New technology: Handheld ultrasound-assisted localization of epidural space

Artificial intelligence for ultrasound scanning in regional anaesthesia: a scoping review of the evidence from multiple disciplines

“… 29 , 30 , 31 , 32 , 33 , 34 , 35 , 36 , 37 Around this point, there has been an increase in published data where the dominant technique was deep learning for both PNB 9 , 11 , 12 , 38 , 39 , 40 , 41 , 42 , 43 , 44 , 45 , 46 , 47 , 48 , 49 , 50 , 51 , 52 , 53 , 54 , 55 , 56 , 57 , 58 , 59 , 60 , 61 , 62 , 63 , 64 , 65 , 66 , 67 , 68 , 69 , 70 , 71 , 72 , 73 and CNB, 7 , 74 , 75 , 76 , 77 though there are some exceptions. 78 , 79 , 80 A number of sources do not explicitly state the techniques used; these are typically clinical case reports or (external) validation studies describing accuracy or efficacy 81 , 82 , 83 , 84 , 85 , 86 , 87 ,…”

“…For some sources, it is not clear how the ground truth for training data was obtained. 9 , 11 , 12 , 28 , 29 , 39 , 40 , 41 , 42 , 43 , 45 , 47 , 50 , 55 , 57 , 59 , 60 , 68 , 70 , 72 , 73 , 74 , 75 , 76 , 78 , 81 , 82 , 83 , 84 , 85 , 86 , 87 , 88 , 89 , 90 , 91 , 93 …”

“…The most common output for models processing CNB ultrasound images was classification/identification of vertebral level, the intervertebral space, or both. 30 , 31 , 32 , 33 , 34 , 75 , 76 , 77 , 81 , 82 , 83 , 84 , 85 , 86 , 87 , 88 , 89 , 90 , 91 , 93 Distance from the skin to the posterior complex/epidural space was the second most common output of models applied to CNB ultrasound. 7 , 80 , 82 , 83 , 84 , 85 , 86 , 87 , 88 , 90 , 91 , 92 , 93 Other functions included identifying the optimal needle insertion point, 29 , 35 , 81 , 88 , 89 , 91 structure identification (via segmentation or bounding boxes), 7 , 35 , 37 or 3D reconstruction of the spine.…”

“…Assessment of accuracy for CNB models was more aligned than the approaches for assessing PNB model accuracy, which involved a greater element of human judgement/interpretation. The most common method was correlating the distance from skin to posterior complex/epidural space as predicted by the model to that observed on manual interpretation of ultrasound, 32 , 34 , 84 , 88 , 93 depth of loss of resistance during needle insertion, 80 , 82 , 85 , 86 or both. 37 , 83 , 92 All reported good correlation between human depth assessment and AI depth assessment on the ultrasound image, and loss of resistance during needle insertion to be deeper than that calculated on ultrasound (by human or AI).…”

“…Nineteen sources used ultrasound scans of the lumbar spine 7 , 29 , 30 , 31 , 32 , 33 , 34 , 35 , 37 , 75 , 76 , 77 , 80 , 83 , 84 , 86 , 88 , 90 , 92 , 93 and two utilised the thoracolumbar spine, 74 , 82 whereas one did not specifically state which region was seen (though use of the lumbar region can be deduced as it is a case report of a patient undergoing total knee arthroplasty). 85 …”