Deep Learning in the Health Record: Discovering Meaning in Clinical Text
Everyone knows that electronic healths records (EHRs) are messy. From missing data, to nonstandard text, to interoperability issues — all this and more make EHRs one of the most difficult types of data to use in data science applications. Despite these formidable challenges, the EHR remains the largest source of clinical data extant today. This talk will focus on several applications of deep learning in EHRs with special emphasis on NLP using clinical text, as well as discussing selected challenges in building end-to-end pipelines using EHR data.
Tasha specializes in a special domain of AI — biologically inspired neural networks. As part of her PhD work at Columbia University, she developed cutting edge technologies to elucidate the reasoning behind neural network based predictions.
At Droice, Tasha manages the AI team to develop state of the art technology that understands a doctor’s thought process through natural language understanding. Prior to joining the machine learning lab at Columbia, she was focused on core research involving mathematical modeling enabling brain computer interfaces. Tasha is a physicist who graduated from Brown University, where she researched the function of neural circuits using optogenetics and mathematical modeling. Tasha believes that the next breakthrough in AI will be the elimination of the black-box behavior of today’s AI.