Using Artificial Intelligence to Improve Pain Assessment and Pain Management: A Scoping Review

Approximate 50 million U.S. adults experience chronic pain. It is a widely held view that pain has been linked to sleep disturbance, mental problems, and reduced quality of life. Uncontrolled pain has led to increased healthcare utilization, hospitalization, emergency visits, and financial burden. Recognizing, assessing, understanding, and treating pain can improve outcomes of patients and healthcare use. A comprehensive synthesis of the current use of AI-based interventions in pain management and pain assessment and their outcomes will guide the development of future clinical trials. This review aims to investigate the state of the science of AI-based interventions designed to improve pain management and pain assessment for adult patients. The electronic databases Web of Science, CINAHL, PsycINFO, Cochrane CENTRAL, Scopus, IEEE Xplore, and ACM Digital Library were searched. The search identified 2131 studies, and 18 studies met the inclusion criteria. The Critical Appraisals Skills Programme was used to assess the quality. This review provides evidence that machine learning, deep learning, data mining, and natural language processing were used to improve efficient pain recognition and pain assessment (44%), analyze self-reporting pain data (6%), predict pain (6%), and help physicians and patients to more effectively manage with chronic pain (44%). Findings from this review suggest that using AI-based interventions to improve pain recognition, pain prediction, and pain self-management is effective; however, most studies are pilot study which raises concerns about the generalizability of findings. Future research should focus on examining AI-based approaches on a larger cohort and over a longer period of time.

Different brain systems support the aversive and appetitive sides of human pain-avoidance learning

Both unexpected pain and unexpected pain absence can drive avoidance learning, but whether they do so via shared or separate neural and neurochemical systems is largely unknown. To address this issue, we combined an instrumental pain-avoidance learning task with computational modeling, functional magnetic resonance imaging (fMRI) and pharmacological manipulations of the dopaminergic (100 mg levodopa) and opioidergic (50 mg naltrexone) systems (N=83). Computational modeling provided evidence that untreated participants learned more from received than avoided pain. Our dopamine and opioid manipulations negated this learning asymmetry by selectively increasing learning rates for avoided pain. Furthermore, our fMRI analyses revealed that pain prediction errors were encoded in subcortical and limbic brain regions, whereas no-pain prediction errors were encoded in frontal and parietal cortical regions. However, we found no effects of our pharmacological manipulations on the neural encoding of prediction errors. Together, our results suggest that human pain-avoidance learning is supported by separate threat- and safety-learning systems, and that dopamine and endogenous opioids specifically regulate learning from successfully avoided pain.

A deep learning method for predicting knee osteoarthritis radiographic progression from MRI

Background The identification of patients with knee osteoarthritis (OA) likely to progress rapidly in terms of structure is critical to facilitate the development of disease-modifying drugs. Methods Using 9280 knee magnetic resonance (MR) images (3268 patients) from the Osteoarthritis Initiative (OAI) database , we implemented a deep learning method to predict, from MR images and clinical variables including body mass index (BMI), further cartilage degradation measured by joint space narrowing at 12 months. Results Using COR IW TSE images, our classification model achieved a ROC AUC score of 65%. On a similar task, trained radiologists obtained a ROC AUC score of 58.7% highlighting the difficulty of the classification task. Additional analyses conducted in parallel to predict pain grade evaluated by the WOMAC pain index achieved a ROC AUC score of 72%. Attention maps provided evidence for distinct specific areas as being relevant in those two predictive models, including the medial joint space for JSN progression and the intra-articular space for pain prediction. Conclusions This feasibility study demonstrates the interest of deep learning applied to OA, with a potential to support even trained radiologists in the challenging task of identifying patients with a high-risk of disease progression.

Adaptive coding of pain prediction error in the anterior insula

Background: To reduce the computational demands of the task of determining values, the brain is thought to engage in adaptive coding, where the sensitivity of some neurons to value is modulated by contextual information. There is good behavioural evidence that pain is coded adaptively, but controversy regarding the underlying neural mechanism. Additionally, there is evidence that reward prediction errors are coded adaptively, but no parallel evidence regarding pain prediction errors. Methods: We tested the hypothesis that pain prediction errors are coded adaptively by scanning 19 healthy adults with fMRI while they performed a cued pain task. Our analysis followed an axiomatic approach. Results: We found that the left anterior insula was the only region which was sensitive both to predicted pain magnitudes and the unexpectedness of pain delivery, but not to the magnitude of delivered pain. Conclusions: This pattern suggests that the left anterior insula is part of a neural mechanism that serves the adaptive prediction error of pain.

Post-operative pain following cardiac implantable electronic device implantation: insights from the BRUISE CONTROL trials

Aims Post-operative pain following cardiac implantable electronic device (CIED) insertion is associated with patient dissatisfaction, emotional distress, and emergency department visits. We sought to identify factors associated with post-operative pain and develop a prediction score for post-operative pain. Methods and results All patients from the BRUISE CONTROL-1 and 2 trials were included in this analysis. A validated Visual Analogue Scale (VAS) was used to assess the severity of pain related to CIED implant procedures. Patients were asked to grade the most severe post-operative pain, average post-operative pain, and pain on the day of the first post-operative clinic. Multivariable regression analyses were performed to identify predictors of significant post-operative pain and to develop a pain-prediction score. A total of 1308 patients were included. Multivariable regression analysis found that the presence of post-operative clinically significant haematoma {CSH; P value < 0.001; odds ratio (OR) 3.82 [95% confidence interval (CI): 2.37–6.16]}, de novo CIED implantation [P value < 0.001; OR 1.90 (95% CI: 1.47–2.46)], female sex [P value < 0.001; OR 1.61 (95% CI: 1.22–2.12)], younger age [<65 years; P value < 0.001; OR 1.54 (95% CI: 1.14–2.10)], and lower body mass index [<20 kg/m2; P value < 0.05; OR 2.05 (95% CI: 0.98–4.28)] demonstrated strong and independent associations with increased post-operative pain. An 11-point post-operative pain prediction score was developed using the data. Conclusion Our study has identified multiple predictors of post-operative pain after CIED insertion. We have developed a prediction score for post-operative pain that can be used to identify individuals at risk of experiencing significant post-operative pain.