CAREER: A deep explainable artificial intelligent framework for electrical impedance myography

Neuromuscular disorders affect millions of individuals worldwide, yet efficient tools to accelerate diagnosis and assess therapeutic interventions are currently lacking. Existing methods for evaluating muscle health face significant limitations, including clinical impracticality due to cumbersome procedures, reliance on highly trained personnel, high costs, and safety concerns stemming from associated pain or the use of ionizing radiation.

The emergence of electrical impedance myography (EIM) offers a promising avenue for assessing muscle health. EIM is sensitive to changes in muscle structure and composition brought about by a variety of neuromuscular disorders as well as by disuse, producing unique disease signatures that will vary with muscle status. Thus, EIM analysis can provide a method to rapidly, quantitatively, and reliably diagnose and monitor neuromuscular diseases at the bedside, act as a tool to help tailor care for individual patients and streamline and improve clinical drug trials.

This CAREER project integrates research with educational outreach by offering students hands-on experience in innovative translational research. This is interlaced with a long-term educational objective of mentoring new generations of students by providing them with experiences in cutting-edge research, developing and implementing activity-based style courses to motivate students’ self-learning in the classroom, encouraging students to choose a science, technology, engineering or math (STEM) degree by participating in research, and assisting undergraduate students in their own translational research efforts.

This CAREER project will establish the scientific foundations of future generation EIM tools and enhance diagnostic accuracy by integrating artificial intelligence algorithms with simulation and analytical methods to extract quantitative muscle insights that are currently inaccessible. The tools developed and data collected are expected to lead to a deeper understanding of the role played by muscle electrical properties in EIM, understanding that is needed for the development of new and more accurate EIM tools for evaluating neuromuscular disorders (NMD).

Research objectives include (1) developing a physics-informed analytical and simulation framework to model the entangled multicellular architecture underlying tissues and automate the extraction of relevant physical and biological information from EIM data, (2) assessing EIM biological variability in silico, and (3) evaluating the robustness of models generating EIM data. In silico simulations and ex vivo measurements will provide proof of principle to optimally determine the minimal yet sufficiently biophysical relevant mechanisms needed to build robust virtual EIM predictions for healthy and prototypical diseased conditions necessary to interpret EIM outcomes in patients.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.