I2S2 Seminar: Clinical AI From Time Series to Free Text

Data-driven clinical decision making, e.g., via machine learning, holds considerable promise for evidence based medicine. This session will present three ongoing advances from BCBI’s AI Lab that tackle very different types of input data ranging from numerical vital signs used to predict pediatric sepsis at Hasbro Children’s Hospital, via automatic localization of pneumonia on chest X-rays at Memorial Sloan Kettering Cancer Center to free-text analysis of PubMed articles. In each of these talks, we will discuss the clinical relevance of the studied task, the studied input data modality, methodological considerations of the machine learning models, and real-world applicability of the derived methods.