ADPKD Image Repository Core

Weill Medical Coll Of Cornell Univ, New York NY

Investigators

Abstract

ADPKD IMAGE REPOSITORY CORE Project Summary: Clinical and experimental models of autosomal dominant polycystic kidney disease (ADPKD) employ imaging biomarkers (e.g., total kidney volume [TKV]) to assess prognosis, determine clinical trial eligibility, and treatment response monitoring. However, existing TKV measurement methods have poor reproducibility, reducing statistical power and increasing required sample sizes and trial durations. Moreover, current risk-stratification models (e.g., Mayo Clinic Imaging Classification [MIC], PRO-PKD, the ADPKD Outcomes Model) are limited in accurately predicting ADPKD disease progression. We have developed automated, machine-learning-based image processing methods that enhance the accuracy and efficiency of total kidney volume measurement as well as assessment of other ADPKD risk factors like visceral adiposity, cyst fraction, and hemorrhagic renal cysts. As ADPKD is a systemic condition, our PKD phenotyping also assesses the effects of APDKD on liver, spleen, pancreas, heart (if within field of view), seminal vesicles, and prostate gland; and enables quantitation of cysts and cyst volumes, extrarenal fluids (pleural effusion, free pelvic fluid), fat (subcutaneous, peritoneal, peri-renal, cardiac) and muscles. Moreover, dynamic serial MRI measurements of bladder volume increasing from one sequence to the next provide a spot assessment of urine output. Stomach segmentations can evaluate gastric confinement. This focus on extrarenal complications of ADPKD also has the potential to correct errors in several biomarkers routinely used to assess health, including body mass index (which can be confounded by weight of cyst fluid), muscle mass (potentially improving creatinine-based calculations of eGFR) and identifying sarcopenia. Furthermore, serial MRIs analyzed by our technique predict TKV trajectory more accurately than MIC. Our comprehensive PKD Image Phenotyping toolkit will be extended in this proposal to pediatric subjects from the PKD-RRC and adults with early disease including test-retest validation of all imaging biomarkers in early disease. All images entered into the Core database will undergo rigorous quality control. The difficulty of obtaining accurate PKD image phenotyping manually is well known among imaging specialists. It requires meticulous, detailed work by expert observers that is time-consuming, expensive, and impractical. Our automated PKD Image Phenotyping resource will be shared gratis with new and established investigators in collaboration with the Administrative Core and the PKD Central Coordinating Site (CCS). We anticipate advances in the understanding of ADPKD in patients, as well as accelerated development and testing of candidate therapies becoming possible by providing PKD scientists and clinicians with easily accessible, accurate, comprehensive, and reproducible PKD Image phenotyping.