Neural Network Blood Vessel Detection in Laparoscopy

In laparoscopic surgery, because of lack of tactile sensation and 3D visual feedback, it is difficult to identify a blood vessel position intraoperatively. We combined the color and geometry in two steps of Neural Network image processing. Application of LVQ (Learning vector quantization) neural networks has enabled the system to overcome the complexity of distribution of samples. The performance of this method is evaluated for detection of the artery and vein in twenty images of the laparoscopic surgery on a pig. Although this algorithm cannot detect blood vessel which are covered with fat or any other tissue it is an effective method for vessel detection, while suppressing capillary and other tissues.