Development of an ultrasound image processing approach to enable automated needle localisation for epidural anesthesia

In this paper, ultrasound imaging is utilised to detect the anatomical structure of lumbar spine, in order to automatically search for a proper epidural needle insertion site. Induced by random speckle noise, ultrasound images are generally obscure and have low resolution, which decrease the readability of the image content. In order to improve the interpretability of ultrasound images, local normalisation modified with difference of Gaussian (DoG) algorithm is used for pre-processing. The pre-processing procedure successfully filters the noise and extracts the main anatomical structures; meanwhile it removes the local means induced by non-uniform wave reflection rate. Then, template matching algorithm, augmented with a position correlator, is utilised to automatically identify the features of interest. Based on the result of feature matching, proper needle insertion site will be identified automatically. The approach has been tested on more than 200 ultrasound images with a 100% success rate. The propose...