W. W. Moses & J.S. HuberDedicated Nuclear Medical Instrumentation for Prostate Cancer
Dedicated Nuclear Medical Instrumentation for Prostate Cancer
W. W. Moses and J. S. Huber
Lawrence Berkeley National Laboratory, Berkeley, CA 94720 USA
Abstract
We describe design concepts for three nuclear medical systems optimized for imaging the human prostate. Functional (rather than anatomic) imaging of prostate cancer can assist in the diagnosis, treatment planning, and follow-up of therapy, and cameras optimized for imaging a single organ / disease potentially have better performance tradeoffs than conventional multi-purpose nuclear medicine cameras. We describe the design considerations for a transrectal single gamma imager, a PET camera with a transrectal probe and an external bank of detectors, and a PET camera with a pair of external curved detector banks. Of these three concepts, the final one appears to be the most advantageous.
Key Words: Prostate Cancer, PET, SPECT, Instrumentation
1 Introduction
Until recently, the imaging of prostate cancer has been confined to modalities that provide anatomical images, such as ultrasound, x-ray CT, or MRI imaging. This imaging is done to assist in several facets of the diagnosis and treatment of prostate cancer, such as guiding biopsy, staging, planning therapy, and monitoring the effectiveness of therapy. In the last half decade there has been progress in using nuclear medical imaging techniques (as well as spectroscopic MRI) to image metabolic aspects associated with prostate cancer. The emergence of radiopharmaceuticals optimized for imaging prostate cancer has spurred interest in cameras optimized for imaging prostate cancer. This paper explores conceptual designs for several nuclear medical imaging systems.
The American Cancer Society estimates that 232,090 new cases of prostate cancer were diagnosed and 30,350 men died of prostate cancer in the United States in 2005. About 90% of all prostate cancers are found in the local and regional stages [1]. Prostate cancer suspicion is typically based on an elevated prostate-specific antigen (PSA) level or on a suspicious node found during a digital rectal exam (DRE). More than half of all cancers detected today are not palpable, and PSA and DRE screenings have high false-positive rates in general clinical practice [2]. The treatment decision is mainly based on biopsy confirmation of prostate cancer, but the diagnostic accuracy of biopsy is problematic [3]. A new imaging technology for sensitive detection of prostate cancer is needed to confirm initial diagnosis, guide biopsy and help guide treatment decisions. A method to monitor response to therapy after an intervention is also needed. Current means of assessing treatment response in prostate cancer are imprecise [4-6], so the administration of multiple courses of a therapy is often necessary before a clear indication of response or progression can be determined. Thus, we need a better prostate imaging technique to monitor response to therapy, assess the efficacy of new treatments, and detect local recurrence sooner.
Prostate cancer imaging with radiopharmaceuticals has used three types of instruments: (1) planar gamma cameras, (2) single photon emission tomography (SPECT) cameras, and (3) positron emission tomography (PET) cameras. As radiopharmaceuticals have been developed that are appropriate for use in any of these camera types, it would seem reasonable to consider prostate-specific designs for all three of these modalities. However, planar gamma cameras image 99mTc compounds that are designed to accumulate in metastatic disease in the bone, and so do not accumulate in the prostate. While they are widely used in the management of prostate cancer, these agents would not be used in a single photon camera that is optimized for imaging the prostate. Thus, we consider only SPECT and PET designs optimized for prostate imaging.
2 SPECT Methods
2.1 Prostate Cancer and SPECT
SPECT imaging of the prostate is dominated by 111In-labeled monoclonal antibodies and 99mTc-labeled peptides. The most widely used compound is ProstaScintTM, an 111In-labeled murine monoclonal antibody directed against prostate specific membrane antigen. The acquisition and interpretation of the ProstaScintTM images are technically demanding due to radiopharmaceutical uptake by background organs—it takes several days to accumulate in “target” tissue and clear from “background” tissue. Nonetheless, the reported sensitivity and specificity of ProstaScintTMimaging for detection of pelvic lymph node metastases is 62% and 72%, respectively [7].
