Although artificial intelligence (AI) can adapt to changing conditions and achieve desired outcomes, how algorithms “understand” and adjust to inputs can be a mystery.

Lin et al. sought to uncover this “black box” in AI-controlled cold atmospheric plasma (CAP) treatments, an approach that induces apoptosis in diseased cells while preserving healthy ones. In previous work, they developed a machine learning (ML) system that predicts the post-treatment state of cancer cell targets and adjusts treatment accordingly. However, they didn’t know how the ML system achieved this outcome without an understanding of specific plasma parameters.

Using an AI-based optical emission spectroscopy (OES) spectra translation algorithm, the authors reverse engineered real-time chemical accumulations above cell medium surfaces. They found that, despite changing conditions, the ML algorithm alters experimental parameters to achieve the same therapeutic outcomes. The application of a Fourier transformation on OES spectra and chemical kinetics analysis revealed how the ML algorithm independently captured additional layers of physics information relying solely on cell viability status, without human input of this information, to achieve the precision and reliability of their AI-controlled CAP model.

“Beyond plasma medicine, similar approaches could advance machine learning-based control in fields like electric propulsion for satellites, plasma-based microfabrication, fusion reactor management, and many other plasma applications” said author Michael Keidar.

Next, the team looks to extend the scope of control that was demonstrated in this paper.

“Instead of limiting the AI to adjusting treatment duration, we plan to authorize and train the AI to control multiple plasma parameters simultaneously, including voltage, gas flow rate, and even additional external electric fields,” said author Li Lin. “In doing so, we aim to tailor therapy to the specific needs of each patient.”

Source: “Low-temperature plasma adaptation in the course of machine learning controls of plasma medicine,” by Li Lin, Qihui Wang, Zichao Hou, Michael Keidar, Physics of Plasmas (2025). The article can be accessed at https://doi.org/10.1063/5.0274614 .

The adage when it comes to public safety has been “if you see something, say something.” Ohio is now employing a new tool where you can say something to an interactive artificial intelligence chatbot; an app that allows people to submit information about potential criminal activity.
 
Ohio Department of Public Safety Director Andy Wilson said the multi-lingual app Safeguard Ohio can allow anyone to upload video, audio, and photos of suspicious activity. Then it lets artificial intelligence to take it from there.
 
“Because AI is involved, it asks the follow-up questions,” Wilson said. “It asks basically everything that needs to be gathered from an informational point of view to get what we need to, number one, understand what’s going on and get it to the right folks.”
 
Users can select from eight categories to report a tip. Those include drug-related activity, human trafficking, terrorism, school threats, and crimes against children.

“People can submit suspicious activity reports using this bot, using this app, sending this information into homeland security and we will get it where it needs to go,” Wilson said.

Ohio Homeland Security (OHS) Director Mark Porter said up to this point, people who want to report suspicious activity would have to call or go to a static form online where they could enter information. He said authorities had seen a decrease in the number of reports over time, getting an average of 30 tips per month until Aug. 6. That’s when the new app went online.

“In the last 30 days, our numbers have tripled in what we are getting,” Porter said. He attributed the increase to the app’s capability to process multiple languages and younger people being more likely to file information using an app and chatbot.

Wilson said reports made via the app can still be made anonymously. But emergencies need to be handled as they always have been.

“This isn’t a substitute for 911. What this is is to catch more of the suspicious activity, not the imminent ‘Hey something is going down,’ but ‘my roommate has a manifesto’ or ‘I saw this person online basically threaten to kill so and so.’ That kind of stuff,” Wilson said. “The AI chatbot will direct the user in case of an emergency, something that’s an emergency or imminent, to call 911.”

Ohio Homeland Security paid approximately $200,000 to the software company Vigiliti for the initial development of the Safeguard Ohio chatbot, backend dashboard for OHS staff, and compatibility with OHS’s current case management system. OHS also signed a two-year contract for $250,000 per year with the company for maintenance of the system and 24/7 access to help resolve any technical issues.

Introduction
Applications of AI in emergency medicine
Benefits of AI in emergency care
Challenges and limitations
Conclusions
References
Further reading

Artificial intelligence is transforming emergency medicine by enhancing triage, diagnosis, and resource management, while also facing challenges related to ethics, bias, and regulation. This article explores its applications, benefits, and limitations in real-world clinical care.

