Mining brain features from schizophrenia studies with Shift-And pattern matching

摘要

An efficient prospect to medical procedures such as diagnosis and therapy is by obtaining knowledge of medical experts from formal reports. Several studies have been carried out on finding differences in brain connectivity between schizophrenia patients and healthy controls with their results reported in natural language with tables and figures. In the area of biomedical research, natural language processing can be employed to retrieve relevant information from articles by scientific and medical experts, based on which a brain network characterizing schizophrenia could be built. Hence, this study presents suitable text mining model for retrieving information about brain region. Meta-analysis is employed to integrate knowledge from different, while relevant information is retrieved from scientific publications with Shift-And Pattern Matching. Evaluation on a set of 1,525 scientific literatures on schizophrenia shows the model has good recall of 73.7%.