Main: schematic of compact gamma camera (CGC). Inset: image of CGC with protective cover removed to show pinhole collimator. 

Contexts in source publication

Context 1

... medical images are typically acquired using large fi eld of view (LFOV) gamma cameras designed for whole body scanning and SPECT imaging. A new generation of small fi eld of view (SFOV) gamma cameras, now in development, has been designed to provide higher resolution capabilities for speci fi c procedures, such as sentinel node localisation [1]. Standardised procedures for assessing the performance characteristics of medical gamma cameras have generally been based on the original standards published by the US National Electrical Manufacturing Association (NEMA) [2]. In the UK and Europe a comprehensive description of procedures to be carried out in clinical departments has been developed by the Institute of Physics and Engineering in Medicine (IPEM) [3]. However, these tests are designed for use with standard LFOV gamma cameras and are not necessarily applicable to SFOV systems. The term SFOV itself is ambiguous, and has been used for cameras with FOVs between 40 Â 40 cm and 40 Â 40 mm. For cameras working towards the higher end of this range, the IPEM standards are often appropriate. For those instruments operating towards the lower end, these standards may not be applicable and, in some cases, the procedures may even be impossible to perform. In this communication the performance of a SFOV gamma camera currently in development is fully characterised following protocols developed speci fi cally for use with SFOV gamma cameras. Performance characteristics are then compared to those of several standard LFOV systems currently in clinical use, and to similar SFOV cameras in development. The portable compact gamma camera (CGC) has been developed by the Space Research Centre, University of Leicester in collaboration with Radiological and Imaging Sciences at the University of Nottingham. The current camera incorporates a number of improvements on the design previously described in the literature [4], particularly in terms of improved shielding and a new cooling system. The CGC consists of a 0.5 mm diameter pinhole with an acceptance angle of 60 in a 6 mm thick tungsten collimator, with a detector placed at a distance of 10 mm from the pinhole centre. An interchangeable collimator is separated from the detector system by a 1 mm thick Al window for protection. The detector is a 0.5 mm thick Tl doped CsI scintillator, consisting of multiple closely packed CsI(Tl) columns each a few m m wide, joined to an electron multiplying charge coupled device (EMCCD) with Dow Corning optical grease. The EMCCD used is the back-illuminated CCD97 produced by e2v technologies [5]. The active imaging area of the EMCCD measures w 8 Â 8 mm but, due to the use of a pinhole collimator, the fi eld of view of the CGC can be larger than this depending upon the magni fi cation factor (related to imaging distance). The detector is Peltier-cooled to temperatures of between 0 and À 14 C. Tungsten shielding 3 mm thick surrounds the detector enclosure. A photograph and schematic of the CGC can be seen in Fig. 1. Individual photon events are detected using automatic scale selection ( ‘ blob detection ’ ) [4]. Bespoke analysis software fi ts a Gaussian distribution to each light splash recorded on the detector. Each fi tted Gaussian equates to a single gamma photon event in the scintillator. The peak amplitude and standard deviation of the Gaussian may be used to calculate the energy of the interacting gamma photon. A new image can then be created, either from recreated light splashes based on the Gaussian information or from the centre points of individual gamma events. Unless otherwise stated, all images in this communication un- derwent an image correction process as follows. Hot pixels, de fi ned as those recording counts above expected thermal noise in more than 5% of frames in a dark image, were replaced with the average signal value of their 4 nearest neighbours. A fl ood image was taken, either with a point source at a large distance (for intrinsic measurements) or with a uniform fl ood source (extrinsic measurements), regularly throughout each day of testing. Dark images, with no incident illumination, were taken regularly during experimen- tation. The dark and fl ood images were fi rst corrected for hot pixels. A master fl at image was then created by subtracting the dark image from the fl ood image (corrected for any difference in exposure time) and then normalising the resulting image to its maximum value. After hot pixel removal an image would be corrected for fl at fi eld effects by subtracting the dark image and then dividing by the master fl at image. 