Electronic Thesis and Dissertation Repository

Thesis Format

Integrated Article

Degree

Master of Engineering Science

Program

Electrical and Computer Engineering

Supervisor

Ladak, Hanif M

2nd Supervisor

Agrawal, Sumit K

Joint Supervisor

Abstract

Computed tomography (CT) scans are acquired prior to cochlear implant (CI) surgery. Three-dimensional segmentations of the inner ear (IE) and round window (RW) based on clinical CTs can improve the CI procedure. Software pipelines are presented here which employ convolutional neural networks to automatically segment the IE and RW. The first pipeline produces high resolution segmentations of the IE and RW in tightly cropped CTs. Mean IE Dice score and RW centroid error were 0.88, 0.57mm and 0.93, 0.18mm in implanted and non-implanted samples, respectively. The second pipeline automatically segments the IE in large field of view CTs of any rotational orientation. Mean Dice scores of 0.83 and 0.89 were achieved in implanted and non-implanted samples, respectively. This is the first known study to present automated segmentations of the IE and RW in CTs with CIs. The pipelines provide quick and accurate segmentations.

Summary for Lay Audience

Cochlear implants are a revolutionary invention which provides hearing to deaf people who have the most common form of hearing loss. Cochlear implants are surgically inserted into the inner ear, a tiny bony labyrinth within the temporal bone of the skull. The round window membrane is a natural opening from the middle ear to the inner ear and is the most frequently used entry point when surgically inserting a cochlear implant into the inner ear. Computed tomography scans are 3D medical images which are acquired prior to the cochlear implant procedure to give surgeons an understanding of the patient's anatomy. Structures within a computed tomography scan, such as the inner ear, can be labeled by hand or by using an algorithm; these labels represent 3D models of the structures and are referred to as segmentations. Segmentations of a patient’s inner ear and round window are useful for the following applications: surgeons can build an enhanced understanding of a patient's anatomy, surgical rehearsal and training platforms could integrate specific patient cases, drilling locations and electrode insertion angles can be determined for robot-assisted cochlear implant surgery, and the creation of customized cochlear implant frequency maps could be automated. Manual labeling of medical imaging data is not feasible for integration to clinical practice as it requires expert knowledge and is extremely time consuming. This work presents an automated approach which quickly and accurately segments the inner ear and round window based on any computed tomography scan with minimal user input. The automated segmentation approach utilizes a supervised machine learning algorithm (convolutional neural networks), and post-processing scripts which finalize the output segmentation. The automated approach was validated by use of clinical and cadaveric scans. Segmentations produced by the network were compared to manual expert segmentations as well as segmentations of high-resolution imaging acquisitions. This is the first known study to present an automated approach which segments the inner ear and round window in scans with cochlear implants. The approach presented in this study utilized a wide variety of imaging acquisitions, resulting in a robust model which has promising clinical implementations.

Download

Included in

Biomedical Commons

Share

COinS