Date of Award

5-2017

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Legacy Department

Bioengineering

Committee Member

Dr. David Kwartowitz, Committee Chair

Committee Member

Dr. Delphine Dean

Committee Member

Dr. Donald House

Committee Member

Dr. Bruce Gao

Abstract

Autosomal dominant polycystic kidney disease (ADPKD) is the fourth most common cause of kidney transplant worldwide accounting for 7-10% of all cases. Although ADPKD usually progresses over many decades, accurate risk prediction is an important task. Identifying patients with progressive disease is vital to providing new treatments being developed and enable them to enter clinical trials for new therapy. Among other factors, total kidney volume (TKV) is a major biomarker predicting the progression of ADPKD. Consortium for Radiologic Imaging Studies in Polycystic Kidney Disease (CRISP) have shown that TKV is an early, and accurate measure of cystic burden and likely growth rate. It is strongly associated with loss of renal function. While ultrasound (US) has proven as an excellent tool for diagnosing the disease; monitoring short-term changes using ultrasound has been shown to not be accurate. This is attributed to high operator variability and reproducibility as compared to tomographic modalities such as CT and MR (Gold standard). Ultrasound has emerged as one of the standout modalities for intra-procedural imaging and with methods for spatial localization has afforded us the ability to track 2D ultrasound in the physical space in which it is being used. In addition to this, the vast amount of recorded tomographic data can be used to generate statistical shape models that allow us to extract clinical value from archived image sets. Renal volumetry is of great interest in the management of chronic kidney diseases (CKD). In this work, we have implemented a tracked ultrasound system and developed a statistical shape model of the kidney. We utilize the tracked ultrasound to acquire a stack of slices that are able to capture the region of interest, in our case kidney phantoms, and reconstruct 3D volume from spatially localized 2D slices. Approximate shape data is then extracted from this 3D volume using manual segmentation of the organ and a shape model is fit to this data. This generates an instance from the shape model that best represents the scanned phantom and volume calculation is done on this instance. We observe that we can calculate the volume to within 10% error in estimation when compared to the gold standard volume of the phantom.

Share

COinS