Assistive system based on nerve detection and needle navigation in ultrasound images for regional anesthesia

The development of Ultrasound-Guided Regional Anesthesia (UGRA) is of great help to practitioners of regional anesthesia as it enables real time visualization of the needle, the targeted nerve, and different anatomic structures. However, the clinician has to perform a complex hand coordination to keep the needle, the nerve and some key regions visible in the ultrasound image plane. Daily clinical practice therefore requires a high degree of training and practical skill to identify the nerve block and steer the needle to it There are two critical steps in UGRA: the recognition of anatomical structures and steering the needle to the target region. An intelligent system, with the ability to identify the regions of interest and to provide the needle insertion trajectory in ultrasound images, can significantly improve UGRA practice and generalize it to medical facilities that lack practitioners. It would also make the UGRA procedure safer (i.e., reducing the risk of nerve trauma). This work presents the first fully automatic system for the detection of regions of interest and generation of the needle trajectory for UGRA. Several problems were addressed, in two stages. The first one consisted in the automatic localization and segmentation of the nerve (target) and arteries (obstacles) in ultrasound images. A new method based on a machine learning algorithm with a multi-model classification process using a sliding window for localization, then an active contour was applied to delineate the localized regions. In the second stage, an algorithm for path planning was also developed to obtain the optimal trajectory for needle insertion based on the result of the first stage (target and obstacle detection). To check the effectiveness of the proposed system, firstly, experiments were performed over individual modules of the detection framework. Secondly, a comparison between the overall framework and the existing method was performed. Two data-sets were acquired in real conditions at different times to prove the robustness of our method. The first data-set contained eight patients and the second data-set, acquired one year later, contained five patients. Experimental results demonstrate the robustness of the proposed scheme and the feasibility of such an assistive system. (C) 2016 Elsevier Ltd. All rights reserved.