Ketamine and facial emotion recognition in treatment-resistant depression: a computational account

Background: Facial emotion recognition deficits are well defined in depression but understudied in treatment-resistant depression (TRD). Over the past decades, subanesthetic doses of ketamine have emerged as an effective treatment for TRD, but its effect on affective processing is still little known. Approaches from computational psychiatry, such as sequential evidence sampling models of decision making, can provide useful insight into the underlying cognitive processes of facial emotion recognition in TRD patients before and after ketamine antidepressant treatment. Aim: Provide a computational account of facial emotion recognition in TRD and the effects of ketamine treatment using hierarchical drift diffusion modeling (HDDM). Method: In this observational case-control study, 24 TRD patients and 35 healthy controls completed a FER task. TRD patients were assessed before, 2-4 hours after the initial racemic ketamine infusion, and at one week of treatment. Healthy controls were assessed at similar intervals, without intervention, to assess baseline differences and potential learning effects related to task repetition. Accuracy and reaction times were analyzed and fit with an HDDM to formalize correct facial emotion recognition as a noisy evidence-accumulation process. Parameters of the best-fitting model were compared across groups and timepoints. Results: Before treatment, TRD patients were slower but as accurate as controls. Following ketamine, the patients were significantly faster than at baseline and compared with repeated testing in healthy controls. These effects were underpinned by longer nondecision times and wider decision boundaries in patients at baseline, both of which decreased after starting ketamine. There was no behavioral or computational evidence of a negative bias in patients. Conclusions: Results suggest that processes related to psychomotor slowing and response caution, rather than affective processing, underlie deficits observed during facial emotion recognition in TRD and can be restored after starting a ketamine treatment.