2.2 Transrectal Single Photon Camera
The design concept that we explored for a single photon camera optimized for imaging is a transrectal camera. The imaging performance of a single photon camera is dominated by its collimator. As the parallel-hole collimator efficiency is independent of the collimator to source distance but the spatial resolution is linearly proportional to the collimator to source distance, optimal performance is achieved by placing the camera as close to the prostate as possible. This suggests a transrectal geometry, as shown in Figure1, as it places the front face of the collimator less than a centimeter from the prostate.
Figure 1. Geometry for transrectal imaging of the prostate with a single gamma camera. The camera is inserted transrectally, bringing the front face of the camera close (within ~1cm) to the prostate. The lines running through the prostate in this figure indicate the direction that is “viewed” by the camera. The count rate in each pixel in the camera is proportional to the integrated activity along one of the lines.
A suitable camera geometry is shown in Figure2. The gamma ray detector is an 8x8 array of pixels, each 3mm square. The detector could either be a solid state detector material (such as cadmium zinc telluride [8-10]) or a scintillator / photodetector combination (such as CsI:Tl coupled to a PIN photodiode array [11, 12]). The detector array would sit behind a collimator. For optimal spatial resolution, the collimator would have square holes that are matched to each element in the detector array. To minimize the septal thickness and ease manufacturing, the collimator would be made of thin sheets of tungsten with holes etched via photolithography, then stacked and glued together. As the number of individual detector elements is relatively high and the leads between the detector elements and their amplifiers must be short (to minimize electronic noise), the electronics must be part of the camera head, and would take the form of a custom integrated circuit mounted on a printed circuit board. This imaging head would be mounted on a support structure that would provide mechanical support for insertion into the rectum and electrical connectivity.
Figure 2. Geometry of a transrectal single gamma camera. The active components of camera are a gamma imager, readout electronics, and a collimator. They are mounted on a mechanical support structure and are encased in shielding (not shown) that blocks gamma rays that do not pass through the collimator.
However, this design has a number of drawbacks. The size of the camera is likely to make it uncomfortable for the patient and so poorly tolerated. Since the camera is likely to be imaging 111In-ProstaScintTM, whose emissions are 171keV and 245keV, the desired detector thickness is 1.5–2.0cm in order to have high detection efficiency (the attenuation length in CsI is 0.9cm at 245keV and 0.4cm at 171keV). One could probably consider the minimum thickness to be 0.5cm. For the collimator, collimators for imaging 111In typically have an aspect ratio (length : hole diameter) of ~10:1. Given a 3mm hole diameter, this implies a length of 3cm, which is rather long. The same aspect ratio can be obtained without having any collimator holes expose more than one detector pixel by dividing the 3mm detector pitch by an integer (i.e., have square holes on a 1.5mm or 1mm pitch, which divide the pitch by a factor of 2 and 3 respectively). However, the septa (thickness between holes) must be ~0.5mm (assuming a tungsten collimator), so these reduction factors (smaller holes) are problematic because the collimator septa either begin to occlude a significant portion of the hole or they become so thin that the gamma rays can penetrate the septa easily. Thus, the desired collimator thickness is 3cm, with 1cm probably representing the minimum thickness. Finally, approximately 25 attenuation lengths of shielding must be placed around the entire camera in order to block gamma rays from striking the detector directly (i.e., without going through the collimator). At 245keV, the attenuation length of lead and tungsten are 1.5mm and 1.1mm, implying desired shielding thicknesses of 3.75cm and 2.75cm respectively. The minimum conceivable shielding thickness is about 1.5cm centimeter of tungsten. Thus, the desired thickness is 6.25cm and the minimum conceivable thickness is 3.0cm.
