990

ACKNOWLEDGMENTS

The authors thank Dr. K. Nawaz, Dr. A. M. Farag, and Mr. S. Baig in the evaluation of thyroid studies.

REFERENCES

I. Jaszczak RJ, Coleman RE. Single photon emission computed tomogra-phy (SPECT), Principles and instrumentation. Invest Radio/1985;20:897-910.

2. Coleman RE, Blinder RA, Jaszczak RJ. Single photon emission com-puted tomography (SPECT). Part II: Clinical applications. Invest Radio/

1986;21: 1-11.

3. AAPM Report No. 22. Rotating scintillation camera SPECT acceptance testing and quality control, 1987.

4. Mueller SP, Polak JF, K.ijewski MF, et al. Collimator selection for SPECT brain imaging: the advantage of high resolution. J Nuc/ Med 1986;27: 1729-1738.

5. Nowak DJ, Eisner RL. Fajman WA. Distance-weighted backprojection: aSPECT reconstruction technique. Radiology 1986;159:531-536.

6. Sorenson JA. Phelps ME. eds. Ph.vsics in nuclear medicine. Second edition. Orlando: Grune and Stratton; 1987.

7. Saw CB, Clarke LP, Serafini AN. Influence of zoom factor on centre-of-

rotation of the SPECT system and the resolution of tomographic images. Nuc/ Med Comm 1987;8:3-10.

8. Larsson SA, Bergstrand G, Bergstedt H, et al. A special cut-ofT gamma camera for high-resolution SPECT of the head. J Nuc/ Med 1984;25: 1023-1030.

9. Keyes JW. Fahey FH. Harkness BA. Tips for high quality SPECT. SNM Complller and Inslrwnentation Council Newsletter. 1990.

203