The Application of Support Vector Machines to Medical decision Support: A Case Study

Support Vector Machines have been successfully applied to a number of classification and regression tasks. This paper outlines their successful application to a medical decision support task aimed at improving the diagnosis of tuberculosis infection. The standard method for diagnosis of tuberculosis is the visual identification of tubercle bacilli in stained sputum smears. The proposed technique based on image capture, image processing and classification of image objects is expected to be more accurate than human visual screening. In this application a Support Vector Machine is successfully used for the classification of image objects. This application is one example of a much broader class of medical imaging tasks to which Support Vector Machines could be applied. INTRODUCTION Since their introduction by Vapnik and coworkers (Vapnik 1995), Support Vector Machines (SVMs) have been successfully applied to a number of real world problems such as handwritten character and digit recogniti...