Proposing a new biomedical passage retrieval framework using clinical laboratory test results as query

Abstract

Clinical Decision Support (CDS) search is performed to retrieve key medical literature that can assist the practice of medical experts by offering appropriate medical information relevant to the medical case in hand. In this paper, we present a novel CDS search framework designed for passage retrieval from biomedical textbooks in order to support clinical decision making using laboratory test results. The framework utilizes two unique characteristics of the textual reports derived from the test results, which are syntax variation and negation information.

For the first part of this dissertation, we proposed a novel framework that consists of three components: domain ontology, index repository, and query processing engine. We first created a domain ontology to resolve syntax variation by applying the ontology to detect medical concepts from the test results with language translation. We then pre-processed and performed indexing of biomedical textbooks recommended by clinicians for passage retrieval. We finally built the query processing engine tailored for CDS, including translation, concept detection, query expansion, pseudo relevance feedback at the local and global levels, and ranking with differential weighting of negation information. To evaluate the effectiveness of the proposed framework, we followed the standard information retrieval evaluation procedures. An evaluation dataset was created including 28,581 textual reports for 30 laboratory test results and 56,228 passages from widely used biomedical textbooks, recommended by clinicians. A total of 20 assessors manually determined whether or not the top 500 retrieved passages were relevant to each test result. Overall, our proposed passage retrieval framework, GPRF-NEG, outperforms the baseline by 36.2, 100.5, 69.7 percent for MRR, R-precision, and Precision at 5, respectively. It also outperforms the best state-of-the-art approach by 4.3, 37.9, and 14.0 percent for MRR, R-precision, and Precision at 5, respectively. Our study results indicate that the proposed CDS search framework specifically designed for passage retrieval of biomedical literature a practically viable choice for clinicians as it supports their decision making processes by providing relevant passages extracted from the sources that they prefer to refer to, with improved performance. The domain ontology and dataset created in this paper are available at: http://kirc.kaist.ac.kr/dataset/.

For the second part of this dissertation, we further developed the passage retrieval framework by incorporating proximity information. To do so, we use knowledge structure, which graphically visualizes key concepts and their relationships in a specific domain where nodes are concepts with their associative relationships. Our framework investigated two new approaches to build knowledge structure compared to the conventional knowledge structure building approach. First, we used word embedding techniques to build initial knowledge structure. Second, instead of exploiting beg of word approach, we investigated node analysis to determine which of the analysis techniques perform effectively to detect the term importance. To do so, we compared two models with/without edge pruning to capture more latent relationship between terms. Our experiment shown that the embedded based knowledge structure outperformed the previous version of knowledge structure building approach and other proximity-aware state-of-the-art models. The strength of this dissertation lies in a wide variety of clinical decision support search tasks. Especially, one of the most frequency medical speciality, clinical laboratory test, was proposed as query to test practice of the proposed framework as we believe that this work will potentially enhance the quality of CDS search in practice.