Optimized data acquisition for spectroscopic CT with Timepix3: combining photon-counting and data-driven readout modes with adaptive pixel masking

Abstract Fully spectroscopic X-ray computed tomography (CT) enables the utilization and analysis of the energy-dependent attenuation properties of samples, providing insight into their underlying material composition. This is beneficial across a wide range of scientific and industrial applications. The energy-dependent attenuation must be recorded with high precision to perform a fully spectroscopic CT scan. This is done by capturing the transmitted energy spectrum in every pixel for each angular projection, enabling accurate calculation of the attenuation spectrum in every voxel and, consequently, energy-resolved CT reconstruction. In this study, the energy-resolved transmission spectrum was recorded using the hybrid photon-counting detector Timepix3 (TPX3), coupled with a 1 mm-thick CdTe sensor. The TPX3 offers two distinct modes of operation: the photon-counting readout mode and the spectroscopic, data-driven readout mode. The photon-counting mode (MPX-mode) is fast and produces minimal data, but it provides no energy information beyond the threshold-crossing criterion. In contrast, the data-driven mode, while enabling reconstruction of the densely sampled energy spectrum, generates a substantial amount of data, especially in CT applications where hundreds to thousands of projections must be acquired. Since accurate spectroscopic recording is essential for spectroscopic CT, data acquisition becomes time-consuming and demands extensive storage and computational resources that substantially exceed typical practice. Consequently, fully spectroscopic CT has not yet become practical for routine use. In this work, a new measurement scheme is presented to address challenges in fully spectroscopic CT imaging by synergistically combining the two readout modes of the TPX3 detector. The objectives are to maximize the overall throughput and acquisition efficiency, reducing data volume and acquisition time, thereby rendering fully spectroscopic CT measurements manageable in a standard laboratory setup. The integration of both readout modes with adaptive pixel masking, whereby any pixel is deactivated after recording a predefined number of events, significantly reduces redundant data. Crucially, this method enables dynamic adjustment of the X-ray tube current during the scan, which drastically accelerates data acquisition. In the presented proof-of-concept experiment, this strategy reduced the raw data volume and acquisition time by a factor of approximately 10 compared to a conventional approach, decreasing the measurement time from an estimated 100 hours to 10 hours and the raw data from approximately 13 TB to 1.3 TB. This work details the proposed measurement principle and presents the results of the proof-of-concept spectroscopic CT scan validating the approach.