Abstract
In medicine, categorizing the characteristics, behavior, and severity of a disease is vital for estimating prognosis and choosing treatment. Despite the volume of rich descriptive data available in medical images and texts, many of the features used for phenotyping are subjective or ambiguous. Machine learning approaches offer opportunities to both standardize descriptions of disease and expand the detail of phenotyping assessments. Using inflammatory bowel disease as an exemplar, we will discuss clinical applications of natural language processing, image segmentation, and computer vision to improve phenotyping for individuals and populations.