Ziv Ben-Zion, PhD

I'm interested in the field of affective neuroscience, the study of neural mechanisms of emotions and emotional disorders. My goal is to conduct research that will allow us to better understand the complex interactions between human behavior, emotions, and our brain structure and function. Specifically, I wish to focus on stress- and anxiety-related disorders, having a personal motivation to improve treatments for individuals suffering from these debilitating disorders. 

Neurobiological processes that take place during the year that follows a traumatic event critically determine who will develop post-traumatic stress disorder (PTSD) and who will not. Among survivors of single traumatic incidents, the chronic disorder frequently follows a failure to recover from early PTSD symptoms. Longitudinal studies further describe diverging early symptom trajectories of non-remission, rapid remission, and delayed remission from early post-traumatic normative responses. To better understand the underlying neurobiology, these observable symptom trajectories must be linked with cognitive deficits and pertinent brain alterations (either present initially or developing simultaneously). To date, however, large-scale, prospective longitudinal studies of PTSD symptom trajectories that involve repeated cognitive and neuroimaging assessment are critically missing.

My doctoral research was designed as an observational prospective study of consecutive trauma survivors admitted to a general hospital’s emergency department (ED) following traumatic incidents. The overarching goal was to uncover the neurocognitive moderators underlying PTSD symptom trajectories. To achieve this goal, we repeatedly and simultaneously evaluated trauma survivors’ clinical symptoms, cognitive functioning, brain structure and function, at 1-, 6-, and 14-months following trauma exposure. Results can be organized into three objectives.

First, we utilized advanced computational methodology to characterize and classify individuals within 1-month following trauma, based on the collected multi-parametric measurements. We successfully identified subgroups of individuals (significantly related to PTSD clinical diagnosis, but not identical to it), with a unique set of potential mechanism-related cognitive and neural biomarkers differentiating between them, in line with previously documented PTSD literature (Ben-Zion et al., 2020). Second, we investigated which objective multi-parametric indices, collected shortly after trauma exposure, could be of value in predicting individuals’ long-term clinical PTSD symptoms. We found both a cognitive construct that emerged as a significant predictor of PTSD development (i.e., cognitive flexibility; Ben-Zion et al., 2018), as well as neuroanatomical risk factors for PTSD severity (Ben-Zion et al., 2019), both of which could guide early management and objective long-term monitoring. Finally, we elucidated the neurobehavioral mechanisms underlying motivational processing in PTSD development. During a competitive decision-making paradigm, we found that increased behavioral risk-aversion and imbalanced neural responsivity to punishments vs. rewards, early after trauma, were predictive of PTSD symptoms 13-months later on (Ben-Zion et al., 2021). 

In summary, by linking observed symptoms with cognitive functioning and neural alternations, findings from this thesis work enhanced our understating of the nature of traumatic stress responses and its aftermath, informing both the pathogenesis of PTSD and the science of resilience and recovery from trauma. Lower cognitive flexibility in PTSD might be manifested as imbalanced neural responsivity to positive vs. negative valance stimuli, potentially involving key brain structures such as the hippocampus and the amygdala. Future studies using similar integrative, mechanism-oriented, exploratory approaches, may lead to improved early treatment and prevention of PTSD, thus improving the life of trauma survivors and increasing the cost-effectiveness of personalized interventions.

