Developing an Interface For Skin Disease Prediction

Current diagnosis of children with skin diseases is very difficult. Symptoms can appear with multiple diseases and symptoms can vary from case to case. Because of these issues, doctors tend to have tunnel vision when they are doing diagnosis. This tunnel vision presents itself when the doctor spots one symptom, and immediately starts to work in the direction of this disease, the doctor looks for symptoms that are common with the disease. This tunnel vision makes him overlook certain things, and he can end up with a wrong diagnosis. To support a doctor when he’s examining a child with a skin disease, a Bayesian network can help the doctor by calculating probabilities of the diseases when it’s given evidence for certain symptoms and/or other knowledge about the patient and it’s history. Because working with a real Bayesian network is to difficult and time consuming for the doctor, an interface is needed. The interface can guide the doctor through the questions, which avoids the tunnel vision problem. An approach to interface development and an initial implementation that deals with this problem was developed.