Premium
Latent variable approach to correct errors in radiographic measurements
Author(s) -
Tu YuKang,
Donos Nikos,
Pometti Daniela,
Nibali Luigi
Publication year - 2010
Publication title -
european journal of oral sciences
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.802
H-Index - 93
eISSN - 1600-0722
pISSN - 0909-8836
DOI - 10.1111/j.1600-0722.2010.00781.x
Subject(s) - radiography , intraclass correlation , medicine , latent variable , variable (mathematics) , orthodontics , dentistry , nuclear medicine , statistics , radiology , mathematics , psychometrics , clinical psychology , mathematical analysis
Tu Y‐K, Donos N, Pometti D, Nibali L. Latent variable approach to correct errors in radiographic measurements.
Eur J Oral Sci 2010; 118: 642–648. © 2010 Eur J Oral Sci Radiographic outcomes are important for the diagnosis and treatment of periodontal diseases. However, the assessment of radiographic measurements is affected by many factors, and it is therefore difficult to ascertain changes in radiographic outcomes. In this study, we proposed a latent variable approach to correct for the distortion in the radiographic measurements in pairs of periapical radiographs taken before and after periodontal treatment. Clinical data from 123 patients treated with non‐surgical periodontal therapy was used to illustrate the latent variable approach in assessing radiographic changes in infrabony defect depth. Results were compared with a correction factor method. Computer simulations were also undertaken to evaluate the performance of these two methods compared with uncorrected, raw measurements by calculating their intraclass correlation coefficients (ICCs). The example data set showed that the latent variable method (LVM) and the correction factor method (CFM) were comparable. Simulations showed that both methods achieved very high ICCs in different scenarios, whilst uncorrected raw measurements had relatively low ICCs. This study suggests that correction for errors in radiographic measurements is required for routine radiographs. Whilst both LVM and CFM achieve excellent results, LVM is more flexible in handling missing values, and may provide better results when treatment effects are large.