Performance analysis of ontology based biomedical document clustering algorithms to improve real time telemedicine systems

Abstract

Recent advancements in the medical field have also led to the massive increase of medical documents in literature as well as clinical practice. Advanced telemedicine and remote medical processing systems are utilizing such medical and clinical documents in analyzing, identifying, predicting and diagnosing the individual health concern and regional health indexes. The massive volume of unstructured biomedical documents becomes an issue for the performance of these systems. The reliability of telemedicine systems depends on the medical experts and any error due to these unstructured documents might lead to severe degradation in its credibility. The general document clustering models are not much suited for medical documents; the specialized models, especially the ontology based models are needed. The objective of this research is to develop improved clustering algorithms for biomedical documents utilizing the ontology [an ontology is a way of showing the properties of a subject area and how they are related, by defining a set of concepts and categories that represent the subject] for enhancing the performance of the telemedicine systems. Different models have been analyzed for biomedical documents using ontology enabled clustering concepts. The performance of the suggested biomedical document clustering algorithms has been tested with the real-time telemedicine system simulated using MATLAB (version 2013a) on documents of different disease domains collected from PubMed repositories. In the first approach, the limitations of the semantic smoothing model are analyzed and in order to overcome them, the ontology based Term Frequency and Inverse Gravity Moment (TF-IGM) enriched semantic smoothing model is developed for improving the clustering efficiency of traditional k-means and hierarchical clustering algorithms. Initially, the stemming, stop word removal is performed and then the improved background elimination is applied to handle the density of general words. TF-IGM is applied for term weighting and the duplicate documents are filtered out using a modified n-grams approach. Finally the clustering achieved using k-means and hierarchical clustering algorithms shows improved performance. newline