This work presents development and testing of image processing algorithms for the automatic detection of landmarks within ultrasound images.

The aim was to automate ultrasound analysis, for use during the process of epidural needle insertion. For epidural insertion, ultrasound is increasingly used to guide the needle into the epidural space. Ultrasound can improve the safety of epidural and was recommended by the 2008 NICE guidelines (National Institute for Health and Care Excellence). Without using ultrasound, there is no way for the anaesthetist to observe the location of the needle within the ligaments requiring the use of their personal judgment which may lead to injury. If the needle stops short of the epidural space, the anaesthetic is ineffective. If the needle proceeds too deep, it can cause injuries ranging from headache, to permanent nerve damage or death.

Ultrasound of the spine is particularly difficult, because the complex bony structures surrounding the spine limit the ultrasound beam acoustic windows [1]. Additionally, the important structures for epidural that need to be observed are located deeper than other conventional procedures such as peripheral nerve block. This is why a low frequency, curved probe (2–5 MHz) is used, which penetrates deeper but decreases in resolution.

The benefits of automating ultrasound are to enable real-time ultrasound analysis on the live video, mitigate human error, and ensure repeatability by avoiding variation in perception by different users.

Previous ultrasound image processing for epidural research used speckle image enhancement with canny and gradient based methods for bone detection [2]. A clinical trial with 39 patients had success detecting the ligamentum flavum (LF) from ultrasound by algorithms in 87% of patients.

Echogenic needles and catheters are now becoming available which are enhanced for extra ultrasound visibility. The Epimed UltraKath ULTRA-KATH™ [3] has a patented design to maximize visibility under ultrasound [4]. The Echogenic Tuohy Needle also includes imprints on the needle tip that reflects ultrasound, allowing for better visualization. Curved needles can also be detected in 2D ultrasound images [5].

This content is only available via PDF.