Machine learning-aided evaluation of oxidative strength of cold atmospheric plasma-treated water

Plasma medicine is gaining attraction in the medical field, particularly the use of cold atmospheric plasma (CAP) in biomedicine. The chemistry of the plasma is complex, and the reactive oxygen species (ROS) within it are the basis for the biological effect of CAP on the target. Understanding how the oxidative power of ROS responds to diverse plasma parameters is vital for standardizing the effective application of CAP. The proven applicability of machine learning (ML) in the field of medicine is encouraging, as it can also be applied in the field of plasma medicine to correlate the oxidative strength of plasma-treated water (PTW) according to different parameters. In this study, plasma-treated water was mixed with potassium iodide-starch reagent for color formation that could be linked to the oxidative capacity of PTW. Corresponding images were captured resulting from the exposure of the color-forming agent to water treated with plasma for different time points. Several ML models were trained to distinguish the color changes sourced by the oxidative strength of ROS. The AdaBoost Classifier (ABC) algorithm demonstrated better performance among the classification models used by extracting color-based features from the images. Our results, with a test accuracy of 63.5%, might carry a potential for future standardization in the field of plasma medicine with an automated system that can be created to interpret the oxidative properties of ROS in different plasma treatment parameters via ML.

Recognizing Cold Atmospheric Plasma Plume Using Computer Vision

Over the last three decades, cold atmospheric plasma (CAP) has been heavily investigated in a wide range of biological applications, including wound healing, microorganism sterilization, and cancer treatment. Atmospheric pressure plasma jets (APPJs) are the most common plasma sources in plasma medicine. An APPJ’s size determines its application range and approach in treatment. In this study, we demonstrated the real-time recognition of an APPJ’s plasma plume output using computer vision (CV), dramatically improving the measurement speed compared to the traditional method of using the naked eye. Our work provides a framework to monitor an aspect of an APPJ’s performance in real time, which is a necessary step to achieving an intelligent CAP source.