Development of nature inspired algorithms for identification of Spine on ultrasound images in spina bifida cases

Spina bifida (SP) is a spine defect observed when a baby is still in the womb. The defect is caused by unfinished closure of the embryonic neural column. This thesis proposes a combination of novel nature inspired methods for locating spinal axis on sonograms, where deformation due to spina bifida is observed. The method involves a flocking dynamics based optimization approach for reducing the size of the search space (bones on the sonogram) and a meta-heuristic evolutionary approach, where the sonogram is divided into columns and bone blobs belonging to the spine are classified. Accordingly, a specific genetic structure and fitness function is utilized and conventional genetic operators are applied to search the actual solution. Results show that both approaches are promising and proposed combination of algorithms generally can distinguish the spinal bones from others even if severe morphological defects exist on the sonograms.