Abstract: FR-PO699
Quantitative Ultrasound for Glomerulosclerosis in Ex Vivo Murine Kidneys
Session Information
- Glomerular Diseases: From Inflammation to Fibrosis - II
November 03, 2023 | Location: Exhibit Hall, Pennsylvania Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Glomerular Diseases
- 1401 Glomerular Diseases: From Inflammation to Fibrosis
Authors
- Singla, Rohit, The University of British Columbia, Vancouver, British Columbia, Canada
- Lau, Yasmine Yuen Man, The University of British Columbia, Vancouver, British Columbia, Canada
- Hughes, Michael R., The University of British Columbia, Vancouver, British Columbia, Canada
- Hu, Ricky, The University of British Columbia, Vancouver, British Columbia, Canada
- Li, Yicong, The University of British Columbia, Vancouver, British Columbia, Canada
- Riazy, Maziar, The University of British Columbia, Vancouver, British Columbia, Canada
- Bissonnette, Mei Lin, The University of British Columbia, Vancouver, British Columbia, Canada
- Mcnagny, Kelly M., The University of British Columbia, Vancouver, British Columbia, Canada
- Rohling, Robert N., The University of British Columbia, Vancouver, British Columbia, Canada
- Nguan, Christopher, The University of British Columbia, Vancouver, British Columbia, Canada
Background
Focal segmental glomerulosclerosis (FSGS) is a condition that can lead to kidney function loss over time, making early detection and diagnosis essential for effective treatment and management. The glomeruli are the main source of ultrasound scattering in the kidney. This study proposes a novel approach to detecting FSGS using a multi-parametric quantitative ultrasound (QUS) approach in ex-vivo murine kidney models. QUS analyzes the spectral content of radiofrequency data in ultrasound to provide user- and system-independent quantifiable measurements, such as backscatter. We hypothesize that as FSGS progresses, there is a measurable increase in backscatter.
Methods
Five mice were recruited to each cohort of a podocalyxin knockout (PKO) model, a puromycin aminonucleoside (PAN) model and healthy controls. QUS was performed using a 128-element linear transducer at 15.625 MHz. From radiofrequency data, we extracted 16 parameters. To avoid overfitting, principal component analysis was used to create a reduced set of three parameters. These were inputs to a support vector machine for multi-class classification using a 5-fold cross validation scheme. Histopathologic analysis, performed by two expert pathologists, was used to determine the burden of FSGS as the ground truth.
Results
The average PKO case had 7% global segmental glomerulosclerosis and 19% FSGS, while the average PAN case had 20% global segmental glomerulosclerosis, 17% FSGS, and 7% total inflammation. Across all spatial locations or views, no significant differences in QUS parameters were found within a cohort. The mean classification accuracy was 80% across the three groups, with the mean precision, recall, and F1 scores being 0.86, 0.80, and 0.78, respectively. Misclassification only occurred in three PKO cases that were considered normal by the algorithm. The Nakagami scale parameter showed significant differences between the PAN model and the others, while the shape parameter showed significant differences between the PKO model and the others.
Conclusion
This study demonstrates that ultrasound measurements alone can effectively discriminate between healthy and diseased models of FSGS in mice. The results provide a foundation for further research into the quantification of kidney disease burden and its eventual use in humans.
Funding
- Government Support – Non-U.S.