While there are serious concerns as to whether a 3.0–6.25cm thick device can be tolerated by the patient, the most serious drawback is that fact that this type of device is likely to produce a 2-D planar image instead of a 3-D volumetric image. This is illustrated in Figure3, which shows how data for a 3-D image is collected with an external gamma camera. In order to create a 3-D volumetric image using computed tomography, you must take a large number (~100) of individual planar projections of the object, with each projection taken from a different angle. Figure3 shows this data being collected at three different projection angles—vertical, 45°, and horizontal. While this is easily accomplished with an external camera, it is obviously impossible to move a camera with transrectal geometry around the prostate. It is conceivable that with advanced collimator designs and processing algorithms, some depth information might be obtained with a stationary camera [13]. An alternate solution might be to use an external camera, but such a camera would be little different than a conventional SPECT camera, and thus not have any advantage for prostate imaging. In short, the transrectal geometry is likely only to be capable of generating a planar image of the prostate, which will have significantly less contrast and diagnostic utility than a 3-D volumetric image.
Figure 3. Projection views necessary for tomography. In order to reconstruct tomographic (3-dimensional) images, projection data of the object must be taken from a large number of “views.” Figures a), b), and c) show an external gamma camera positioned to acquire data from three different views (vertical, 45°, and horizontal respectively), as well as a projection of the object taken at each view.
3 PET Methods
3.1 Prostate Cancer and PET
In general, PET has higher sensitivity, spatial resolution, and cost than SPECT. The most commonly used PET radiopharmaceutical in oncologic imaging is 18F-fluorodeoxyglucose (FDG), and the first PET studies of prostate cancer were performed with FDG. Unfortunately, FDG is not very prostate specific and bladder accumulation of radioactivity often obscures prostate tumors [14]. Thus, SPECT imaging with ProstaScintTM is regarded as being superior to PET imaging with FDG for predicting the presence of prostate cancer [15], particularly for less aggressive disease [16].
Fortunately, newly developed PET radiopharmaceuticals, such as 11C-methionine, 11C-choline, 11C-acetate, or 18F-fluorocholine, have recently demonstrated outstanding results in the sensitive detection of prostate cancer. Hara and co-workers find that: 11C-choline clears the blood quickly; its uptake in prostate tumors provides excellent tumor/normal contrast; and bladder accumulation is minimal [17]. Therefore, 11C-choline is an extremely attractive PET tracer for imaging prostate tumors [18-27]. Figure4 shows 11C-choline images of a prostate cancer patient before and after therapy, demonstrating the ability to detect prostate carcinoma and follow therapy efficacy using 11C-choline. Several other radiopharmaceuticals are also currently under investigation for prostate cancer imaging, including 11C-acetate [28-30], 11C-methionine [31-34], and 18F-fluorocholine [35, 36]. Therefore, we were also motivated to also consider PET camera designs that are optimized for imaging the prostate.
Figure 4. PET images of choline cancer. 11C-choline image of a patient with prostate cancer before (a) and after (b) treatment. These grayscale images indicate a high (white) uptake in the prostate center compared to a low (gray) uptake elsewhere. Images provided by Hara and co-workers [17].
3.2 Transrectal PET Camera
The first prostate-specific PET camera design that we considered is shown in Figure5. Like the single photon camera, this design employs a transrectal probe, but because PET requires detection of a pair of annihilation photons, it also includes an external bank of detectors. The advantages of this geometry over conventional PET geometries parallel those for the transrectal single gamma system. By placing a detector transrectally, it is extremely close to the prostate, and the external bank of detectors is also placed as close to the patient’s abdomen as possible. This increases the sensitivity and also has the potential to improve the spatial resolution. When the object is much closer to one detector in a PET camera than the other, the spatial resolution is determined by the closest detector. If this detector has higher resolution than conventional detectors (while maintaining high efficiency), then high-resolution, high-efficiency imaging can be achieved [37]. As the volume of the transrectal detector is much smaller than the volume of the external detector bank, one can use a transrectal detector that costs considerably more or is considerably more complex than a conventional module without affecting the commercial viability of the design. In fact, this design uses a factor of ~3 less detector volume than a conventional PET camera.
Figure 5. Geometry for transrectal imaging of the prostate with a PET camera. One PET detector module is inserted transrectally, bringing the front face of that module close (within ~1cm) to the prostate. The other end of each chord is defined by an external bank of PET detector modules The lines running through the prostate in this figure indicate the directions that are “viewed” by the camera. The count rate in each chord in the camera is proportional to the integrated activity along one of the lines.
