Imagine a doctor being able to measure pain as precisely as a thermometer measures temperature. That’s the promise of AI in medicine, and it’s already taking shape through innovative research. Sungjoon Hong, a student at the College of Osteopathic Medicine (NYITCOM), is working on a project that uses artificial intelligence to assess the effectiveness of epidurals. By analyzing patterns that the human eye might miss, this approach turns subjective patient reports into objective, measurable data. Hong’s path into this field began during his first semester when he met Associate Professor Milan Toma, Ph.D., whose expertise lies in AI-assisted medical diagnostics. Their collaboration highlights a growing convergence of technology and healthcare.
This story is just one example of how AI in healthcare is reshaping medical practice. From diagnostics to treatment planning, artificial intelligence is helping clinicians make faster, more accurate decisions. The following sections explore seven specific ways that medical AI research is revolutionizing the field, turning once-futuristic ideas into practical tools for modern clinics.
1. Turning Subjective Pain into Objective Data with AI
Pain has always been a personal experience. When you tell a doctor your pain level on a scale of one to ten, that number is entirely based on how you feel in that moment. It is subjective, influenced by your mood, your tolerance, and your memory of past pain. This makes it difficult for clinicians to measure treatment effectiveness with precision. That is where AI in medicine is starting to change the game. Instead of relying on a patient’s report alone, machine learning models can now analyze physical signals to create an objective pain measurement.
Consider how an epidural works during labor. The nerve block causes blood vessels in the lower body to dilate, which increases the temperature of the patient’s feet. This is a measurable, physical response. Using thermal imaging in medicine, a researcher named Hong developed a machine learning model to automatically evaluate that temperature change in women who had received an epidural. The model segments the thermal image of the feet and analyzes the temperature shift over time. This turns a vague sensation into a clear, data-driven result. For you, the patient, this means AI-driven diagnostics could soon provide your care team with a real-time, unbiased readout of how well your pain relief is working — no guesswork required.
2. Improving Epidural Failure Detection Rates
That unbiased, real-time assessment is especially valuable in one area where current methods are surprisingly subjective: confirming that an epidural block is working. Epidural blocks are a common form of pain relief during childbirth and certain surgeries, but they fail up to 12 percent of the time. The standard test to check them involves a nurse or doctor applying a pinprick or an ice cube to your feet and asking if you feel anything. This approach relies heavily on patient feedback and the clinician’s interpretation — and it can easily miss a partial failure. Artificial intelligence in medicine offers a more reliable alternative. AI thermal analysis uses a camera to continuously monitor skin temperature patterns on the affected area. By analyzing these temperature changes over time, AI in anesthesiology can detect when an epidural is not providing adequate nerve block, often before you might notice any discomfort. This tool turns a subjective, intermittent check into a continuous, data-driven assessment, helping to lower the epidural failure rate and improving your overall experience.
3. How U-Net Deep Learning Processes Medical Thermal Images
That same thermal imaging approach gets an even sharper upgrade when paired with a specialized deep learning model called U-Net. While thermography captures temperature patterns, the real challenge is making sense of all that data quickly and accurately. That is where Hong’s integration of the U-Net architecture comes into play. U-Net was originally built for medical image segmentation—essentially, it is a deep learning model designed to pick out specific structures in scans, like organs or tissues, with pixel-level precision.
Hong adapted this U-Net architecture to focus on thermal images of feet for temperature analysis. Instead of a doctor manually tracing regions of interest or eyeballing subtle temperature shifts, the model automatically segments the image into clear, measurable zones. It acts as an automated assistant, reducing the manual interpretation workload and speeding up diagnosis. For you, this means less waiting for results and fewer chances of human error slipping through. The deep learning medical imaging process handles the heavy lifting, flagging abnormal patterns that might indicate circulation issues or nerve damage. It is a practical example of how AI in medicine can turn a complex thermal snapshot into a straightforward, actionable insight for your healthcare provider.
4. A Medical Student’s Rapid AI Research Achievements
That kind of practical, data-driven insight shows how AI in medicine can improve care. But behind these tools are people pushing the field forward—sometimes at an astonishing pace. Take Hong, a medical student who published three first-author papers in under a year. That is a remarkable output for any researcher, let alone someone still in medical school. His work focuses on the intersection of AI and medicine, driven by a deep interest in pain management and physical medicine. This is not just about academic accolades; it shows that medical student research can make a real impact in AI publications.
