Otolaryngology - Theses
Permanent URI for this collection
Search Results
1 - 3 of 3
-
ItemImproving the preservation of acoustic hearing for cochlear implant recipients( 2022)Cochlear implants have been widely used to provide an alternative way of hearing for those who have experienced hearing loss, specifically in the high frequencies. Over the years, cochlear implant recipient criteria have expanded to those who are not only classified as severely to profoundly deaf, but also to include those who still have substantial residual hearing, particularly in the low frequencies. The increase in cochlear implant candidacy motivated the development of a new modality for treating patients with residual low-frequency hearing, termed electro-acoustic stimulation. Electro-acoustic stimulation involves electrical stimulation from the implant for mid to high frequencies complementary to low-frequency amplification through hearing aids. The resulting hearing experience has been shown to be advantageous over cochlear implant stimulation alone, allowing for better music appreciation, and improved speech recognition in noise. To achieve electro-acoustic stimulation, the residual low-frequency hearing must be preserved during and following cochlear implantation. Unfortunately, this is not always the case as implantation often results in either immediate or delayed loss of residual hearing loss in the postoperative period in more than half the individuals. Causes for these losses are speculated to be associated with intra-cochlear trauma occurring during electrode array insertion in conjunction with the consequential inflammatory response that ensues. To further improve hearing outcomes in cochlear implantation recipients, the issue currently presented is whether the detection of cochlear injury can be achieved and result in improved hearing preservation rates post-implantation. The investigations in this thesis aim to explore the implementation of a monitoring system, namely impedance, to identify whether an intra-cochlear injury has occurred, such as cochlear wall injury leading to the infiltration of blood into the cochlea. Impedance is a passive electrical measurement that evaluates the resistance and conductivity of the medium surrounding the recording electrodes. Specifically, this thesis will explore relationships between intra-operative impedance measurements and clinical outcomes such as hearing thresholds. Additionally, the impedance will be monitored post-operatively to determine if impedance fluctuations align with physiological events anticipated to occur in the cochlear following implantation that is unfavourable for hearing preservation. Lastly, investigations in this thesis will explore a possible therapeutic intervention for the removal of intra-scalar blood if it were to enter the cochlea during implantation.
-
ItemDeep Learning for Three-Dimensional Multi-Modal Medical Image Processing: From Classification to Segmentation( 2022)Deep learning is a state-of-the-art method in machine learning, proven successful in many application domains. This research aims to use a deep learning framework for three-dimensional (3D) medical image segmentation. Multi-modal images are used in this process as they provide additional information that can make segmentation easier. For example, in computed tomography (CT), more rigid materials such as bone are better defined, while in magnetic resonance imaging (MRI), softer materials such as anatomical structures are better defined. However, multi-modal images for the same person may not be in the same orientation and have different resolutions. Therefore, aligning (or registering) multi-modal images prior to segmentation is an important task. Furthermore, the type of image under consideration is essential in the segmentation process. For instance, a method used to segment the brain may not be applicable in segmenting the heart. Therefore, to fully automate the process of segmentation, it is crucial to classify the multi-modal images and register them before performing a segmentation. Thus, in this thesis, we introduce methods based on deep learning for the classification, registration, and segmentation of multi-modal medical images. For each of these tasks, mainly due to practical limitations such as availability of datasets, we develop and validate our methods for one application/dataset. Firstly, we explore the problem of classifying 3D multi-modal images of different organs. To this end, we introduce a rotation and translation invariant classification model. We use the fact that most human organs are (approximately) symmetrical to simplify the problem. We extract a two-dimensional (2D) representative slice of the 3D organs and use that 2D slice as the input to a deep learning model to perform the classification. We prove experimentally that our method is comparable to existing classification techniques when the assumptions of viewing direction and patient orientation are met. Moreover, we establish that it shows high accuracy even when these assumptions are violated, where other methods fail. Secondly, we introduce a novel deep learning method for registering 3D multi-modal medical images of the head. We use image augmentation methods to create synthetic images to supplement an existing dataset. We use a validated method of registration (one plus one evolutionary optimization) to generate ground truth data and use the symmetry of the human head as an initial alignment to aid the optimization process. Before performing the registration, we also use a classification model to identify the imaging modality (MRI and CT) in order to determine the order of input for the registration to make the approach fully automatic. Then, we combine deep and conventional machine learning methods to predict the transformation/registration parameters. We show that the proposed methods outperform similar existing methods on publicly available MRI and CT images of the head. Lastly, we introduce a deep learning framework to perform brain tumor segmentation. To achieve this, we present a method of enhancing an existing MRI dataset by generating synthetic corresponding CT images, where prior domain knowledge of the registration method is applied to get a paired multi-modal ground-truth dataset. We fine-tune our network architecture and training strategies to segment brain tumors and show that the proposed model outperforms similar existing methods.
