Mei-Hua Hall, PhD

Dr. Hall’s Psychosis Neurobiology Laboratory was founded in 2013. Her lab focuses on two interconnected areas of research. One area is geared toward deepening our understanding of the neurobiological underpinnings of psychotic disorders and mapping out the trajectories of patient recovery. The second area leverages the power of machine learning (ML) and natural language processing (NLP) to delve into the vast information contained in patients’ electronic health records (EHR).

The objective is to develop cutting-edge NLP-based tools for risk prediction and to provide valuable clinical and computational resources to the psychiatric community.

Dr. Hall receives funding from the National Institute of Mental Health and other sources for her research.

Dr. Hall’s lab is currently conducting several projects. One project uses an integrative approach, combining neurophysiological, neurochemical, genetics, clinical, and behavioral measures to understand the neurobiological mechanisms underlying schizophrenia and bipolar disorder. This project also aims to clarify how genetic and environmental risk factors influence psychopathology, neurocognitive traits, and brain function.

Robust genome-wide genetic studies have identified many risk genes and variations within those genes associated with schizophrenia and bipolar disorder, though their effects on the brain are largely unknown. The lab collects a variety of neurophysiological, clinical, and cognitive functioning data on patients with these disorders. The goal is to link genetic risk factors associated with psychosis disorders to brain neurophysiological characteristics.

Through collaboration with the Broad Institute Stanley Center, the lab investigates the relationships between genetic risks in patients with schizophrenia and bipolar disorder, as well as specific biomarkers, including sleep spindles, auditory evoked related potential responses, and neurocognitive measures, to elucidate the underlying mechanisms.

Another project of the lab uses multimodal approaches to stratify patients who experience their first episode psychosis into homogeneous subgroups based on patients’ unique neurobiological profiles and to relate these profiles to later functional recovery outcomes. This study longitudinally follows up with patients at the early stage of their illnesses every six months for two years. An expanded biomarker panel (cognition, MRI, EEG, clinical symptoms) and functional outcome measures are collected in each patient. The goals are to identify and better characterize individual longitudinal patterns of functioning recovery and to explore the potential risk and protective factors associated with functioning outcomes.

In addition, the lab uses a computerized cognitive training program for patients with schizophrenia and bipolar disorder with the goal of improving certain cognitive functions (such as memory and attention) and psychosocial outcomes. In recent years, training programs based on learning-induced neuroplasticity have shown promising improvement effects on cognitive function.

This suggests that cognitive rehabilitation training induces neurobiological remodeling in brain circuitry relevant to a range of cognitive functions. This project examines brain changes induced by cognitive training to reveal the underlying mechanisms mediating these changes.

A new line of research focuses on interoception (the processing, interpretation, and regulation of bodily signals by the brain) in psychotic disorders. This project utilizes a multimodal approach (EEG, EKG, psychophysiology, and eye tracking) to characterize interoception and its association with clinical symptoms. The goal is to gain more understanding of the underlying mechanisms that may aid the development of new interventions.

Concurrently, the lab focuses on using natural language processing (NLP) and machine learning techniques to create sharable psychiatric EHR data resources and tools for evaluating and predicting psychosis patient readmission risk as well as risk of harm. This project includes experts in psychiatry and computational linguistics.

Manual chart review is impractical and not cost-effective, so the lab has developed a “computational version of chart reviews” pipeline that systematically extracts important clinical features from patient EHRs and converts these data to features for use in training machine learning models relating to trauma and readmission risk factors. In addition, the lab creates a psychiatric-specific temporal relation annotation scheme for building accurate readmission prediction models.

The ultimate goal of these endeavors is to develop precise, actionable tools for clinical practice, enhancing our understanding of the interaction between neurobiological mechanisms and environmental risks in psychotic and mood illnesses.