We propose to establish an Exploratory Center for Interdisciplinary Research in Benign Urology at the Children?s Hospital of Philadelphia (CHOP) and the University of Pennsylvania (Penn), the central mission of which is to apply machine learning to improve the understanding of the pathophysiology, diagnosis, risk stratification, and prediction of treatment responses of benign urological disease among children and adults. The proposed CHOP/Penn Center for Machine Learning in Urology (CMLU) addresses critical structural and scientific barriers that impede the development of new treatments and the effective application of existing treatments for benign urologic disease across the lifespan. Structurally, urologic research occurs in silos, with little interaction among investigators that study different diseases or different populations (e.g. pediatric and adult). Scientifically, analysis of imaging and other types complex data is limited by inter-observer variability, and incomplete utilization of available information. This proposal overcomes these barriers by applying cutting-edge approaches in machine learning to analyze CT images that are routinely obtained for evaluation of individuals with kidney stone disease. Central to the CHOP/Penn CMLU is the partnership of urologists and experts in machine learning, which will bring a new approach to generating knowledge that advances research and clinical care. In addition, the CMLU will expand the urologic research community by providing a research platform and standalone machine learning executables that could be applied to other datasets. The Center?s mission will be achieved through the following Aims, with progress assessed through systematic evaluation:
Aim 1. To expand the research base investigating benign urological disease. We will establish a community with the research base, particularly with the KURe, UroEpi programs, other P20 Centers, and O?Brien Centers. We will build this community by providing mini-coaching clinics to facilitate application of machine learning to individual projects, developing an educational hub for synchronous and asynchronous engagement with the research base, and making freely available all source codes and standalone executables for all machine learning tools.
Aim 2. To improve prediction of ureteral stone passage using machine learning of CT images. The CMLU has developed deep learning methods that segment and automate measurement of urinary stones and adjacent renal anatomy. In the Research Project, we will compare these methods to existing segmentation methods and the current gold standard of manual measurement. We will then extract informative features from thousands of CT scans to predict the probability of spontaneous passage of ureteral stones for children and adults evaluated in the CHOP and Penn healthcare systems.
Aim 3. To foster collaboration in benign urological disease research across levels of training and centers through an Educational Enrichment Program. We will amplify interactions across institutions and engage investigators locally and nationally by providing summer research internships, and interinstitutional exchange program, and an annual research symposium.
The proposed CHOP/Penn O?Brien Center for Machine Learning in Urology addresses critical structural and scientific barriers that impede development of new treatments and the effective application of existing treatments for benign urologic disease across the lifespan. This application overcomes these barriers by applying cutting- edge approaches in machine learning to analyze complex imaging data for individuals with kidney stone disease.The Center?s strategic vision of using machine learning to generate knowledge that improves diagnosis, risk stratification strategies, and prediction of outcomes among children and adults will be achieved through the implementation of a Educational Enrichment Program and a Research Project.
