AI Pain Recognition Tool Detects Pain Before, During and After Surgery

Researchers presenting at the ANESTHESIOLOGY 2023 annual meeting demonstrated that an automated pain recognition system leveraging artificial intelligence (AI) may be capable of detecting pain in patients before, during, and after surgery.

Current methods to assess pain are often subjective. The Critical-Care Pain Observation Tool (CPOT) requires care teams to rate a patient’s pain based on body movement, muscle tension, and facial expression, while the Visual Analog Scale (VAS) asks patients to rate their own pain.

These methods can be biased and limited in scope.

“Traditional pain assessment tools can be influenced by racial and cultural biases, potentially resulting in poor pain management and worse health outcomes,” explained Timothy Heintz, BS, lead author of the study and a fourth-year medical student at the University of California San Diego, in the press release. “Further, there is a gap in perioperative care due to the absence of continuous observable methods for pain detection. Our proof-of-concept AI model could help improve patient care through real-time, unbiased pain detection.”

Improved methods for pain detection have the potential to prevent adverse outcomes such as anxiety and depression, along with decreasing hospital length of stay, the researchers indicated.

The AI model was given 143,293 facial images from 159 non-pain episodes and 115 pain episodes in 69 patients who underwent various elective surgical procedures. The tool, which utilizes a combination of deep learning (DL) and computer vision, was then trained to interpret each raw facial image based on whether or not it represented pain.

After being trained on enough images, the model was found to focus on facial expressions and muscles in particular areas of the face, such as the lips, nose, and eyebrows.

Following training, the tool’s output aligned with VAS results 66 percent of the time and with CPOT results 88 percent of the time.

“The VAS is less accurate compared to CPOT because VAS is a subjective measurement that can be more heavily influenced by emotions and behaviors than CPOT might be,” Heintz noted. “However, our models were able to predict VAS to some extent, indicating there are very subtle cues that the AI system can identify that humans cannot.”

If the technology is validated in additional studies, the research team highlighted that it may improve pain assessment and management in clinical settings.

In that scenario, cameras mounted on the ceiling and walls of a patient’s surgical recovery room could feed up to 15 images per second to the AI tool, which could then assess pain for conscious and unconscious patients.

Doing so could also help nurses and other members of the care team focus less on intermittently assessing a patient’s pain and more on other aspects of care.

Moving forward, privacy concerns must be addressed, but the researchers plan to incorporate additional monitoring features—including sound, movement, and brain and muscle activity—into the model.

This research represents just one potential application of AI in surgical care.

In July, Desirée Chappell, CRNA, vice president of clinical quality at Northstar Anesthesia in Irving, Texas, and Jonathan Tan, MD, vice chair of Analytics and Clinical Effectiveness at the Children’s Hospital Los Angeles (CHLA), who serves as an assistant professor of Clinical Anesthesiology at CHLA and the Keck School of Medicine at the University of Southern California sat down with HealthITAnalytics to discuss the potential role of these advanced technologies in anesthesiology.

They noted that while AI does show significant promise in the field, navigating the hype around these tools is key to using them effectively while prioritizing patient safety.