• No se han encontrado resultados

Detector technology:

There are a variety of detectors en ab le of imaging positron-emitting tracers and these can be broadly divided into four classes: i) the thallium doped sodium iodide (Nal (Tl)) gamma camera with lead collimators; ii) the dual-head rotating Nal (Tl) camera with modified electronics for coincidence detection; iii) the dedicated Nal (Tl) PET camera with a ring detection system and iv) the dedicated bismuth germanate oxide (BGO) PET camera with a ring detection system (with either a full or partial ring). The performance of each camera type depends on the figures o f merit of the particular machine and with the wide range in performance comes a wide difference in price. Most Nuclear Medicine departments can afford to run an adapted Anger gamma camera, which is PET capable, but a state-of- the-art full ring, multicrystal dedicated BGO PET camera costs between £1 million and £1.5 million sterling.

The first electronic gamma camera was described by Anger in 1952 and the design remains relatively simple; its primary componoits are a lead collimator and a large Nal (Tl) scintillation crystal (usually between 40-50cm in diameter and 0.6-1.2cm thick) coupled to a hexagonal array of photomultiplier tubes (PMT). Photons pass through the collimator and interact with the scintillation crystal, which results in the emission of light. The light is detected by the PMT and converted into an electrical

signal. The information produced is a two-dimensional representation of three- dimension radionuclide distribution. The concept of SPECT, also introduced by Anger in 1967, allowed the transition from two-dimensional planar imaging to

multiple two-dimensional cross-sectional images. These were produced by rotation of the gamma camera about the patient axis and processing data sets using computer software. The development o f the gamma camera favoured imaging isotope decay by electron capture (y-ray photons in the 50-300 KeV range) rather than positron decay. Imaging positron decay using a Nal (Tl) gamma camera has the disadvantages of lower sensitivity (described below) as well as an intrinsic inefficiency for detection of high-energy photons. I will briefly describe the options available for commercially available detection systems now.

13.2.1 Nal (Tl) gamma camera with lead collimators:

The Nal (T l) gamma camera with lead collimators is the cheapest solution for PET. Nuclear Medicine departments can a d ^ t existing cameras, however, with the cost saving comes a low detection sensitivity. The reason for the poor performance of these cameras can be put down to the inefficiency o f the lead collimator and secondly, due to the physical properties of the Nal (T l) crystals described above. In order to detect coincident gamma rays a camera can either use coincidence electronics or lead collimators to define the direction of the incident gamma rays. Lead collimators absorb 95% or more of the gamma rays which leads to a decrease in sensitivity by one to two orders o f magnitude compared to the electronic devices [Wong WH et al., 1999].

The Nal (Tl) crystals are the components of the system that actually detect the incoming gamma rays. Unfortunately, the Nal (Tl) crystals used are very thin and optimized for detecting 140 KeV gamma rays produced in technetium ( ^ T c ) imaging. The high-energy 511 KeV gamma rays produced by positron emitting isotopes penetrate the crystal leading to approximately 70% escaping detection. Overall, only 2% of the emitted gamma ray pairs are detected resulting in an image resolution o f 2.5-3cm.

13.2.2 Dual-head rotating Nal (T l) camera with modified electronics for coincidence detection

This is a rotating dual-head Nal (Tl) camera with coincidence electronics. Ehmination of the lead collimator increases the detection sensitivity by 2 0-fold, but the

penetration and therefore loss of 511 KeV gamma rays remains a problem with the Nal (Tl) crystals. Overall, this camera assembly is 5-10 times more sensitive than the standard Nal (Tl) gamma camera and can detect lesions between 1.5-2cm [Ziegler et al., 1997]. The draw back of this system is that the camera bdiaves as one large detector, which becomes inactive as soon as a gamma ray is detected. It remains in this state until all the stimulated scintillation light has been emitted. Therefore, if there were a second gamma ray coming in the signal would “pile up” leading to inaccurate detection o f energy. This does not occur with a lead collimator and the only way to reduce signal pile up is to reduce the injected dose of tracer (by upto 80% of that used for the more sophisticated dedicated BGO PET camera). This further reduces image quality and reduces the detection efficiency by nine to ten times.

In addition, the coincidence detection efficiency is geometrically dependent resulting in a high efficiency at the centre of rotation and low efficiency at the periphery. The solution is to ensure that the camera field is significantly larger than the field of view. This camera does have certain advantages in terms o f increased detection sensitivity over the simple Nal (Tl) gamma camera, but this has to be weighed against the

disadvantages of having to reduce the injected dose and limitations of the geometry of the assembly.

