Automatic Design of a Novel Image Filter Based on the GA-EM Algorithm for Vein Shapes

Medical doctors and clinical technologists operate specific, complicated diagnostic systems to assess venous diseases. Instead of using such expensive equipment, low-cost infrared cameras could capture vein images noninvasively and simply. On the other hand, the obtained image may have low contrast and a low signal-to-noise (S/N) ratio and this should be sufficiently improved by filtering processes. Therefore, an efficient image filtering method to estimate venous changes will enable the early detection of disease. In this study, a novel filtering method based on the genetic algorithm (GA) with the expectation maximization (EM) algorithm was newly proposed for the visualization of venous shapes, its effectiveness was evaluated by images acquired from a near-infrared (780 nm) charge coupled device (CCD) camera. The novel filter was automatically designed by the GA to efficiently improve the worse S/N ratio of venous images, even with an unknown correct image answer. In future studies, the proposed filtering method could be employed to easily detect peripheral swelling from vein images.