MinneBOS - Boston's Field Guide to Data Science & Emerging Tech

In this talk, I will discuss about our proposed advanced deep learning models for classifying free text radiology reports based on the presence of pulmonary emboli (PE). The models are trained on a subset of Stanford training set (2512 reports) and evaluated on reports collected from four major healthcare centers. Our experiments suggest feasibility of broader usage of neural networks in automated classification of multi-institutional imaging text reports for various applications including evaluation of imaging utilization, imaging yield, and clinical decision support tools.