Dr. Saab and his team are exploring a new, more objective way to diagnose pain that is rapid, less expensive and provides more accurate results.
In a new study led by Carl Saab, PhD, and published in NeuroImage, researchers have developed a novel methodology that combines electroencephalography (commonly called EEG) and artificial intelligence to diagnose pain. This marks a giant step forward in objectively diagnosing pain, which has long been a difficult and inaccurate process that relies almost entirely on self-report.
Chronic pain is a nearly universal experience, afflicting millions of people and costing billions in treatment and loss of productivity. Misdiagnosing pain can make treatment difficult and has contributed to the overprescription and abuse of opioids, a major health crisis in the United States.
According to Dr. Saab, who recently joined the Department of Biomedical Engineering from Brown University, where he led this project, “Our approach offers a significant improvement over the current gold standard for assessing pain, which is what we call the ‘smiley face report’ that you are typically asked to fill out at primary care or the emergency clinic—a frown for pain, a smiley face for no pain.”
Dr. Saab and his team developed and trained an artificial intelligence platform to objectively detect and classify pain using EEG scans (which non-invasively map the brain’s electrical activity) from three groups: healthy patients; patients diagnosed with a chronic pain condition called radiculopathy; and patients with chronic back pain who were scheduled to undergo surgery to implant a pain-relieving assistive device.
The researchers trained the computer algorithm to compare and correctly distinguish patterns among the EEGs of the different subject groups, all matched for age and gender. They found that the machine learning algorithm performed significantly better than chance in correctly distinguishing between healthy patients without pain and those with chronic pain. It also accurately distinguished patients scheduled for surgery from all the other subjects.
While earlier studies of EEG-based pain detection have focused on long EEG durations, this new algorithm investigated shorter events at the sub-second scale, thus providing a novel scientific understanding of the mechanisms of pain in the human brain.
“EEGs give us so much data and helpful information on sensory states that artificial intelligence tools are necessary to decipher at all,” noted Dr. Saab. “Now, with the use of our algorithm, pain diagnosis may become much more precise and, importantly, can be an automated process.”
Not only will the algorithm help automate and streamline the process of diagnosing pain, but Dr. Saab says it may also help automate medical decision making by providing caregivers actionable information.
Currently, pain medicine specialists identify patients who may be good candidates to receive the assistive device surgery (implantation of a spinal cord stimulator) based on subjective self-reports from patients and recommendations from a panel of healthcare experts after analyzing patients’ long-term health histories. This algorithm may help simplify and streamline a patient’s decision to undergo surgery by providing quickly generated and objective empirical evidence.
“Considering that assessing new or unusual pain is a common onramp for diagnosing the majority of health conditions, this is an exciting new methodology that could potentially transform healthcare. It may also pave the way to apply a similar approach for assessing mental health conditions like depression and anxiety.”
Using artificial intelligence methods, Drs. Gaj and Nakamura have created a fully automated method to count lesions in the brain scans of patients with multiple sclerosis.