Keywords: Depression | Neuroscience | PTSD
Presentation Type: Symposium
Background: Mental illnesses such as depression and PTSD comprise differentiable symptom clusters, which may better relate to underlying behavioral and neural disruptions than aggregate diagnoses. In the present study, we used a well-understood computational model of learning to characterize how specific symptom clusters of PTSD and depression relate to differences in processing positive and negative information and the effect of CBT on these differences.
Methods: 169 participants (69 with MDD, 39 with PTSD, 29 trauma-exposed controls, and 32 community controls) completed a gain and loss learning task during fMRI scanning and completed structured clinical interviews and self-report measures of symptom severity. Participants’ behavior on the task was fit to a reinforcement learning model and the resulting correlates of learning were regressed against neural activity. Behavioral and neural differences were assessed among levels of symptoms (reexperiencing and hyperarousal in PTSD and anhedonia and negative affect in depression). A subset of participants with depression also completed 12 weeks of standard CBT and were re-assessed post-treatment.
Results: Within PTSD, hyperarousal and not reexperiencing symptoms were related to increased attention-modulated updating during loss learning (behavioral associability weight: hyperarousal t=2.29, p<.05, reexperiencing t=0.79, p=.4; neural associability value: p<.05). Conversely, participants with depression showed decreased modulation (t=7.25, p<.001); in the depressed participants, anhedonic symptoms were related to reduced updating of reward value (reduced learning rate and decreased neural coupling of learning signals), while during loss learning, negative affect was related to more negative behavioral valuation and reduced ventromedial prefrontal cortex signaling of value. After CBT, disrupted gain and loss learning both showed changes correlated with improvements in symptom severity.
Discussion: Computational models such as reinforcement learning uniquely connect neural, behavioral, and symptom-level disruptions in depression and PTSD and provide new insights into the presentation and treatment of these disorders.
Virginia-Tech Carilion Research Institute
Friday, November 17
8:30 AM – 10:00 AM
Friday, November 17
12:00 PM – 1:30 PM
The asset you are trying to access is locked. Please enter your access key to unlock.