Image Credit: JHEVPhoto / Shutterstock.com

Introduction

Artificial intelligence (AI) is an interdisciplinary field that integrates computer science, mathematics, and related disciplines to create algorithms that can perform tasks conventionally restricted to human intelligence. AI algorithms utilize data-driven analysis, probabilistic modelling, and iterative optimization to learn, solve problems, and make decisions.¹

Unprecedented computational power, widely available and open-access electronic health data, as well as algorithmic breakthroughs, are rapidly transitioning AI from a conceptual technology to an integrated component of modern healthcare.¹ Despite projected growth of the global AI healthcare market, its incorporation into clinical practice remains limited due to the relative nascency of this technology and lack of standardization.²

In emergency medicine, AI has gained traction not only in clinical decision support (CDS) but also in digital twin modeling of patients, predictive analytics for emergency department (ED) flow, and integration with prehospital emergency medical services (EMS).^3,8,9

Additionally, recent primers emphasize the importance of familiarizing nonexpert clinicians with AI principles, terminology, and limitations to support safe and informed adoption.¹¹

Applications of AI in emergency medicine

AI-driven triage algorithms can analyze large datasets without bias and with significantly greater depth than conventional models, enabling clinicians to prioritize patients more effectively compared to traditional methods.⁵ In fact, machine learning models consistently demonstrate superior discrimination and performance capabilities for predicting emergency outcomes like hospital admission or intensive care unit (ICU) transfer and conditions like stroke, sepsis, and myocardial infarction.^4,5

Medical imaging and the interpretation of these images are among the most mature applications of AI, as numerous deep learning algorithms have been trained to analyze X-rays, computed tomography (CT) scans, and ultrasound images.¹ For these applications, AI technologies have successfully detected abnormalities like intracranial hemorrhage, fractures, and pneumothorax with high accuracy to support clinicians and reduce conventional diagnostic delays.¹ Explainable AI (XAI) methods are increasingly being incorporated into these models to enhance clinician trust by making diagnostic outputs more interpretable.^7,11

AI-powered CDS systems have also been developed to integrate real-time data from electronic health records (EHRs) and provide timely recommendations.¹ For example, AI models have been used to analyze electrocardiograms (ECGs) to predict impending cardiac arrest. Machine learning-assisted alerts have also been shown to improve the time to antibiotic administration.¹ More recently, scoping reviews highlight that CDS tools in emergency departments have been used to improve sepsis management, diagnostic accuracy, and disposition planning.³ Published case examples include Duke’s “Sepsis Watch” system and Viz.ai for subdural hematoma detection, which illustrate real-world clinical adoption.¹¹

AI-based predictive analytics can mitigate ED crowding by forecasting patient arrivals and anticipating surges. This application of AI allows hospitals to transition away from a reactive to a proactive staffing model that ensures the optimal allocation of limited resources like beds.^1,6

AI-powered symptom checkers and chatbots can simultaneously guide patients in self-assessing the urgency of their condition. Emergency dispatchers can also utilize natural language processing to recognize conditions, such as out-of-hospital cardiac arrest, faster and more accurately, despite limitations in first-responder knowledge.¹ EMS applications include AI-driven decision support for ambulance routing, prehospital risk stratification, and remote monitoring to improve patient outcomes before hospital arrival.^6,11

Another emerging domain is the use of digital twins, virtual patient models that simulate disease progression and treatment response, which could help personalize emergency care interventions and optimize resource use.⁹

Benefits of AI in emergency care

AI algorithms can rapidly process and synthesize vast quantities of data, thereby leading to faster and more precise assessments.⁴ This significantly reduces conventional image interpretation delays, with some AI models demonstrating performance superior to that of human specialists in specific tasks.¹

AI can provide several benefits to the existing public health infrastructure. By accurately predicting patient volume, AI can enable hospitals to better manage patient throughput, reduce system inefficiencies, alleviate overcrowding, and shorten patient wait times.⁶ These predictive tools also support disaster preparedness and surge capacity planning, strengthening system resilience.^4,5

For administrative purposes, AI can automate routine and time-consuming tasks using ambient listening technologies and generative AI-based clinical summaries. The adoption of AI into these aspects of healthcare has the potential to reduce clinician burnout, as well as improve both patient satisfaction and provider well-being.^1,4 Furthermore, AI can facilitate continuous quality improvement by identifying patterns in adverse events and enabling evidence-based policy development.^7,11

High-tech hospital uses artificial intelligence in patient carePlay