The good, the bad, and the facts : multimodal representation of medical conversations for patient understanding

Abstract

Medical patients face significant challenges for managing their health information. In particular, cancer patients have a uniquely difficult experience where they must endure the physical and emotional effects of their illness while simultaneously navigating overwhelming amounts of medical information. In this thesis, I focus on the challenge of capturing, reviewing and extracting information from medical appointments for patients enduring serious health conditions such as cancer. First, I propose a novel multimodal-interface to help patients review and understand information they received from conversations with their doctors. This interface captures medical conversations as text and audio, with important positive and negative information highlighted. I conducted 25 user studies where I enacted fictional conversations between a doctor and a patient to evaluate whether this method of representing information would help patients review and understand their appointments. Results from the user studies show that the web interface serves as a useful tool for reviewing the content of the conversations, however its effect on patient understanding cannot yet be determined. Second, I propose a machine learning algorithm to automatically classify the positive and negative information in medical conversations based on analysis of the text and prosody in speech. The model with the highest performance on my dataset achieved an accuracy of 90.6% and Fl-score of 0.888. While I focus on challenges within the medical field, findings from this thesis may be relevant to emotional conversations in any setting such as sportscasting, political debates and more.