Three-dimensional ultrasound bone imaging for spinal and epidural placement
Abstract Number: OP1-5
Abstract Type: Original Research
Introduction: Ultrasound guidance of neuraxial anesthesia has gained recent popularity among anesthesiologists. However, conventional 2D ultrasound imaging has proven to be limited in the obese and morbidly obese population  due to numerous factors including: user skill required to interpret ultrasound images, limited imaging depths, and artifacts (i.e. false appearance of tissue structures in the image) generated from spinal bone reflections .
Methods: To circumvent the limitations of conventional ultrasound, a handheld ultrasound device was developed with 3D bone imaging capabilities. Fully automated segmentation of spinal bone surfaces was achieved using a new active shape model based signal processing algorithm. In this technique, a statistical model of spine surface shapes was developed using a “training set” comprising 6,120 spine surface profiles produced from a computed aided design (CAD) model. Additionally, specialized ultrasound transducers were implemented to reduce bone-derived artifact signal compared with conventional ultrasound. Feasibility of the proposed device was demonstrated using an excised deer spine and in vitro phantom model. Bone imaging artifact signal was quantified in terms of image contrast for both conventional ultrasound and the proposed imaging system. Error from 3D bone surface renderings with and without the active shape model approach was quantified using a CAD spine model embedded in a tissue-mimicking gelatin phantom (example images in Fig below).
Results: The proposed device demonstrated both 2D (Fig A) and 3D (Fig B) bone imaging capabilities. Custom transducers yielded reduced artifact bone imaging with average contrast improvement of approximately 6 decibels (dB). The active shape model based segmentation approach (Fig A; red line denotes the segmented spine surface) yielded 3D reconstructions (Fig B) that exhibited root mean squared (RMS) error of 1.8 mm compared to 2.2 mm without the active shape model.
Conclusions: The proposed handheld ultrasound device enabled intuitive imaging of bone anatomy with inherently safe ultrasound. Using the proposed techniques, ultrasound-based image artifacts were reduced compared to standard ultrasound. In addition, highly accurate 3D spinal bone renderings were demonstrated (Fig B) at imaging depths up to 10 cm.
 Chin et al, Anesth 2011
 Grau et al, Can J Anesth 2003