Quantifying ultrasound medical image segmentation for peripheral nerve blocks: a comparison of expert evaluations

Editor—The advent of ultrasound-guided regional anaesthesia (UGRA) has revolutionised the practice of regional anaesthesia including peripheral nerve blocks. Compared with traditional surface landmark techniques that anaesthesiologists use to gauge the location of nerves, UGRA offers continuous visualisation of the nerves and surrounding sonoanatomy. 1 Ito H. Shibata Y. Fujiwara Y. et al. Ultrasound-guided femoral nerve block. J Anesth. 2008; 57: 575-579 Google Scholar This live visualisation allows for a more precise spread of the injectate around a nerve, faster sensory onset, and an overall increase in blocking success rate. 2 McCartney C.J. Lin L. Shastri U. Evidence basis for the use of ultrasound for upper-extremity blocks. Reg Anesth Pain Med. 2010; 35 (–S5): S10 Crossref PubMed Google Scholar As a result of the rapid adoption of this imaging technology, identification (segmentation) of anatomical structures in ultrasound images has become a significant research area, particularly in relation to the development of machine learning models for tasks related to diagnostics or anaesthetic procedures. 3 Pesapane F. Codari M. Sardanelli F. Artificial intelligence in medical imaging: threat or opportunity? Radiologists again at the forefront of innovation in medicine. Eur Radiol Exp. 2018; 2: 35 Crossref PubMed Scopus (390) Google Scholar In this study, we aimed to determine a clinically relevant threshold for image segmentation by assessing intra-labeller and inter-labeller variability in anaesthetists when segmenting ultrasound images. Leading in the development, standardised evaluation, and adoption of artificial intelligence in clinical practice: regional anaesthesia as an exampleBritish Journal of AnaesthesiaPreviewA recent study by Suissa and colleagues explored the clinical relevance of a medical image segmentation metric (Dice metric) commonly used in the field of artificial intelligence (AI). They showed that pixel-wise agreement for physician identification of structures on ultrasound images is variable, and a relatively low Dice metric (0.34) correlated to a substantial agreement on subjective clinical assessment. We highlight the need to bring structure and clinical perspective to the evaluation of medical AI, which clinicians are best placed to direct. Full-Text PDF