A Comparison of Hand Tracing and Computerised Cephalometric Analysis

Abstract

The aim of this study was to determine if automated computerised cephalometric analysis with Planmeca Romexis® (Version 5.1.1.R) dental imaging software is as accurate as traditional hand tracing in terms of landmark identification, angular and linear measurements. Pre-treatment lateral cephalograms for a selection of patients undergoing comprehensive orthodontic treatment were collected and irreversibly anonymised. Using a sample of eight radiographs, an inter-examiner reliability investigation was carried out to measure the level of inter-rater agreement of cephalometric analysis by hand tracing between the primary investigator and two Consultant Orthodontists. Using a sample of thirteen radiographs, an intra-examiner reliability investigation was carried out to measure the level of intra-rater agreement for cephalometric analysis by hand tracing of the primary investigator against themselves at two different time points. To compare the results of automated computerised cephalometric analysis using Planmeca Romexis® (Version 5.1.1.R) dental imaging software with those obtained by traditional hand tracing, 68 lateral cephalograms were manually traced by the primary investigator to include 16 anatomical landmarks and 15 angular and linear measurements. The radiographs were then analysed by the automated software programme. The accuracy of the computer software programme was assessed by comparing the automated landmark locations and measurements to the hand traced landmark locations and measurements. Accuracy for automatic landmark identification was recorded as successful detection within a 2.0mm precision range of the hand traced landmark. Accuracy of automatic angular and linear measurements was recorded within 2° or 2mm of the hand traced measurements respectively. Intraclass correlation coefficients were calculated to measure the level of inter-rater agreement. Bland-Altman plots were constructed to analyse the level of inter-rater agreement. Planmeca Romexis® (Version 5.1.1.R) dental imaging software programme had a landmark successful detection rate (SDR) in a 2.0mm precision range of 82.08% with a mean radial error of 1.25 ± 0.91mm. This is comparable with the inter-rater reliability between two clinical experts. Gonion and A-point were the least accurately identified landmarks. The average success rate for all angular and linear measurements within 2° or 2mm respectively was 80.98%. ICCs for all measurements ranged between 0.73 and 1.00 indicating moderate to excellent levels of agreement. Automated cephalometric analysis with Planmeca Romexis® (Version 5.1.1.R) dental imaging software is suitable for clinical use, however, more accurate measurements may be possible if landmark locations are checked and amended if necessary prior to performing an analysis.