Flexible needles detection in ultrasound images using a multi-layer perceptron network

In percutaneous surgical procedures, the precise positioning of the needle tip is of paramount relevance for good performance. Nevertheless, locate the tip of a inserted needle in a patient still represent a challenge. An imaging technique widely used for this task is the ultrasound imaging, however the low signal/noise ratio often make virtually impossible the use of simple image processing techniques. In this work, we propose an alternative way to detect needles in ultrasound images using Multi-Layer Perceptron neural networks. The results show good needle detection in most of cases, giving an average error of 3 pixels in 529 pixels per image.