چكيده به لاتين
In recent years, significant advancements have been made in the development of high-power ultrasound transducers, leading to the emergence of high-intensity focused ultrasound (HIFU) therapy. This non-invasive and non-ionizing treatment approach shows promise in effectively ablating masses through coagulation necrosis. However, a crucial challenge lies in accurately guiding the HIFU transducer to the desired focal area for creating thermal lesions. Ultrasound monitoring has been proposed as an alternative to MRI, which presents certain limitations in monitoring this treatment. Utilizing ultrasound for HIFU guidance offers notable advantages, including cost-effectiveness, portability, and the ability to operate concurrently with HIFU. This thesis focuses on the development of a novel method for guiding HIFU treatment using ultrasound, specifically utilizing entropy-based techniques and extracting the twinkling artifact. These methods possess significant advantages, including exceptionally low computational complexity and the feasibility of implementation on standard medical center devices. The exploration and analysis of these characteristics form a crucial aspect of this research. This research utilized a series of B-mode ultrasound images of extracorporeal porcine muscle subjected to HIFU sonication, obtained from previous studies. The objective was to investigate the changes in selected entropies before, during, and after HIFU treatment, employing a suitable pattern for analysis. Among the entropies examined, the threshold, normalized Shannon, and Norm entropies yielded the most favorable outcomes, exhibiting increases of 71%, 68%, and 31% in the post-treatment frames compared to the pre-treatment frames, respectively. The entropies were then calculated within 1 〖mm〗^2 window and stored in a new matrix. Visualizing these entropy matrices enabled effective monitoring of lesion formation, enhanced by techniques such as overlapping windows and computing the difference between the pre-HIFU matrix and those during and after HIFU. Employing YOLO segmentation, the images provided reports on the total lesion area every 84 to 178 ms, contingent on the specific entropy used for image construction. The predicted area demonstrated an error range of 6% to 13% across different entropies. During the subsequent phase of this project, a comprehensive collection of Doppler images was obtained by extracting Doppler signals. The Doppler signals were derived through the sequential sampling of radio frequency (RF) signals. Within this series of images, a remarkable twinkling artifact was notably observed within the lesion area. Importantly, the presence of this artifact in the Doppler images exhibits significant potential for delivering precise guidance in monitoring the progression of the lesion.
