Guang Lin leading machine learning research funded by NIH

02-15-2024

 

Advanced imaging and machine learning may improve the way doctors detect and treat cardiomyopathy, the leading cause of death for children with Duchenne muscular dystrophy (DMD) and Becker muscular dystrophy (BMD).

Purdue mathematics and mechanical engineering professor Guang Lin is co-principal investigator of a $3.6M grant from the National Institutes of Health (NIH) to change the grim outlook for these patients.

Co-principal investigator Craig Goergen, the Leslie A. Geddes Professor in the Purdue University Weldon School of Biomedical Engineering, says, “The progression of cardiac disease is variable and poorly understood in DMD and BMD patients. There are no blood or imaging biomarkers that can predict the pace of progression or the risk of early mortality in these patients. Given this and the variability between patients, clinical trials are challenging.”

To meet that challenge, DMD Cardiac Care Consortium and medical centers around the country will partner to generate a national registry of DMD and BMD patients. With the clinical data and high-fidelity cardiac magnetic images from the registry, Lin’s group will use novel, data-driven, personalized machine learning models to determine cardiovascular measures associated with cardiomyopathy.

Lin’s focus is to develop advanced machine learning tools to discover innovative and interpretable, data-driven, population-based, and personalized causal models to determine the risk of mortality and rapid progression for individual patients. Says Goergen, “The results will provide clinicians all over the world with a method to assess their patient’s risk in real time and improve the quality of life for these pediatric patients.”

Participants in the multi-center collaboration include:

• Purdue University
  • Craig Goergen, PhD, Biomedical Engineering (PI)
  • Guang Lin, PhD, Mathematics and Mechanical Engineering (Co-PI)
• Riley Hospital for Children, Indiana University School of Medicine
  • Larry Markham, MD, Pediatric Cardiology
• Vanderbilt University Medical Center
  • Jonathan Soslow, MD, MSCI, Pediatric Cardiology
• Nationwide Children’s Hospital

PROJECT SUMMARY/ABSTRACT

Duchenne and Becker muscular dystrophy (DMD/BMD) are devastating diseases with no cure resulting in loss of ambulation, respiratory failure, cardiomyopathy, and premature death. Dystrophin associated cardiomyopathy (defined here as CM) is the leading cause of death in DMD/BMD, and an under-studied concern in DMD and BMD mutation carriers (MDC). CM progression is variable and poorly described in the current era. There are no blood or imaging biomarkers that can predict the pace of progression or the risk of early mortality. More importantly, there are no established cardiac outcome measures.

Novel, targeted therapeutics are necessary to treat CM, but these significant knowledge gaps make clinical trials challenging. A better understanding of DMD/BMD/MDC cardiovascular disease progression and the identification of surrogate outcome measures are critical for the field to advance. To address these obstacles, we propose to leverage the Duchenne muscular dystrophy cardiac care consortium (DMDCCC). Created with grants from the NHLBI and the FDA, this consortium consists of eight high-volume sites with similar DMD/BMD/MDC cardiovascular treatment and diagnostic protocols, including surveillance CMR imaging every 1-2 years. This proposal will create a comprehensive prospective registry of DMD/BMD/MDC patients with meticulously collected clinical data and cardiac magnetic resonance (CMR) images; we anticipate enrollment of 950 patients with over 4000 CMR studies. This cohort will be used to better define the progression of CM and to determine associations with mortality. The central hypothesis of our proposal is that integrated statistical modeling based on advanced imaging can improve prediction of CM progression and mortality. Aim 1 will create a comprehensive cohort of DMD/BMD/MDC patients and model the progression of CM. Aim 2 will determine cardiovascular measures that are associated with CM mortality or rapid progression using novel, data-driven, personalized machine learning models. Aim 3 will create a portal for DMD/BMD/MDC centers to determine patient risk. This multi-PI proposal leverages expertise in clinical care, cardiac imaging, biomedical engineering, complex image analysis, and neural networks. To our knowledge, this study will create the largest cohort of DMD/BMD/MDC patients with CMR images, allowing for a better understanding of CM progression and identifying biomarkers that associate with poor outcomes. The resulting risk portal will provide clinicians all over the world with a method to assess their patient’s risk in real time, allowing intensification of therapy for those deemed high risk. By building on prior productive collaborations, particularly that of the DMDCCC, this proposal will expand our understanding of CM, improving clinical care and future cardiac-specific therapeutic trials.

Research reported in this publication was supported by the National Heart, Lung, And Blood Institute of the National Institutes of Health under Award Number R01HL167969. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.