Assisted query formulation for multimodal medical case-based retrieval

Medical information retrieval systems support health care experts in diagnostic and treatment decisions through the management of large amounts of clinical data. However, the ever growing data produced in medical environments and the proficiency of non-professional users pose several challenges to a retrieval system. In this paper, we propose a medical retrieval system, supporting semantic multimodal queries for medical case-based search. The system explores many of the commodities available in commercial search engines and provides the user with key tools to support medical information discovery: multimodal queries and semantic suggestions of medical terms. It is build upon state-of-the-art information retrieval and data fusion techniques. We also propose a new data-fusion technique, which we called Inverted Squared Rank (ISR), to better deal with the combination of ranked lists from various heterogeneous systems: similar to Reciprocal Rank Fusion approaches like RR [10] and RRF [3]. The proposed rank fusion method outperforms RR and RRF in most measures and is particularly better on the top 10 results. The system is at http://medical.novasearch.org/