Summary: A precision neuro-engineering and computational intelligence breakthrough successfully engineered an artificial intelligence platform to decode and objectify human pain. The research overcomes the historic diagnostic boundaries of subjective self-reporting.
By utilizing a dual-model, self-correcting AI algorithm to audit electroencephalogram (EEG) signals triggered by thermal stimuli, the platform maps localized brainwave activity to provide an uncorrupted, real-time biological ruler for physical suffering.
Key Facts
- The Subjective Diagnostic Deficit: Historically, clinical medicine has been entirely dependent on the Visual Analogue Scale (VAS), a highly subjective framework where patients manually rate their own pain levels. This reactive approach produces wildly inconsistent evaluations for identical physical stimuli and completely fails vulnerable populations who cannot communicate, such as intensive care patients, individuals with impaired consciousness, young children, and the elderly.
- The Dual-AI Self-Correcting Algorithm: Led by Principal Researcher An Jinung (DGIST) and Professor Jeon Seong-chan (GIST), the team bypassed standard machine learning limits. Instead of training an AI on biased, subjective human scores, they deployed an innovative architecture where two distinct AI models cross-compare their prediction results. The system selectively trains itself only on highly reliable, matching data points, successfully filtering out individual biases in pain expression.
- Universal Calibration Across Unseen Environments: Rigorously validated using EEG data across a cohort of 41 participants, the self-correcting model vastly outperformed traditional neural networks. Crucially, the platform demonstrated the capacity to maintain stable, highly accurate pain-intensity predictions when exposed to entirely new stimulus environments that the AI had never encountered during its training phase.
- Isolating the F7 and F8 Neuro-Biomarkers: The study yielded a massive neurophysiological discovery by mapping the exact cranial coordinates where pain registers. Investigators isolated delta wave activity located specifically within the left and right anterior temporal lobes, mapped electrophysiologically to the F7 and F8 frontal nodes,proving this specific brainwave signature tracks directly with the physical intensity of pain.
- Transitioning Toward Real-Time BCI Platforms: As first author Dr. Jeong Ui-jin notes, the ultimate trajectory of this technology is to expand past static testing into a Brain-Computer Interface (BCI)-based real-time monitoring network. This would allow hospitals to track patient suffering continuously without requiring verbal communication.
- Immediate Clinical Deployment Vector: Backed by the National Research Foundation of Korea, the framework is designed to function as a universal pain AI platform. Its immediate real-world applications include real-time pain monitoring before and after complex surgeries, objective tracking in intensive care units (ICUs), and long-term diagnostic tracing for chronic pain disorders.
Source: DGIST
Daegu Gyeongbuk Institute of Science and Technology (DGIST, President Lee Kunwoo) announced that a research team led by Principal Researcher An Jinung at the DGIST Industrial AX Innovation Institute (he also serves as an adjunct professor in the Interdisciplinary Engineering), in collaboration with Professor Jeon Seong-chan’s team at Gwangju Institute of Science and Technology (GIST), developed technology that uses artificial intelligence (AI) to analyze electroencephalogram (EEG) signals triggered by thermal stimuli and objectively classify pain intensity.
As pain perception varies from person to person, previous methods relied heavily on the Visual Analogue Scale (VAS), a subjective scale expressed by patients. This resulted in inconsistent evaluations, even for the same stimulus, and posed significant limitations in accurately assessing pain for patients who have difficulty communicating, such as those with impaired consciousness, children or the elderly.
Principal Researcher An’s team developed a technology that uses AI to analyze EEGs generated during various thermal stimuli to classify pain intensity. Notably, shifting from conventional methods that simply learned from patients’ subjective pain scores, the team implemented an innovative algorithm in which two AI models compare their prediction results and selectively learn only from highly reliable data. Using this, the bias in pain expression, which varies from person to person, was effectively reduced.
