Identifying cavitation regions using spectral and intensity data; application to HIFU

The high power intensities in HIFU often result in bubble production, either through cavitation or boiling, which are believed to be a primary contributor to tissue necrosis. Bubbles are associated with the bright hyperechoic regions in ultrasound B-mode images. As the only changes observed on tissue are subtle during treatment, some HIFU therapy protocols rely on the observation of significant brightness changes as the indicator of tissue lesions. The occurrence of a distinct hyperechoic region around the focus is often associated with cavitation. In general, the hyperechoic regions show good correlation with ablated tissue (observed directly following subsequent removal of the tumour in an operation, or using MRI), but the sensitivity of this techniques is sub-optimal. Reliable detection of cavitation and a method to distinguish between different types of events is therefore, an important goal for better control of the treatment. This thesis presents a novel method to provide detection of cavitation activity as an aid to assisting treatment. The image intensity information is used to identify hyperechoic regions spatially and temporally. However, hyperechoic regions may appear for reasons other than cavitation – for example because of tissue interfaces. The spectral information is useful to distinguish from other events and thermal generation of bubbles. Thus the spectral estimation methods are becoming of increasing interest in early and robust detection of cavitation activity. There are three main contributions related to this thesis: identifying the boundaries and maintaining a history of cavitation events from their brightness and intensity statistics through using a probabilistic method, determining not just the presence of cavitation but also its local changes at a high spatial resolution through analysing spectrally the RF signals from the imaging transducer on a pixel by pixel basis, and finally combining the advantages of both methods to improve the overall reliability of automatic cavitation detection. In addition, the spectral information extracted here is capable potentially of distinguishing between cavitation and boiling. The method is assessed using a simulation of a synthesised cavitation, and the applied to detect cavitation following HIFU in ex-vivo calf liver.