Hong’s rapid progress did not happen in a vacuum. Toma, who mentored him, described Hong as one of the most self-driven students he has worked with. Toma also maintained an open-door policy, encouraging anyone interested in research to step in and explore. That combination of personal initiative and a supportive environment allowed Hong to contribute meaningfully to pain management research. For you, this story highlights a key point: the next breakthrough in AI in medicine might come from someone just starting their career. It is a reminder that curiosity and dedication can accelerate learning and produce tangible results, even in a field as complex as healthcare.
5. Acknowledging AI’s Limitations in Medical Settings
That same passion for progress must be paired with a clear-eyed view of what AI can and cannot do. Hong, whose curiosity drove real results, also openly acknowledges the boundaries of AI in medicine. His primary commitment as a future physician is patient safety, which means recognizing these limits is not a weakness but a responsibility. For you, this highlights a key point: powerful as these tools are, they are not infallible. AI models need large, diverse datasets to avoid bias in healthcare; without them, results can be skewed or unreliable. Interpretability remains a challenge—some AI systems act as a “black box,” making it hard to explain how they reach a diagnosis. Clinical validation of AI is still an ongoing process to ensure these tools perform reliably outside controlled conditions.
Related reading: our post Salesforce Data Thefts Continue via Klue App offers more practical ideas on this.
Hong emphasizes that AI should augment, not replace, clinical judgment. This distinction matters for patient safety. When applying AI in medicine, think of it as a support tool that flags patterns or possibilities a human might miss. The final decision always rests with a qualified professional. Recognizing these AI limitations helps you use the technology more wisely. It avoids over-reliance and keeps the focus where it belongs: on the patient. Understanding what AI cannot do yet is just as valuable as celebrating what it can.
6. AI’s Role in Pain Management and Physical Medicine & Rehabilitation
Beyond surgical precision, another area where AI in medicine shines is in helping people recover function and manage chronic pain. Fields like pain management and Physical Medicine & Rehabilitation (PM&R) are built around this goal: helping patients regain mobility, reduce discomfort, and return to daily life. For professionals like Hong, the deep sense of fulfillment comes from guiding people through vulnerable moments. This human-centered passion naturally aligns with the analytical power of AI, which can enhance rehabilitation programs without replacing the human touch.
How AI Monitors Recovery Over Time
One of the biggest challenges in rehabilitation technology is tracking progress beyond a clinic visit. You might feel better one week and worse the next, but standard check-ins only capture snapshots. AI for long-term recovery changes that by analyzing data from wearables, motion sensors, and even medical imaging. Over weeks or months, the system can spot subtle trends — like gradual improvement in gait or increased range of motion — that you or your clinician might miss. This allows care plans to be adjusted in real time, making AI in pain management a practical tool for sustained, measurable progress. The technology doesn’t replace the rehabilitation specialist; it gives them a clearer, data-backed picture of how you are truly healing.
7. Expanding AI Applications to Other Medical Conditions
This same AI technique isn’t limited to epidurals alone. The combination of thermal imaging and a U-Net neural network can be adapted for other conditions where temperature changes signal trouble. For instance, AI in vascular disease could detect poor circulation by analyzing heat patterns in your limbs. Similarly, thermal imaging for neuropathy—common in diabetic patients—could spot early warning signs before you feel pain or numbness. The method’s strength lies in its ability to measure subtle thermal shifts that the human eye might miss. Other student researchers at NYITCOM, including Andrew Tisser, Perry Rosen, Arianna Falletta, Noah Hoonhout, Dongchan (Alex) Lee, and Chris Kyriakides, are exploring AI in dermatology and other specialties, pushing the boundaries of what this approach can do. As these applications mature, you can expect a broader clinical adoption where AI serves as a reliable, non-invasive assistant across many fields of medicine.
Frequently Asked Questions
How can AI turn subjective pain into objective data?
AI in medicine analyzes physiological signals like heart rate variability, skin conductance, and facial expressions to quantify pain levels. By training on large datasets of patient responses, it learns patterns that correlate with reported pain. This gives you a more consistent, data-driven metric rather than relying solely on a patient’s verbal description.
Is the AI-based epidural assessment method better than the standard pinprick/ice cube test?
AI-based methods can offer more objective and continuous monitoring compared to manual tests. While standard tests depend on patient response and clinician skill, AI evaluates subtle physiological changes in real time. This can reduce variability and catch early signs of complications, but clinical validation is still ongoing.
What are the risks or downsides of using AI in pain management?
Key risks include data privacy concerns and potential bias in training data that could affect certain patient groups. AI systems also require high-quality input data and may produce false positives if not properly calibrated. You should view AI as a support tool rather than a replacement for clinical judgment.
