Differential diagnosis of eczema and psoriasis using categorical data in image processing

A number of attempts have been made at using image processing and machine learning to automate the diagnosis process of dermatological diseases. Psoriasis and eczema are two conditions that are usually misdiagnosed, and were selected after an investigation. The reason for this high rate of misdiagnosis is due to the margin of error present in the method doctors use to perform a diagnosis. This method consists of using the visual aspects of the disease along with non-visual patient history questions. This method is not as effective when certain diseases do not have definitive answers to these history questions and the doctors have to base the diagnosis almost entirely on the visual aspects displayed on this skin. The proposed solution addresses this problem by combining both the visual as well as nonvisual feature vectors into a single classifier, a support vector machine, which once trained, enables diagnosing to a reliable level of accuracy at 84%.