The precision of micro-tomography in bone taphonomic experiments and the importance of registration.

Micro-computed tomography (µCT) provides micrometric 3D images and has been used in forensic studies for anthropology pubis measurement or insect description for post mortem interval estimation. Studies have suggested using registration, a superimposing images method between a reference and a target. This technique avoids positioning bias and increase the precision of µCT. However, no clear study has reported the precision with µCT analysis before or after registration in a forensic field. One fresh post mortem sample of a human cranial vault was collected. Two successive µCT acquisitions (resolution 10µm) of it were performed without repositioning. The data from the second acquisition were copied and registered by two trained operators (operators 1 and 2). Operator 1 performed a second registration process after 1 week (operator 1 bis). The images were analysed. The bone volume (BV), bone surface (BS), number of trabeculae (TbN), trabecular thickness (TbTh) and mean trabecular distance (TbSp) were compared before and after registration. The mean (±SD), the coefficient of variation (%CV), and the precision error of the standard deviation absolute value and of the coefficient of variation between operators 1 and 2 (inter-subject variability) and between operator 1 and 1 bis (intra-subject variability) were calculated. We also collected the second phalanx of the 2nd, 3rd and 4th fingers on the hands of a second individual. Two successive scans (resolution 27µm) were performed without repositioning. A comparison (mean±SD of BV, BS, TbN, TbTh, TbSp) was made between the first and second scans with and without registration, and an ANOVA repeated measures procedure was performed. For the vault, we show that after 30 registrations for each operator (1, 2 and 1 bis), the mean and %CV were very close for each variable and between operators. For BV and BS, the difference in the mean value was approximately 0.01 (mm3 and mm2, respectively). The precision error was higher in the inter-subject registrations for each variable. The precision error magnitude for all variables was very low (<0.01) in absolute value and of %CV. For the fingers, the difference between the first and second scans may be approximately 50% without registration. We found that the second scan without registration is significantly different for BV (p=0.006), BS (p=0.007), TbN (p=0.019) and TbSp (p=0.002). Knowing the precision of the device (with and without registration) is important to ensure that the accuracy of the µCT results.