Spatiotemporal and Deep Learning Analysis of Cardiac Imaging for Predictive Risk Stratification in Duchenne Muscular Dystrophy

点击翻译按钮获取中文摘要

PROJECT SUMMARY/ABSTRACT Heart disease is the leading cause of death for individuals with Duchenne muscular dystrophy (DMD). DMD is a devastating and progressive neuromuscular disease with no known cure. This X-linked genetic disorder affects nearly 1 in 5000 boys and manifests as debilitating muscle weakness and progressive cardiomyopathy (CM). While CM in some individuals with DMD progresses rapidly and fatally in their teenage years, others can live relatively symptom-free into their thirties or forties. Early identification and treatment can improve quality and length of life, but currently, there are no standard imaging biomarkers that can detect early onset or rapidly progressing DMD CM. Additionally, research in this area has lagged due to small population study sizes and limited standardized imaging data. To that end, this project will utilize the largest standardized imaging data registry of DMD CM created through the collaboration of 6 of the largest medical institutions with DMD CM expertise. Following the objective set up by the National Heart, Lung, and Blood Institute (NHLBI) to “develop and optimize novel diagnostic and therapeutic strategies to prevent, treat, and cure HLBS diseases” we propose the following Aim: 1) Identify imaging biomarkers of DMD CM onset and progression using novel image analysis. Utilizing recently developed methods of spatiotemporal analysis of 4D (3D plus time) cardiac imaging data, we can evaluate localized kinematic parameters in the heart that may be sensitive to subtle changes in disease physiology. With this project we also follow a second major objective set forth by NHBLI to “Leverage emerging opportunities in data science to open new frontiers in HLBS research” through another Aim: 2) Apply deep learning neural network to DMD registry to evaluate CM onset and progression. Utilizing the large DMD CM imaging registry, we will apply deep learning techniques for automated segmentation and analysis of cardiac parameters to evaluate patterns of early-onset and rapid progression. These results will help to bridge a crucial gap in optimizing clinical treatment for a devastating pediatric disease and pave the way for future research and innovation through the definition of robust imaging This fellowship research and training will be carried out at Purdue University under the direct mentorship of Craig Goergen, PhD who is a leading expert in cardiovascular imaging and biomechanics research and at Indiana University School of Medicine with Larry Markham, MD, Division Chief of Pediatric Cardiology and renowned physician scientist with a focus on DMD CM. Guang Lin, PhD (Purdue University), a data science expert, will provide expertise in the deep learning algorithm development. Han Kor, MD (Nationwide Children’s Hospital), May Ling Mah, MD (Nationwide Children’s Hospital), and Jonathan Soslow, MD (Vanderbilt School of Medicine) are all practicing pediatric cardiologists with expertise in DMD cardiac imaging and will provide access to data, clinical insight, training, and mentorship for this project. biomarkers and clinical trial endpoints.