The aim of this work was to develop a procedure for analysing and processing large amount of medical data in order to help physicians to make better decisions during the diagnosis process and to forecast high risk patients for developing a specific disease.
The suggested solution is a procedure that combined different techniques that were applied to pre-process, analyse and visualize patient medical record data in order to foresee the patient high risk of developing a specific disease. This is achieved through the analysis of inter, intra-family relationships among patients and their similarities in terms of diagnosed diseases. The above procedure led to the development of a system prototype that will help the physician to make better decisions during the diagnosis process.
The methods followed during the research process in order to reach to the proposed solution consists of primary research, such as an interview with the physician and secondary research such as related work, journal articles, papers, internet sources.
The collaborative filtering technique was used to explore, not only the patients’ intra-family relation, but also, the inter-family relations, in order to define a similarity score among patients. The patients are grouped based on their similarity score.
The Latent Semantic Index clustering technique was used to analyse the group of similar patients in order to foresee the risk of developing a specific disease.
Cluster visualization was used to visualize the results. In this way, the physician can further investigate and understand the results i.e. why a patient has a high risk of developing a specific disease.