Listing 1 - 2 of 2 |
Sort by
|
Choose an application
In orthodontic treatment, for many decades, panoramic radiographs and lateral cephalograms are considered the standard two-dimensional radiographic techniques required for treatment planning and follow up. Nevertheless, both imaging techniques present with several limitations such as geometric distortion and superimposition of anatomical structures. During recent years, there has been an upward trend in utilizing 3D images, especially from CBCT, as an aid in orthodontic diagnosis and treatment planning but the scientific evidence is still lacking in many aspects. Therefore, the primary goal of this doctoral thesis was to investigate the use of 3D images in orthodontics compared to conventional 2D modalities including panoramic radiography and lateral cephalography. Subsequently, an attempt was made to develop 3D reference systems to increase the reproducibility of several crucial cephalometric landmarks in 3 dimensions. Finally, the Frankfort horizontal plane was revisited, focusing more on its 3D version.This thesis begins with Chapter 1, explaining the general principles of orthodontic treatment planning and imaging modalities traditionally used to achieve the information needed to perform an orthodontic treatment. At the end of the chapter, the overall aims and hypotheses of this doctoral project were presented in detail. In Part I: Chapter 2, a systematic review on 3D cephalometry was presented. This systematic review focused on the scientific evidence for the diagnostic efficacy of 3D cephalometry, especially for landmark identification and measurement accuracy. It was clearly observed that this topic is fairly new and the scientific evidence of the diagnostic efficacy of 3D cephalometry is still limited and more concrete studies need to be performed. Methods of conducting research in this area are very crucial as radiation exposure to young patients is one of the main factors for ethical concern.In Part II, it was aimed to investigate and compare the use of panoramic radiography and the 3D data. In the first chapter of part II, Chapter 3, an attempt was made to compare in vitro subjective image quality and diagnostic validity of reformatted panoramic views from CBCT with digital panoramic radiographs, regarding orthodontic treatment planning. Results revealed that although digital panoramic radiograph still showed better image quality, some reformatted panoramic view from particular CBCT devices could achieve comparable image quality and visualization of anatomical structures. Next in this panoramic imaging part, the agreement between CBCT and panoramic radiographs for initial orthodontic evaluation was assessed. Chapter 4 showed that the agreement between CBCT and panoramic radiograph was good and CBCT could offer the same amount of information necessary for initial orthodontic evaluation.Subsequently, cephalometric imaging modalities were investigated in Part III of this doctoral thesis, beginning with Chapter 5. In thischapter, the linear measurement accuracy of three imaging modalities: two lateral cephalograms and one 3D model from CBCT data, was evaluated. The results showed better observer agreement for 3D measurements. The accuracy of the measurements based on CBCT and 1.5-meter SMD cephalogram was better than a 3-meter SMD cephalogram. These findings have confirmed that the linear measurements accuracy and reliability of 3D measurements based on CBCT data was good when compared to 2D techniques. Chapters 6, 7 and 8 concentrated on the reproducibility of cephalometric landmarks in 3 dimensions and attempted to develop a more robust system for 3D cephalometry. In Chapter 6, a new reference system was designed in Maxilim® software to improve the reproducibility of the sella turcica landmark in 3D. The results showed that the new reference system offered high precision and reproducibility for sella turcica identification in 3 dimensions.In Chapter 7, this time, a new reference system was developed in order to systematically improve the reproducibility of mandibular cephalometric landmarks (Pog, Gn, Me and point B) in 3D. It offered moderate to good overall precision and reproducibility for mandibular cephalometric midline landmark identification.Chapter 8 was the last study on 3D cephalometry in this doctoral thesis. The aim was to evaluate the Frankfort horizontal plane (FH), which is widely used in 3D cephalometric analysis. In this chapter, the precision and reproducibility of landmarks that form the Frankfort horizontal plane (Po, Or) and newly chosen landmarks (IAF, ZyMS) was investigated. The angular differences of optional planes compared to the Frankfort plane in 3D were measured. It was demonstrated that the precision and reproducibility of Po and Or was moderate. IAM and ZyMS showed good precision and reproducibility. From the newly proposed planes, the ones closest to the original FH are the plane formed by connecting Or-R, Or-L and mid-IAF (Plane 3) and the plane formed by connecting Po-R, Po-Land mid-ZyMS (Plane 4). This study demonstrated the possibility of using new planes when traditional FHs were not feasible.Lastly, in Chapter 9, the general discussion and conclusions were thoroughly discussed and presented. The findings of the present doctoral thesis elaborated the use of 2D and 3D images for orthodontic treatment and showed the possibility and new development to improve the use of 3D cephalometry. Although the scientific evidence on clinical use of 3D cephalometry is still limited, this project helps to provide a solid base for future studies. New studies should focus on the implementation of 3D cephalometry in clinical practice and evaluate how this new technology may improve the treatment outcome of orthodontic patients. In the near future, one ultra-low dose CBCT scan may be able to yield all necessary information and replace a cascade of 2D radiographic images while still keeping the ALARA principle. Whether those strategies are equally valid, remains to be proven.
Choose an application
An image processing pipeline is presented for automated localisation of anatomical structures or landmarks. Applications in the field of post-mortem human identification and for oral and maxillofacial surgery planning rely on anatomical landmarks in medical images. More often three-dimensional cone-beam CT images are used on which annotation of landmarks is more difficult than standard X-ray images. Techniques exist to automatically localise and even delineate these landmarks but they usually require manually defined initialisation points. Since manual delineation is often time-consuming, inaccurate and subjective, automation is desired. The process applies the random decision forest algorithm which enables to analyse images acquired by different modalities and search for a variety of anatomical landmarks. A random decision forest is a collection of decision trees wherein each tree predicts a value for every data point of the image. By gathering the predictions of all trees and combining them for all image points, a final prediction is made for the position of the anatomical landmark. The forest is called random since every tree is built up in a random fashion. As a consequence, trees are decorrelated and give independent predictions which will improve the final forest outcome. A tree makes a prediction based on information about the data point. This information is delivered by features. They inspect the image appearance in the close neighbourhood of the data point as well as image areas further away. These are local and contextual features. Characteristic zones in the image, e.g. a bright spot, are detected by these features and provide clues about the position of the target structure. Also symmetry features are assumed to be valuable. The performance of this procedure is validated on two-dimensional panoramic radiographs depicting the oral cavity and on three-dimensional cephalograms which visualise the hard tissues of the head. In both datasets, anatomical structures of interest could be automatically localised. Nonetheless, the influence of selected parameters and the quality and size of the datasets cannot be underestimated. Therefore further investigation is required to build up the robustness of the processing pipeline.
Listing 1 - 2 of 2 |
Sort by
|