Clinical Text Analysis and Mining using Artificial Intelligence

Clinical text, such as progress reports, safety reports, includes large amounts of detailed patient and disease information. In this talk, I will focus on learning and case identification problems on clinical text, and present how we can develop artificial intelligence-based approaches that extract knowledge and support the clinical decisions. First, I will introduce the pathological feature assessment for melanoma (skin cancer) patients using natural language processing techniques. Then I will present an attentive deep neural network model that automatically identifies the allergic events from hospital safety reports. I will show the generalizability and interpretability of the proposed model and demonstrate how does the model extract the clinical knowledge which is complementary with human knowledge