Processing Compressible 4D-CT Medical Image by using Cellular Neural Network

For 3-dimension image we use definition �voxel� (volumetic cell) instead of pixel (picture cell) of 2D. A single voxel particle consists parameter x, y, z for spatial dimensions. Data of voxel can be used to describe large objects and applied in video game, geology, astronomy, satellite image� and especially medical image processing. For surgery image collected from CT scanning picture such as brain, lung, heart, chest, hip� the object periodically change their location and size (respiration, heart rhythm, muscle and bone joint movement�) so the time element (t) must be added to parameter group of voxel. So the processing of motion image has relation with 4D-CT scan technique. The 4D-CT image processing requires IR, the virtue of which is to determine compatible cells among two consecutive image in movement. There are many methodes based on various indentification criteria (on optical flow, calculation model, geometrical characteristics�). They are popularly used in processing and calculation of medical images. However there are still many constraints ( such as accuracy or calculation optimization). With regards to medical image in general and lung tumour diagnosis in particular there are often many differnces because of cell flexibility and disruption. In general, those methodes yield good results if the strength of light is constant (Horn/ Schunk approach) but this condition is unrealistic because of lung movement based on respiratory rhythm which leads to variations of light on object (lung tissue) and causes discrepancies. We call CT image of affected tissues as compressible image. In this report we study speed calculation of optical flow of compressible CT 4D image.4D image of large voxel volume and calcualtion complexity is a considerable challenge for existing PC based on sequence processing. In order to increase speed calculation we propose neuron cell parrallely processing network capable of calculating 4D optical flow in real time mode.