Ultrasound imaging in neurosurgery: Delineation of tumours for resection control

Abstract

Complete removal, as opposed to only partial removal, of all macroscopic tumor tissue is associated with longer survival for most of the brain tumours. Ultrasound is an imaging technique that can contribute to increase the degree of tumour being removed by the surgeon, and thereby indirectly contribute to increased survival. Ultrasound may also contribute to safer surgery and less risk for neurological damage. However, this requires that the ultrasound images are able to depict the tumour in an adequate manner, throughout the whole operation.

The studies included in the thesis investigate whether or not ultrasound is able to adequately depict the transition between tumour and normal brain. In two studies we have used 3D ultrasound and a navigated biopsy forceps to collect biopsies in the peripheral tumor in order to compare ultrasound image findings and histology.

The results of the biopsy studies show that there is a high correspondence between histology and ultrasound findings, for biopsies sampled prior to start of resection. However, the quality of the ultrasound images is affected by the progress of the surgery. Ultrasound image volumes acquired before start of resection show a realistic delineation of the solid tumour. The images from 3D ultrasound volumes acquired during and at the end of surgery are somewhat more difficult to interpret for the surgeon. The surgery itself introduces more noise and interference in the ultrasound images, making the images less specific in separating normal tissue from tumour. The reasons for the increased noise are discussed, and it is suggested how the noise can be identified and reduced.

In three other studies ultrasound data has been processed off-line to estimate strain in the brain tissue. The aim was to investigate whether or not processing of ultrasound data could detect different elasticity in tumor tissue and normal brain, and if so, check if this could differentiate between tumor and normal brain. The work led to the first article, of our knowledge, demonstrating that it is possible to image brain tumors using ultrasonic elastograms, i.e. images depicting strain in the tissue. The pulsation of the arteries during a cardiac cycle impose movements of the surrounding brain tissue, and this was found to cause sufficient deformation to allow generation of strain images. Furthermore, the results showed that the degree of strain is significantly less in tumor tissue than in normal brain tissue. Analysis of image contrast, showed that the strain images have significantly higher contrast between tumor and normal brain than conventional ultrasound images. Ultrasound imaging of strain has thereby the inherent potential to differentiate better between pathological and normal brain tissue than conventional ultrasound. The results suggest that ultrasound elastography methods should be further developed for intraoperative use. This could complement the future use of ultrasound imaging in neurosurgery.