Thesis
The ability of machine learning to predict skeletal maturation using the cervical vertebrae maturation (CVM) method on traced lateral cephalograms
University of Iowa
Master of Science (MS), University of Iowa
Spring 2021
DOI: 10.17077/etd.006045
Abstract
Introduction Growth and development are modulated by environmental and host factors, with the amount, rate and timing varying greatly among individuals. Common indices used to predict craniofacial growth for orthodontic diagnosis and treatment planning purposes are the hand-wrist radiograph, cervical vertebrae maturation (CVM) method, statural height increase and presentation of secondary sex characteristics. The reliability and reproducibility of these methods have come into questions and make growth prediction a difficult task.
Materials and Methods
This study evaluated whether artificial intelligence, using a novel principal component analysis (PCA), can effectively assess skeletal maturation using the cervical vertebrae method from traced lateral cephalograms based off the CVM method. Longitudinal cephalometric images from patients (N=50) were selected from the AAOF Craniofacial Growth Legacy Collection at the University of Iowa College of Dentistry. The cervical vertebrae (C2, C3 and C4) were traced at six different time-points using ImageJ software and corresponding X,Y-coordinates exported for analysis. PCA was applied to normalized vertices finding 20 principal component shape descriptors. Statistical correlations between the descriptors and age was assimilated using ordinary least squares (OLS) regression.
Results
PCA confirmed substantive intrinsic variation in vertebral geometry that was confirmed with quantitative analysis. The concavity at the inferior border of C2, C3 and C4 was captured by the 2nd, 4th and 15th principal components. The trapezoidal characteristics of C3 and C4 was captured by the 2nd, 3rd, 4th and 19th principal components. The aspect ratio of C3 and C4 was captured by the 1st, 2nd and 15th principal components. There was a moderate-strong relationship of shape descriptors with age (RMSE = 3.29 years, R-squared = 0.642), and a strong correlation of shape descriptors with cephalometric number (RMSE = 0.783, R-squared = 0.781).
Conclusions
AI is capable of characterizing shape changes in the cervical vertebrae C3, C3 and C4 into principal component shape descriptors. The changes are specifically related to concavity of the inferior border and shape progression from trapezoidal to rectangular horizontal to square over time. PCA confirmed the correlation between shape descriptors and geometric changes, as well as their correlation to age and corresponding lateral cephalogram. The novel PCA and shape descriptors are capable of predicting CVM stages and shows potential as a diagnostic tool for craniofacial growth in regards to orthodontic diagnosis and treatment planning.
Details
- Title: Subtitle
- The ability of machine learning to predict skeletal maturation using the cervical vertebrae maturation (CVM) method on traced lateral cephalograms
- Creators
- Megan S Utter
- Contributors
- Kyungsup Shin (Advisor)Stephen Baek (Committee Member)Thomas E Southard (Committee Member)Michael A Callan (Committee Member)Shankar Rengasamy Venugopalan (Committee Member)
- Resource Type
- Thesis
- Degree Awarded
- Master of Science (MS), University of Iowa
- Degree in
- Orthodontics
- Date degree season
- Spring 2021
- DOI
- 10.17077/etd.006045
- Publisher
- University of Iowa
- Number of pages
- ix, 42 pages
- Copyright
- Copyright 2021 Megan S Utter
- Language
- English
- Description illustrations
- illustrations (some color)
- Description bibliographic
- Includes bibliographical references (pages 41-42).
- Public Abstract (ETD)
Growth of the human body and face is difficult to understand, yet important for orthodontic care and diagnosis. Various indices have been developed in order to better comprehend adolescent growth, one of which focuses on maturation of the cervical vertebrae name the cervical vertebrae maturation (CVM) method. Recently the reliability and reproducibility of this method has come into question.
This study sought to evaluate if artificial intelligence (AI) using a novel analytical tool, can effectively assess skeletal maturation using cervical vertebrae from longitudinal orthodontic radiographs from the AAOF Craniofacial Growth Legacy at the University of Iowa College of Dentistry. The cervical vertebrae (C2, C3, C4) were traced using ImageJ software at six different time points exported for analysis. After normalizing the vertebral tracings, 20 shape descriptors were found. Statistical correlation was evaluated between the descriptors and age.
Quantitative analysis confirmed the shape variation observed in the cervical vertebrae, mainly the concavity of inferior border of all three vertebrae and change in shape of the C3 and C4 vertebrae. There was a moderate-strong relationship of shape descriptors with age and cephalometric number.
This study demonstrated that AI is capable of characterizing morphological changes of the cervical vertebrae into novel shape descriptors which relate the concavity of the inferior border and shape progression from trapezoidal to rectangular horizontal to square over time. Correlation between the shape descriptors with age was confirmed with statistical analysis. Thus, AI is capable of predicting cervical vertebrae maturation and shows potential as a diagnostic tool in orthodontics.
- Academic Unit
- Orthodontics; Craniofacial Anomalies Research Center
- Record Identifier
- 9984097368602771
Metrics
8 File views/ downloads
42 Record Views