Optimizing laboratory X-ray diffraction contrast tomography: Effects of detector binning

In laboratory-based diffraction contrast tomography (LabDCT), pixel binning on 2D detectors is an effective strategy to reduce exposure time and improve acquisition efficiency. However, its impact on reconstruction accuracy remains unclear. To address this, we systematically compared LabDCT datasets acquired without detector binning and with 2 × 2 binning, using synchrotron-based DCT (SRDCT) as the ground truth reference. The results show that unbinned (bin1) acquisitions yield higher grain indexing rates and improved grain boundary position accuracy compared with binned data (bin2), particularly for small grains near the sample surface. This improvement is attributed to the longer exposure time, finer spatial sampling (i.e. smaller effective pixel size), and a lower effective detection-size threshold during spot segmentation. To mitigate the information loss induced by binning, we employed a deep-learning spot-segmentation model (Mask R-CNN), which substantially improves reconstructions from binned images and narrows the performance gap relative to unbinned data. Finally, guidelines for further optimization of LabDCT acquisition and reconstruction are proposed.