In Daegu, South Korea, researchers have solved a problem that has vexed doctors for decades: how to measure pain objectively instead of relying on what patients tell them they're feeling. A team led by Principal Researcher An Jinung at the DGIST Industrial AX Innovation Institute, working with Professor Jeon Seong-chan's team at Gwangju Institute of Science and Technology, has developed artificial intelligence that reads the brain's own pain signals—recorded as electroencephalogram waves—to classify pain intensity without the distortion of human subjectivity.
The breakthrough matters because pain has always been stubbornly personal. When a doctor asks a patient to rate their pain on a scale of one to ten, that number tells us more about the individual's pain tolerance, cultural background, and communication style than about the actual intensity of what they're experiencing. This subjectivity creates real problems in hospitals. Dosing decisions go wrong. Patients with impaired consciousness, young children, and older adults who struggle to communicate their pain often suffer needlessly because doctors can't verify their experience. Even for the same stimulus applied to the same person twice, the subjective Visual Analog Scale—the gold standard for decades—produces inconsistent results.
An's team addressed this by developing an algorithm that works differently from conventional AI. Rather than simply learning from patients' self-reported pain scores, the system uses two AI models that compare their predictions against each other and selectively learn only from the most reliable data points. This mutual skepticism between models filters out the noise of individual bias, zeroing in on what's actually happening in the brain.
Testing with EEG data from 41 participants, the researchers found that the model significantly outperformed conventional approaches and, remarkably, remained stable and accurate when faced with new thermal stimuli it had never encountered before. The team also made a neurophysiological discovery: delta wave activity in specific brain regions—the F7 and F8 sites in the left and right anterior temporal lobes—correlates closely with pain intensity. This finding establishes a biological anchor for pain measurement and opens the door to developing what researchers call "brain-based digital biomarkers."
"This study directly addresses the bias in subjective self-reported labels, which was the chronic limitation of EEG-based pain analysis," An said. The vision extends beyond the laboratory. An plans to integrate multiple bio-signals into a universal pain AI platform suitable for clinical use, while first author Jeong Ui-jin, a postdoctoral researcher, envisions the technology monitoring pain before and after surgeries, tracking chronic pain patients, and providing objective assessment in intensive care units where communication is often impossible.
The implications ripple outward quietly. Imagine a patient in a coma whose family wonders if they're suffering. Imagine a pediatric ward where doctors can finally verify whether a child's pain management is working. Imagine chronic pain patients whose experiences are finally documented by the organ that experiences them—the brain itself. The research, published in IEEE Transactions on Neural Systems and Rehabilitation Engineering, represents a shift from asking people how much it hurts to asking their brains directly. The next frontier, Jeong notes, is real-time monitoring through a brain-computer interface, bringing this objective pain assessment from the research setting into the continuous care of actual patients.