Zhang, Zhenwei
(2021)
Investigations of videofluoroscopy via machine learning: Novel ways for swallowing disorders assessment.
Doctoral Dissertation, University of Pittsburgh.
(Unpublished)
Abstract
Videofluoroscopic swallow studies are widely used in clinical and research settings to assess swallow function and to determine physiological impairments, diet recommendations, and treatment goals for people with dysphagia. It can be used to analyze biomechanical events of swallowing, to differentiate between normal and disordered swallow function. It is also important for clinicians to understand the association between various possible physiological measures and penetration-aspiration, in order to determine the boundary values for these measures that can be validated for impaired swallows. In recent years, deep learning technique have achieved tremendous success in various medical imaging applications, including, but not limited to brain studies, disease diagnosis and prevention.
In this dissertation research, we attempted to further this research in two key areas. First, we evaluated the potential association between the trajectory of hyoid bone movement and the risk of airway penetration and aspiration during VFSS examination using generalized estimation equations. In addition, the model was built based on aspects of hyoid bone displacement to predict the extent of airway penetration. Second, we aimed to explore the potentials of deep learning techniques to address different dysphagia problems. These algorithms to automatically evaluate and assess VFSS dysphagia studies are highly sought after in the dysphagia clinical and scientific communities. To demonstrate the feasibility of deep learning techniques on VFSS, we computed and compared the state of art object detection networks for hyoid bone tracking algorithm, which was the first attempt to utilize deep learning techniques in the VFSS field. In physiologic measurements, scaling of images based on length of vertebrae bodies to compensate for size differences among different patients is a crucial component of the analysis. In order to detect key anatomical points needed for a routine swallowing assessment in real-time, we presented a novel two-stage convolutional neural network trained with missing annotations to localize and measure length of the vertebral bodies. Finally, we sought to measure the amount of residue remained in vallecular area. We implemented an ensemble method with several networks to segment and calculate the residue scale.
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
| Item Type: |
University of Pittsburgh ETD
|
| Status: |
Unpublished |
| Creators/Authors: |
|
| ETD Committee: |
|
| Date: |
3 September 2021 |
| Date Type: |
Publication |
| Defense Date: |
27 April 2021 |
| Approval Date: |
3 September 2021 |
| Submission Date: |
31 May 2021 |
| Access Restriction: |
2 year -- Restrict access to University of Pittsburgh for a period of 2 years. |
| Number of Pages: |
151 |
| Institution: |
University of Pittsburgh |
| Schools and Programs: |
Swanson School of Engineering > Electrical and Computer Engineering |
| Degree: |
PhD - Doctor of Philosophy |
| Thesis Type: |
Doctoral Dissertation |
| Refereed: |
Yes |
| Uncontrolled Keywords: |
dysphagia, machine learning, image processing, classi�cation, feature analysis |
| Date Deposited: |
03 Sep 2021 17:15 |
| Last Modified: |
03 Sep 2023 05:15 |
| URI: |
http://d-scholarship.pitt.edu/id/eprint/41199 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[feed]](/images/feed-icon-32x32.png){kind=icon}