4 Â 4 pixel binning was used in all cases, producing a 104 Â 112 pixel usable image array with pixel sizes of 64 Â 64 m m. Binning was used as individual CCD pixels (16 m m) would grossly over- sample for expected resolutions of the order of several hundred m m. Edge pixels were removed to eliminate the effect of defects at the machined edge of the scintillator. Raw data were converted to gamma photon counted images using the blob detection method. Intrinsic spatial resolution was calculated using an edge response function (ERF) method. A 10 mm thick lead block, with a 2 Â 20 mm slit, was positioned 40 mm in front of the uncollimated camera face (50 mm from the detector). A 3 mm diameter 14 MBq 99m Tc source was placed 200 mm above the slit. At this height it can be assumed that the photons from the source impinge perpendic- ular to the slit and detector and are parallel to one another. A 10,000 frame image was taken (35 min acquisition time) with the CGC, an example of which is shown in Fig. 2. An additional image was taken with a 10 mm thickness of Perspex placed between the slit and the camera as a scattering medium. Peak pixel counts were 480 and 390 respectively. In some instances the slit collimator was not perfectly orientated parallel to the detector array ’ s principal axes. This was accounted for by a least squares fi tting algorithm which calculated the centre line of the slit image and its angle to detector array axes. The ERF (the relative intensity compared to distance from edge) was then calculated. Although it would be possible to use the ERF as a measure of resolution, the IPEM standard recommends the full width half maximum (FWHM) of a line spread function (LSF). The full width tenth maximum (FWTM) is also reported as it is not expected for the LSF to be Gaussian. The LSF was calculated simply as the derivative of the ERF. Figure 3 shows the ERF and LSF for the example image in Fig. 2. This analysis was completed for each edge. The modulation transfer function (MTF) was also calculated by performing a Fast Fourier Transform of the LSF. This describes the camera response in frequency space. An example MTF is shown in the bottom graph of Fig. 3. The MTF shows the best response at low frequencies, as would be expected, and drops quickly as frequency increases. Bar patterns cease to be visible at around the 10% point of the MTF, so this is the value typically quoted, although occasionally 5% or 3% values are given instead. Table 1 shows a range of typical resolution measurements, using an average of both edges of the slit, for slits orientated vertically and horizontally across the detector face. General trends are as expected, with resolutions degrading with increasing Perspex thickness. It has long been established that the spatial resolution using a pinhole collimator will vary based on the distance between the pinhole and the source [6]. The nature of a pinhole collimator makes measurement of spatial resolution at the collimator face (as would be applicable for a parallel hole collimator) unhelpful for clinical assessment. Resolution measurements were taken at a range of distances, using a 0.5 mm diameter pinhole, and used to calculate a relationship that may be applied to any situation. A 1 mm diameter capillary tube fi lled with 40 MBq of 99m Tc was used as a line source. Varying numbers of 4 mm thick blocks of Perspex were placed directly in front of the camera face. The capillary was then positioned directly in front of the Perspex blocks. In this way the capillary was imaged at a range of distances from the camera with Perspex fi lling the intervening space. The resolution of a pinhole camera will vary across its fi eld of view, so all phantom images were taken within the 30 FOV of the pinhole where this effect is negligible [7]. Image acquisition time was 100 s. The capillary tube was orientated at a range of angles to the detector array and this orientation was corrected for as described in Section 3.1. For each image the pro fi le of the capillary tube was taken and FWHM and FWTM values calculated. These values were then corrected for magni fi cation to give the resolution of the camera in terms of the object it was seeing rather than the image it was producing. A plot of the calculated resolutions against Perspex thickness can be seen in Fig. 4. Least squares fi tting produces a linear fi t with R 2 1⁄4 0.998 for the FWHM data and R 2 1⁄4 0.998 for the FWTM data. As would be expected, resolutions degrade with increasing depths of Perspex. FWTM values increase faster than FWHM values due to the decrease in signal to noise ratio with increasing thickness of Perspex. At greater thicknesses, unscattered counts are reduced and scattered counts act to enhance the background noise level. The overall effect is larger on FWTM values where signal is already relatively low. For LFOV cameras resolution is stated as a single value, measured at the collimator face. Clearly this would not be a practical measurement in this case. The FWHM resolution at the non-magnifying position (13 mm) was found to be 1.28 mm, and this varies according to the linear relationship between distance, d (with scattering medium), and resolution. For the CGC this relationship ...

Context 2

... the use of a pinhole collimator, the field of view of the CGC can be larger than this depending upon the magnification factor (related to imaging dis- tance). The detector is Peltier-cooled to temperatures of between 0 and À14 C. Tungsten shielding 3 mm thick surrounds the de- tector enclosure. A photograph and schematic of the CGC can be seen in Fig. ...