1 3 .2 3 Dedicated Nal (Tl) PET camera with a ring detection system

This system is comprised of six Nal (Tl) heads with coincidence detection electronics placed in a fixed ring that surrounds the patient [Karp et al., 1990]. In addition, the thickness o f the Nal (T l) detectors is 2.5cm compared to the 1cm thickness of a dual­ head detector. As well as producing a uniform detection field around the patient the detection sensitivity is increased by four times compared to the dual-head coincidence camera. The problem of signal pile up remains and even though electronic processing of the detected activity can reduce the dead time o f these cameras, the injected dose of

radiotracer has to be reduced by 60-80% of a dedicated BGO PET camera [Karp et al., 1990].

This dedicated PET camera is a cost effective alternative to the expensive state-of-the- art BGO PET cameras, but it is a compromise in terms of absolute detector sensitivity and resolution.

13.2.4 Dedicated BGO PET camera with a ring detection system

The gold standard PET camera is the dedicated BGO PET camera. Unlike the other cameras already described this assembly consists o f several thousand small BGO crystal detectors, which are constructed into a ring that surrounds the patient [Degrade et al., 1994]. This system has two advantages. Firstly, the BGO crystals improve sensitivity because they are better suited for the 51 IKeV photons produced due to their higher density compared to Nal (Tl). The multicrystal layout lets the operator work with much higher count rates so that the injected dose can be increased, which in turn improves image quality. In addition, eliminating the rotation o f the detector ring means that the quality o f studies using short-lived isotopes such as * ^ 0 and is

improved. Overall, these improvements in hardware mean that the practical spatial resolution o f these cameras is far superior to those previously discussed and currently is reported at between 4.5 and 6mm. Ultimately, the resolution of BGO PET detectors

for cancer lesions depends very much on the differential uptake of tracer between tumour and normal tissue. The current view is that malignant lesions of 6 - 10 mm

can be accurately detected.

Image reconstruction and quantitation

The key advantage of functional imaging with PET is the ability to relate detected radioactivity to specific metabohc parameters. This allows comparison between lesions in one patient over time, for example to monitor treatment, or between different patients in order to compare disease. So that one is able to understand the complex and contentious debate surrounding the application o f quantitation to routine clinical use 1 need to first describe the process of image reconstruction.

A comprehensive review o f the physics of image reconstruction is not attempted. I will, however, give a brief overview of the basic details that a clinician should be

familiar with. Images can be acquired after tracer injection in a static or dynamic mode. In the former, the body is imaged in sections of 15cm for both emitted activity and then transmitted activity from a germanium source. This is done at a point in time after tracer injection and gives activity data sets at that particular point in time. Dynamic imaging is the acquisition o f multiple images over a set time period over a region or regions of the body. Dynamic imaging can be used initially to optimise the timing o f subsequent imaging with a particular tracer, but can also be used to study tracer distribution and uptake in a specific region o f interest. This can be of particular use in breast cancer and its metastatic lesions where [^*F]fluoro-2-deoxy-D-glucose (FDG) uptake increases over a three hour period after initial injection [Borner et al., 1999]. Both these techniques allow functional metabolic parameters to be calculated if transport models for the tracer have been validated and blood tracer activity is also known. This means that an area of high FDG uptake could be quantified with respect to glucose metabolism and tissue perfusion, for example.

Image reconstruction is the process o f producing a two or three-dimensional image from the stored data sets using one of two methodologies. The two algorithms used are Iterative and Fourier techniques. Iterative reconstruction techniques lead to a more accurate reconstructed image by allowing for the effects of noise. The basic principle employed is that a ‘best guess’ image is produced and using a number o f iterations the image is improved by comparing this ‘pseudo-projection’ with the raw data. This technique is computationally demanding and therefore, Fourier techniques are more routinely used.

Filtered Back Projection versus Ordered Subsets Expectation Maximisation

The commonest technique used in Fourier reconstruction is filtered back projection (FBP). The fundamental process involves backprojection of the collected projected data onto a reconstruction matrix. Each transaxial slice will have its own matrix and the process is in effect the reverse of the projection process. The resultant image is usually crude and blurred due to overestimation of the activity concentration (figure

1.7). A process known as Ramp filtering is applied prior to backprojection in order to

a

Figure 1.7 The plane containing the hot spheres from the reconstructed images o f a phantom (15 min fo r both emission and transmission). Different image

reconstructions are shown: (a) FBP with measured attenuation correction; (h) FBP with segmented attenuation correction; (c) OSEM with segmented attenuation correction [Visvikis et aL^ 200IJ

FDG