UCSF

Pain is known to have sensory, cognitive, and affective aspects. However, the mechanism behind individual perception and anticipation of pain remains a question in the field. In this investigation, 26 subjects who suffered from varying levels of severity of chronic low back pain (cLBP) were recruited into a neuroimaging study which included structural magnetic resonance imaging (MRI) and evoked pain paradigm functional MRI (fMRI) experiments. During evoked pain paradigm fMRI experiments, participants received randomized cues on the intensity of the upcoming pain stimulations (either high, low, or uncertain). Their structural images were input into segmentation tools to measure the volume of 31 regions of interest (ROIs). fMRI images were collected, preprocessed, and then processed to build individual activation maps. Daily experience with cLBP was also collected through self-reporting PEG (Pain, Enjoyment, General activity) scores. We assessed to what extent brain structure volumes were associated with self-reported PEG scores. Further, we built logistic regression models with LASSO penalization for each subject separately to test three things: 1) if neural patterns of each cLBP patient were separable when perceiving high pain and low pain stimulus; 2) if high pain and low pain anticipation brain activation patterns of cLBP patients were distinguishable during known anticipation cues; and 3) if brain activation during known anticipation cue states could be used to decode each cLBP patient’s anticipation bias during uncertain cue states. All analyses focused on structural and neuronal activation measures form a priori selected brain regions including subfields of insula, nucleus accumbens, substantia nigra, anterior cingulate cortex, amygdala, caudate nucleus, putamen, pallidum, subgenual frontal cortex, and thalamus. The linear regression models showed that the volumes of left insula middle short gyrus, right insula anterior inferior cortex and bilateral anterior cingulate cortex were negatively associated with PEG scores, which reflect their daily experience with cLBP. The LASSO model built for individuals separating high pain and low pain perception had an average area under the curve (AUC) 0.773 ± 0.206 and the accuracy of prediction was 0.65 ± 0.179. The LASSO model built for each individual separating high pain anticipation and low pain anticipation has an average AUC of 0.861 ± 0.218 and accuracy of prediction of 0.75 ± 0.21. Furthermore, the linear regression models assessing the association of 1) regional activation during pain perception, 2) regional activation during pain anticipation, 3) individual anticipatory bias decoded for uncertain cue states with PEG scores were not statistically significant in this cLBP cohort. Thus, we were not able to explain individually perceived severity of cLBP by their neuronal activation and anticipatory bias, but their brain morphometry changes in this study.