My completed projects include identifying pathological fMRI brain activity in disorders such as Schizophrenia, measuring whether EEG correlates of cognitive ability may change with aging, and creating a "speech recognition" algorithm to identify whether and why an infant is crying. I believe that communicating our science to the public is the best method of moving academic research out of the ivory tower and into practice. More recently, my diabetes EHR screening work was featured in NPR and Medscape.
As Principal Investigator and Director of the Laboratory of Computational Neuropsychology, my lab integrates data across modalities to create biophysical measurements of neuropsychological disorders. More generally, we are focused on bioinformatics and translational science - harnessing data and statistics for biomedical applications. Toward this aim, I lead several projects:
- ChatterBaby: I am the founder of the ChatterBaby project which launched in 2013. This project won first prize in UCLA's 2016 Code for the Mission competition- MHealth Apps. The ChatterBaby algorithms use artificial intelligence to translate a baby's cry into a distinct need, predicting acute infant pain with over 90% accuracy. These algorithms are now being used to identify pathological infant vocal patterns in children who are and are not at risk for autism spectrum disorder.
- Early prediction of cognitive decline: In Alzheimer's disease and in vascular dementia, extreme variability in hemodynamics captured through fMRI may predict future decline. I was recently awarded a K25 grant from the NIA to support my work further, using Dynamic Causal Modeling to obtain measurements of neurovascular health.
- Public Sector Innovation: My laboratory works with Litmus at NYU to identify how acoustic features may impact and predict prison violence.
- Placebo effect: measuring the placebo effect in vivo in the brain using fMRI, to increase our ability to separate effective drugs from the ineffective. This technology will particularly be useful in assessing drug efficacy for rare diseases, when it may be unethical and impractical to assign patients to a placebo group. I have also created a new trial design to mitigate the placebo response.
- Diabetes risk: using electronic medical records and data mining to create new detection algorithms for diabetes screening. Diabetes is the fifth leading cause of preventable death in the United States, and frequently isn't diagnosed until there is at least one complication.
- Psychometrics: the PANSS is a symptom scale used to evaluate disease severity in Schizophrenia, a disease particularly debilitating to individuals and devastating for families. Accurately measuring Schizophrenia helps to identify effective treatments for this disease.