During my postdoctoral training at Yale, I’ve continued to work on this unique dataset with the additional expertise of my mentors and lab members. First, I conducted a conceptual replication of a recent promising study that identified distinct brain-based biotypes associated with different longitudinal patterns of post-traumatic symptoms (Stevens et al., 2021). Results did not replicate in our sample, suggesting that caution is warranted when attempting to define subtypes of psychiatric vulnerability using neural indices before treatment implications can be fully realized (Ben-Zion et al., In Press). As one of the first replications attempts in the fields of neuroimaging and psychiatry, I believe that this work will critically push the fields toward future replication studies to identify more stable and generalizable brain-based biotypes of psychopathology. Second, I’ve explored the longitudinal association between structural brain changes and PTSD symptom trajectories during the first critical year following trauma exposure. Utilizing a novel longitudinal segmentation pipeline assessing hippocampus and amygdala subregions volume, together with an advanced Bayesian multilevel modeling approach, we found that lower initial volumes of the hippocampus (specifically, the subiculum and CA1 subregions), and larger amygdala volumes, are associated with non-remitting PTSD, thus supporting the idea volumetric abnormalities serve as predisposing vulnerability factors for PTSD development (Ben-Zion et al., 2022). Moreover, no time-dependent volumetric changes were observed (from 1- to 14-months post-trauma) across all individuals or between those who developed PTSD and those who did not, thus not supporting the idea of progressive, stress-related atrophy of hippocampal or amygdala subregions. Third, collaborations with experts in computer and data sciences demonstrated the use of deep-learning models (Sheynin et al., 2021) and advanced statistical learning methods (Schultebraucks*, Ben-Zion* et al., 2022) to predict the risk to develop chronic PTSD from early cognitive and neural measures. Future work could further delineate the mechanisms that underlie such prediction models, and potentially improve single-patient characterization and preventive treatment.

Furthermore, I’ve contributed to several research projects led by fellow postdocs in my labs, while also learning from them and improving my scientific skills and knowledge. First, using robust Bayesian statistics and two independent samples of veterans, we showed that the amygdala’s response to pain is lower in PTSD individuals, and is associated with greater emotional numbing symptoms (Korem et al., 2022). Second, we demonstrated that machine-learning models, based on routinely collected nursing data in an acute care setting, can reliably predict patients at risk for delirium (Spiller et al., 2022). Finally, I’ve co-led a pre-registered scoping review aiming to map the availability of translated and evaluated screening questionnaires for PTSD. Overall, we screened 866 studies, of which 126 were included, with a total of 128 translations of 12 different PTSD questionnaires (Hoffman*, Ben-Zion* et al., 2022). Results showed very large heterogeneity in the translation and validation processes, making a quality assessment impractical. This highlights both the need for more transparency in translation processes and the need for more rigorous evaluation methods. Furthermore, translations into languages spoken in middle- or low-income countries were underrepresented, emphasizing that more investment is needed in those neglected countries and language groups. Finally, as the majority of existing translations (73%) were not accessible, we developed an online open repository for translated & validated PTSD screening questionnaires (tobiasrspiller.github.io/PTSD-Screener-Repo/).

Currently, I’m learning the field of computational decision-making neuroscience (Prof. Levy) and expanding my clinical knowledge to anxiety- and stress-related disorders (Prof. Harpaz-Rotem). My main research aims to characterize neural computations of value (how good or bad something is), uncertainty (how precise these value estimations can be), and prediction error (the discrepancy between expected and observed information) across different processes (learning, decision-making, memory) and domains (positive vs. negative valance). My hypothesis is that estimations of this basic set of computations in different processes and domains rely on both shared and distinct neural mechanisms; and that different psychopathology symptoms are associated with the type and magnitude of alterations in these computations. To test these computations within individuals, I’ve developed a naturalistic computer-game task of a new virtual world, in which participants learn the associations between different stimuli and outcomes; choose between different stimuli; and update their learning when circumstances change based on previous memory. In the short term, we aim to conduct a large online behavioral study (N=1000) and an fMRI study (N=100) testing our ecological task within individuals from the general population. Using computational modeling and advanced classification techniques, we will characterize individuals based on their behavioral and neural patterns, and explore this objective classification in relation to stress- and anxiety-related symptoms. In the long term, we hope to use this task to study various mental health symptoms (beyond stress and anxiety) based on objective neurobehavioral individual profiles of value, uncertainty, and prediction error in different processes and domains. This approach could inform mechanism-based objective diagnoses and individually tailored effective interventions.