M.Sc Thesis

M.Sc StudentAlchanati Shira
SubjectCreating Panorama Pictures from Ultrasound Images for
Needle Localization Betweem Vertebrae
DepartmentDepartment of Biomedical Engineering
Full Thesis textFull thesis text - English Version


The visualization of the vertebrae by ultrasound imaging is very limited due to their complex structure and the considerable differences of the acoustic properties between soft tissues and hard tissues, such as bones. The ultrasound waves can advance through soft tissues while hard tissues mostly reflect them, resulting in a bright lamina and an acoustic shadow beneath it. This limits ultrasound imaging for back procedures even though it could serve as a real-time monitoring tool. Today, although high accuracy and precision are required for performing and monitoring procedures such as epidural anesthetic induction, anesthesiologists still insert the needle to the required location by manual inspection of the vertebrae. This procedure is carried out “blindly”, even though the spinal cord could be damaged in such a process. Back surgeries also lack monitoring in real-time and the surgeons mostly use pre-operative Computed Tomography images to determine where the screws should be inserted.

Our goal was to enable better visualization of the spine, in a way that could improve accuracy while performing such procedures, while using conventional ultrasound imaging. In addition, we aim to improve visualization of the needle in order to assist in needle localization. Ultrasound is the modality of choice, since it can be used for real-time monitoring and its radiation is not harmful

For this purpose, we acquired data from backs of subjects and from a vertebrae phantom (including a needle, in a water bath or a jelly-based medium). We used a phased array transducer for imaging due to its ability to steer the beam and allow sonication focused at different angles. The data acquired was both the B-mode data and raw IQ data. A signal and image processing methods were developed that included applying image stitching and generalized sampling methods to create a panorama that compounds the frames of the vertebrae from several locations. The assumption was that imaging the same area from different points of view and intertwining the images will complete the vertebrae’s appearance. The raw data was pre-processed, before being beamformed into an ultrasound image, using a measure similar to the phase coherence factor.

Our results show that the panorama enhances the visibility of the vertebrae by creating a fuller visualization of them, as opposed to images that display only a small part of one vertebra - only the part perpendicular to the transducer. The panorama also denoises the surroundings of the vertebrae. Finally, the panorama improves the visibility of the needle by combining the low resolution frames of the needle into a higher resolution panorama. These results were obtained by processing of both B-mode data and raw data. However, processing B-mode data requires fewer computations, as the ultrasound system already provides processed data.

We conclude that panorama pictures provide clearer information about the shape of the vertebrae and their surroundings and contribute to needle localization.