Edge-on semiconductor x-ray detectors - towards high-rate counting computed tomography

The main obstacle today for photon-counting x-ray computed tomography (CT) in the clinical domain is the large x-ray flux rate employed in commercial CT systems enabling ultra-low scanning times. State-of-the-art direct-conversion semiconductor detectors with dedicated readout ASICs are able to reach rates of up to 5–10 million recorded events per second and channel. The requirement coming from clinical CT would be around one billion primary events per second and square millimeter active detector surface. Decreasing the (transverse) pixel size of typically 1 mm² for medical CT down-to about 200 μm×200 μm is effective to reduce the count rates per channel. However, for pixel sizes around and below that level, a number of physical effects like K-escape or charge sharing set in with degrading effects on image quality and spectral information content of the transmission data. A possible way to bridge the gap of about one more order of magnitude in count rate is to apply the sub-structuring of the active sensor volume not only in the directions transverse to the x-ray beam but also along the beam. In this paper we present simulations for a so-called “edge-on” detection concept in which the different sensor layers (in different depths) are read out using independent channels. The main observation reported in this paper is that already a very simple concept of data acquisition or data usage suffices to reconstruct high-quality images with a noise increase restricted to the outermost parts of the object under consideration. The count-rate problem is severe only in the primary beam and in “the first few millimeters of object”, as the event rate is exponentially damped inside the patient, drastically reducing the remaining effective count rate per channel.