Study On Clinical Drug Decision Algorithm Based On Heterogeneous Information Network

With the rapid growth of biomedical literatures and clinical datas,the researches of big data to improving medical and health care are in full swing.How to extract effective medical knowledge for doctors to make clinical decision,from literatures and clinical data,is a very hot topic in recent years.Most of the current clinical decision support systems are based on Bayesian theory,Association rules,N eural network and other machine learning or data mining technologies,and the heterogeneous information network analysing as a data mining technology also can help doctors making clinical decision by constructing and analysing a heterogeneous information network from medical datas.This paper achieved two clinical decision algorithms by analysing heterogeneous information network based on literatures and clinical data,the main work of this study include:(1)An algorithm named Med Rank can be used to recommend influential medications from MEDLIN E by analyzing information network.The algorithm extracted the articles,authors,journals,drugs and other entities for a given disease and formed a heterogeneous star information network.Then calculating every entities’ possibility through authority ranking function which was just like Page Rank algorithm.Finally,Med Rank algorithm can recommend effective treatments for doctors based on the assumption that “a good treatment is likely to be found in a good medical article published in a good journal,written by good author(s)”.But there was no definition of “good” for the articles,journals and authors,an improved algorithm named HIC-MedRank was proposed by introducing H-index of authors,impact factor of journals and citation counts of articles as criterion for defining good articles,journals and authors,and recommended antihypertensive agents for the patients suffered from Hypertension with Chronic K idney Disease by considering published time,support institutions,publishing type and some other factors of articles.(2)Heterogeneous information network analysis technology was further applied in clinical data,the data was extracted from the third Xiangya Hospital of Central South University.This paper took recommending anti-hypertensive methods for hypertension and complicated with other diseases as an example.After obtaining datas,this paper removed invalid datas and features,then selected pivotal features to constituting a patient-centered heterogeneous information network,and finally ranked those anti-hypertensive methods by analysing network.In addition,this article can also get the rank lists of doctors and departments,recommend ing superior doctors and departments for patients.The experiments were conducted on Medline datasets and clinic datasets,the results show that the recommendation drugs of HIC-MedRank algorithm are more precise than the drugs of Med Rank,and are more recognized by attending physicians.The consistency rate is up to 80% by comparing with the JNC guidelines.And the results also show that the recommendation of C linic Rank is better than the recommended results of FP-growth algorithm and is more consistent with the drug use of doctors.The doctors and departments’ satisfication with the recommended drugs are more than 70%.