-
ItemExtending the application of virtual reality simulation in temporal bone anatomy and advanced surgical training( 2021)Cochlear implant surgery has a strong foundation in the treatment of profound hearing loss. In the last decade, there have been marked developments in technology which have enabled the expansion of eligibility criteria. Cochlear implantation is now offered to patients with residual low-frequency hearing or those with unilateral or asymmetrical hearing loss. It is evidenced that to optimise hearing outcomes for patients with residual hearing, it is necessary to reduce the trauma during the insertion of the cochlear implant, in particular by precise surgical technique. Current temporal bone surgery training, including cochlear implant surgery, is based on an apprenticeship model, where registrars observe and practice with consultant supervision. Prior to performing surgery on a patient, it is common practice to perform cadaveric temporal bone dissections. In addition to concern regarding decreasing availability of cadaveric temporal bones are financial constraints and regulations that reduce teaching time available in surgery. The generally low caseload, specifically relevant to cochlear implant surgery, minimises the opportunities for apprenticeship training. As such, this traditional model of training is not maintainable. Similar pressures face the training of anatomy of the temporal bone to medical students and junior doctors. While otologic presentations make a sizable proportion of presentations to emergency departments and general practice, due to the reduction in medical school training time, education in otology is in decline. Virtual reality (VR) surgical training is an attractive adjunct to the current training pathway as it provides a cost effective platform where risk-free, repetitive practice is readily available. VR also has several unique benefits. By presenting automated feedback, VR training allows for self-directed learning. In addition, automated assessment tools have been validated to objectively measure performance. While the effectiveness of VR simulation for mastoidectomy training has previously been well established, to the author’s knowledge there have been no VR simulators adapted to teach more complex temporal bone surgery such as cochlear implant surgery, or clinically oriented temporal bone anatomy. The aims of this thesis were: 1) to determine the viability of expanding the role of VR simulation in otology, including anatomy education and advanced temporal bone surgery and 2) to explore patient factors that relate to surgical technique in cochlear implant surgery. To these ends, several investigations were performed. Firstly, a randomised control trial was conducted to determine whether a clinically oriented VR temporal bone simulator module improved anatomy knowledge of medical students and junior doctors. Participants were randomly allocated to three groups of differing display modality: stereoscopic 3D, monoscopic 3D and 2D presentations. The participants completed a pre-tutorial questionnaire before working through the self-guided tutorial. The module was followed by a post-tutorial questionnaire and a retention questionnaire at 6 weeks. The questionnaires assessed factual anatomic knowledge, spatial relationships and clinically oriented knowledge as well as student’s perception of the display modality. It was observed that the module was effective in imparting factual knowledge in all modalities. The students exposed to the 3D technologies performed better in the spatial relationship and clinically oriented questions. The Stereoscopic 3D modality showed particular benefit for ease of use. Secondly, a training module specifically designed for the surgical approach to cochlear implant surgery was assessed with a prospective pre- and post-study of ENT registrars. All participants were exposed to the same training module that included concurrent and terminal feedback on temporal bones with a range of common anatomical variations. The participants’ performances were compared before and after the training. The assessment temporal bones used at the end of training consisted of a mirror image of the one used prior to training and one novel temporal bone. It was found that there was a significant improvement with a large effect size after training for both the previously encountered temporal bone and the novel temporal bone. Thirdly, a conceptual anatomical study was performed using the University of Melbourne temporal bone simulator data. The study explored patient factors that affect the surgical technique used in approaching cochlear implantation. In particular, the relationship of surgical preparation of the facial recess to the acceptable electrode trajectories into the cochlea. It was found that acceptable trajectories through a round window membrane approach most likely originated superiorly in the facial recess, adjacent to the facial nerve. Conversely, acceptable trajectories through a cochleostomy approach most likely originated inferiorly in the facial recess, adjacent to the junction of the facial nerve and chorda tympani. Furthermore, the skeletonisation of the facial recess was found to be critical in the preparation of the temporal bone for a cochleostomy approach. The results presented throughout the thesis will help guide medical educators in the areas of otology and cochlear implant surgery. These results suggest the viability of an expanded application of virtual reality temporal bone simulations to use in anatomy education and advanced temporal bone surgery training.