The size issues discussed with the transrectal single photon imager also exist for the transrectal PET probe, but are less of a problem. This is because PET relies on electrical collimation rather than mechanical collimation, so there is no need for a collimator and the shielding requirements are much reduced. However, the requirements for this PET probe are significantly higher than they are for conventional PET detector modules. First, it will be exposed to a much higher gamma flux than conventional PET detector modules, as it is much closer to the patient than detectors in a conventional PET camera and there is substantially less shielding. Therefore, the transrectal PET probe must have much higher rate capability (and therefore much lower dead time) than conventional PET detector modules. In addition, the probe must be a “volumetric” detector. That is, most PET detectors identify the position of each 511keV photon interaction accurately (within ~5mm) in two coordinates (in a cylindrical camera geometry, the axial and azimuthal directions), but significantly less accurately (~25mm) in the third coordinate. Thus, it is essentially a 2-D detector. In order to maintain high spatial resolution, the transrectal probe needs to measure the third coordinate with accuracy similar to what is measured in the other two coordinates. This can be solved by using a detector that measures the “depth-of-interaction,” and a number of detector designs have been proposed [38-46] that could potentially provide the requisite performance.
However, there are several drawbacks to this design. Building a detector probe that can simultaneously achieve high resolution, high efficiency, and meet the count rate / dead time requirements is extremely challenging. There are also some potential patient safety issues. Many of the potentially suitable detector designs utilize photomultiplier tubes, which have fragile glass envelopes and require operating voltages of approximately one kilovolt. The design must be robust enough that no harm will come to the patient even if the photomultiplier tube were to break while inserted into the patient’s rectum. However, the largest drawback of this design is the same as the largest drawback in the transrectal single photon probe design, which is the type of image produced. This design can only collect data from a single (or small number of) projection angles around the prostate, and so will produce images that are little more than two dimensional. Although the radiotracer contrast in the patient when using PET radiopharmaceuticals is better than 111In-ProstaScintTM, the resulting images are unlikely to be of diagnostic quality.
3.3 External PET Camera
The final prostate-specific PET camera design that we considered is sketched in Figure6. It consists of two external banks of detector modules arranged in an elliptical geometry (45cm minor axis, 70cm major axis). This geometry brings the detector modules as close to the prostate as possible, which increases the sensitivity. The axial extent is only 8cm, which is approximately half that of a conventional PET camera. This provides better shielding, which reduces the number of scatter and random background events. This design requires about one-fourth of the detector volume of a conventional PET camera, which significantly reduces the cost.
Figure 6. Geometry for the prostate-specific PET camera. (a) Drawing of a transaxial view through prostate, showing the patient centered between two PET detector banks. The individual detector modules are angled to point towards the prostate. (b) Drawing of the sagittal view. The bottom arc is fixed below the patient bed, whereas the top arc adjusts vertically for patient access and compactness. Both detector banks are tilted and positioned as close as possible to the prostate, which improves sensitivity and minimizes attenuation.
The bottom detector bank is fixed below the patient bed, whereas the top bank can be moved upward for patient access and downward for maximum sensitivity. Each bank consists of 2 axial rows of 20 Siemens/CTI EXACT HR+ block detectors for a total of 80 detectors per camera. We use these HR+ block detectors because they are three attenuation lengths thick for good detection efficiency with narrow detector elements for good spatial resolution. The detector modules point toward the center of the camera (i.e., close to where the prostate is located) to minimize penetration artifacts. However, this geometry accentuates the penetration artifacts for sources located near the edge of the camera field of view. In short, we sacrifice the size of the volume that is imaged with good spatial resolution to a cylinder at the center of the camera that is approximately 15cm in diameter and 8cm high, and in return obtain better sensitivity, lower scatter and randoms backgrounds, and lower cost. A patient of average size is not fully encircled by detector modules, which results in irregular and incomplete sampling due to the side gaps. However, the angular coverage is nearly complete (unlike the transrectal designs), so we are able to create 3-D volumetric images. Using a 3D iterative penalized maximum likelihood reconstruction algorithm with randoms and attenuation correction, we are able to reconstruct nearly artifact free images in the region of interest [47-50].