Application of Support Vector Machine in Medical Diagnosis
Anto Satriyo Nugroho
School of Life System Science & Technology, Chukyo University Japan
Email : nugroho@life.chukyo-u.ac.jp   URL : http://asnugroho.net

Support Vector Machine (SVM) received increasing attention, and it has
been successfully applied in various fields. SVM in principle is a linear
machine which is trained to find the optimal discriminating hyperplane in
the feature space by maximizing the margin between the classes. The training
phase of SVM is conducted based on structural risk minimization, different
from neural networks that follow the empirical risk minimization. This
is the reason why SVM are often reported to achieve better generalization
than conventional pattern recognition method such as neural networks.

Computer Aided Diagnosis, on the other hand, requires a method with high
accuracy, while the information is often noisy, high dimensional setting,
small sample and incomplete. To deal with problem with such characteristics,
we develop a computer aided diagnosis system using Support Vector Machine
that is combined with individual merit base feature selection. The
experimental results showed that SVM achieved good results, and robust
to the existence of irrelevant features.