Gamma camera technology

[Gyro]

I had a bone scan yesterday and watching the large white cover plate passing about 30mm over my body got me curious about how gamma cameras actually work. Looking on wikipedia, I found that the technology is rather surprising. Very different from a modern electronic x-ray plate where there is a dense array of semiconductor sensors behind the scintillatior sheet (I think mikeselectricstuff did a video teardown of one).

The first difference is that there is a big finely perforated Lead columnator plate, anything from 25 to 55mm thick (I'm glad I didn't know that as it slowly passed my ear!). This is followed you a scintillator plate, typically doped Sodium iodide. Now for the surprising bit... Instead of a silicon detector array, gamma cameras use a hexagonal array of photomultiplier tubes optically coupled [Edit: via a light guide]  to the plate. These have a huge face area compared to the resolution of the camera and the position of each light flash is determined by comparing the amplitude of the outputs of all of photomultipliers in the array.

The number of scintillations must be huge (or so my geiger counter told me when I got home). Ok, there is a lot of data redundancy as the plate takes about 15 minutes to traverse the length of the body (assuming that you can determine what is redundant). To be able to accurately determine the exact position of an individual flash using only amplitude information (no time of flight as far as I can find), using a coarse array of photomultipliers seems nothing short of miraculous - something that you would probably kick out at the feasibility stage! The images I've seen look to be surprisingly high resolution.

I thought I would share anyway - sometimes things don't operate anywhere close to how you might think they do. Before I looked, I had visualized some sort of fast moving scanning PN junction(s) assembly, probably with a fine slit in a lead shield, rotating behind the cover plate. Now I know why it was so quiet.

https://en.wikipedia.org/wiki/Bone_scintigraphy

https://en.wikipedia.org/wiki/Gamma_camera

[Kleinstein]

Using the amplitude information with photomultipiers may look like a strange way, but take into account the developement. The simply scintillator counters still use photomulpiers, using photodiodes there is on a new developement. With a rather large scintillator it is also not so far off using more than 1 photomulitplier to cover more of the surface and the position information than comes natural. One may want to look at the amplitude of the pulses anyway to distuish between different energies and suppress some background.

Compared to a geiger counter scintillators have something like 100 to 1000 times higher sensitivty as there is just more mass than in the gas of the geiger tube.

[Gyro]

I don't know if it would be practical to develop a linear array type camera (like a flatbed scanner). I can visualize something like a U shaped Lead trough with a line of fine holes at the required resolution followed by a scintillator in strip form, then some kind of optical mask with a line array of photodiodes either directly positioned above or via an optical fibre array.

Photomultipliers are obviously more rugged in the presence of radiation and have huge gain. I'm not sure whether photodiodes and associated gain stages would have sufficient sensitivity and s/n to compete in such a line array situation, where the gamma rays are highly columnated and the single row of holes limit the exposure compared to a perforated plate. Obviously they want to keep the injected isotope dose to the minimum practical (I read that it is equivalent to around 2 years background). Such a setup would be much simpler in image processing terms though, although processing is cheap these days.

I guess, if it was that simple, somebody would have one on the market by now. The current setup uses columnating plates (as far as I could judge) around 600mm x 800mm, (I'm not sure how many hexagonal photomultipliers that makes).  At the thicknesses I mentioned, they really are Lead slabs, which must weigh upwards of a tonne each. The machine has two assemblies on Y shaped frame that rotates from horizontal (for access) to vertical (plates horizontal) for scanning. In addition to the linear movement of around 2m (maybe a little less due to plate size), the top assembly also moves up and down to match the contour of the body. I'm not sure if the second assembly actually is a camera (it would make sense to capture an image from below the table too), or simply a counterweight to assist in rotation. Either way, the cantilevered load on the foundation and bearings must be huge. If there was a simpler way I'm sure they would have used it by now!

[Wolfram]

Signal to noise ratio is alpha and omega when it comes to patient imaging, so anything with worse SNR will give a higher patient dose for the same diagnostic data. It's hard to beat bulk scintillators and large area PMTs in this application. As of the last decade or so, Geiger mode multipixel APDs have become available (often called SiPM or MPPD/MPPC), but cost per area is far higher than PMTs as far as I know.

More advanced fine pitch sensors and thin scintillators are used in X-ray imaging, but there you can get away with a less sensitive detector for multiple reasons. Firstly, radiation is only present during imaging, compared to gammagraphy where you get internal radiation during the entire biological/radiological half life of the imaging isotope. Secondly the gamma decay used in gammagraphy is isotropic, so it has to be collimated heavily in order to get spatial resolution, compared to X-rays where you have a directional beam that's limited to the imaging area. For this reason, gammagraphy has thousands of times higher patient dose for a given image SNR and resolution, making it critical to collect as many photons as possible.

I suspect a lot of the RND money that could go into gammagraphy is being invested in PET instead.

[Kleinstein]

The dedectors for PET (positron emission tomography) are likely very similar. Just detectors on all sides and looking for correlation to reduce the background. The finally detected signal is a pair of 512 keV gamma.