Consequent to testing with EEG data from 41 participants, the model they developed demonstrated significant improvements in performance compared with conventional models and maintained stable predictions in new stimulus environments, in which the model had not been trained yet. Furthermore, they revealed that delta wave activity in the left and right anterior temporal lobes (F7 and F8) is closely associated with pain intensity, thereby establishing a neurophysiological basis for developing 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,” stated An. “We intend to develop this into a universal pain AI platform that can be utilized in actual clinical settings by integrating various bio-signals.”
First author Jeong Ui-jin, a postdoctoral researcher, stated, “We hope this technology will be widely used for pain monitoring before and after surgeries, chronic pain tracking, and objective pain assessment in intensive care units,” adding, “Moving forward, we will devote ourselves to research so that it can be expanded into a brain-computer interface (BCI)-based real-time monitoring system.”
This study was conducted with support from the National Research Foundation of Korea’s Mid-Career Researcher Support Program and the Future Promising Convergence Technology Pioneer (Challenge Type) Program.
The findings were published in the May issue of IEEE Transactions on Neural Systems and Rehabilitation Engineering, a prestigious international journal in rehabilitation engineering.
Key Questions Answered:
A: Because traditional medicine has lacked an objective, physical ruler for suffering. For decades, clinics have relied entirely on the Visual Analogue Scale (VAS), which requires patients to verbally describe or point to their pain level. If a patient has impaired consciousness, is too young to speak, or is an elderly individual struggling to communicate, doctors are left guessing because there was no way to read pain straight from the human nervous system.
A: By forcing two separate AI models to cross-examine each other’s homework. Traditional AI models fail because they try to learn from highly flawed, subjective self-reports. The DGIST and GIST teams solved this by programming an algorithm where two distinct AI systems compare their predictions in real time, choosing to learn exclusively from data that both models flag as highly reliable, effectively neutralizing individual emotional bias.
A: They are the exact neurophysiological coordinates where physical pain prints its digital signature. The researchers discovered that delta wave activity within the left and right anterior temporal lobes, specifically tracked at the F7 and F8 electrode sites, scales perfectly with pain intensity. Isolating this exact brainwave highway provides a definitive, objective digital biomarker that allows scientists to build real-time brain-computer interfaces to monitor suffering.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- Journal paper reviewed in full.
- Additional context added by our staff.
About this neurotech and AI research news
Author: Wankyu Lim
Source: DGIST
Contact: Wankyu Lim – DGIST
Image: The image is credited to Neuroscience News
Original Research: Closed access.
“EEG-based Pain Classification via Sample Selection to Mitigate Subjective Label Bias” by Euijin Jung; Sung Chan Jun; Jinung An. IEEE Transactions on Neural Systems and Rehabilitation Engineering
DOI:10.1109/TNSRE.2026.3692232
Abstract
EEG-based Pain Classification via Sample Selection to Mitigate Subjective Label Bias
Quantifying pain intensity is essential for enabling personalized pain management. Recently, electroencephalography (EEG)-based approaches have been investigated to estimate pain levels, particularly for patients who are unable to communicate their pain due to cognitive or neurological impairments.
However, most existing methods are trained using self-reported pain labels, which are inherently subjective. This subjectivity often leads to biased models that limit the reliability of predictions.
To address this issue, we propose a novel method that incorporates reliable sample selection for EEG-based pain level classification during training. The proposed approach quantifies sample informativeness and estimates label reliability. Each sample is then assigned a priority level, and those identified as either unreliable or uninformative are excluded to enhance model robustness.
We evaluate the method using EEG data from 41 participants exposed to warm, cool, and thermal grill illusion (TGI) stimuli, with pain labels collected via the Numerical Rating Scale (NRS). A 5-fold cross-validation procedure is employed to ensure robustness in both quantitative and qualitative evaluations. The proposed model achieves statistically significant improvements over baseline models in multi-class classification with 3, 6, and 10 classes.
Furthermore, we demonstrate that our method generalizes well to previously unseen types of thermal stimulation, underscoring its potential for objective pain assessment in non-communicative patients. Additional analyses reveal pain-related EEG features, indicating that delta-band activity at the left and right frontotemporal electrodes (F7 and F8) is strongly associated with perceived pain intensity.
