Artificially intelligent software has been developed to advance medical treatments utilizing cold atmospheric plasma (CAP), a jet of electrified gas. The software, developed by researchers at the Princeton Collaborative Low-Temperature Plasma Research Facility (PCRF), combines machine learning and physics to predict the chemical composition of CAP emissions. This breakthrough has significant implications for treating cancer, promoting tissue growth, and sterilizing surfaces.
Predicting CAP emissions with AI
The software, known as a physics-informed neural network (PINN), learned to forecast the diverse array of chemicals emitted by CAP jets. It accomplished this by analyzing data collected during real-world experiments and integrating the fundamental principles of physics. This innovative application of AI, termed machine learning, enables the system to continually improve its predictions based on provided information.
Cold atmospheric plasma: A multifaceted tool
Cold atmospheric plasma (CAP) has been employed in various medical applications, including cancer cell eradication, wound healing, and bacteria elimination on food surfaces. However, the precise mechanisms behind these effects have not been fully understood by scientists.
According to Yevgeny Raitses, a managing principal research physicist at the U.S. Department of Energy’s Princeton Plasma Physics Laboratory (PPPL), the AI-driven software marks a significant step toward comprehending how and why CAP jets work. This newfound understanding could lead to refined and more effective utilization of CAP technology in medical treatments.
A collaborative effort
The project was a collaborative effort between researchers at PPPL and George Washington University (GWU), under the banner of the Princeton Collaborative Low-Temperature Plasma Research Facility. PPPL, renowned for its pioneering work in plasma research, has expanded its mission to incorporate AI applications in fields such as medicine and manufacturing.
Sophia Gershman, a lead PPPL research engineer, emphasized the difficulty in accurately determining the chemical composition of CAP jets due to the need to consider interactions on a nanosecond timescale. Machine learning offers a solution to this complexity, allowing for precise calculations that were previously practically impossible.
Data generation and training
The project began with a small dataset obtained through Fourier-transform infrared absorption spectroscopy. This initial data served as a foundation for generating a more extensive dataset. Inspired by natural selection, an evolutionary algorithm was employed to train the neural network. Through successive iterations, the AI system improved its accuracy by selecting the best datasets and refining its predictions.
Accurate calculations for CAP Jets
The team successfully developed a software solution capable of accurately calculating chemical concentrations, gas temperature, electron temperature, and electron concentration within cold atmospheric plasma jets. This achievement is particularly notable because CAP jets can have extremely hot electrons while maintaining near-room temperature for other particles, making it suitable for medical treatments.
Personalized plasma treatment on the horizon
Michael Keidar, a professor of engineering at GWU, highlighted the long-term goal of implementing real-time calculations to optimize CAP treatment during medical procedures. Keidar is currently working on a prototype for a “plasma adaptive” device that could be personalized to each patient’s unique needs. By monitoring patient responses and utilizing machine learning, the device could adjust plasma settings to maximize effectiveness.
While this study examined the CAP jet’s chemical composition over time, it focused on a single point in space. Future research will need to expand the investigation to consider multiple points along the jet’s output stream. Additionally, integrating the surfaces treated by the plasma into the analysis will be crucial to understand how it impacts the chemical composition at the treatment site.
This groundbreaking research, funded by the U.S. Department of Energy and the Princeton Collaborative Research Facility, paves the way for improved medical treatments using cold atmospheric plasma. With the integration of AI, the potential for personalized and optimized plasma treatments offers hope for more effective healthcare solutions.