Acquisition and Frequency Spectroscopic Evaluation of Broadband Clinical Ultrasound Raw Data for Liver Cirrhosis and Focal Pathologies Using Neural Networks for Tissue and Pathology Differentiation

What this Trial Is About

Researchers are evaluating how well neural networks trained on ultrasonic raw radiofrequency data can assess liver diseases in patients undergoing clinical ultrasound exams. The study aims to compare the performance of these neural networks against elastography and those trained on b-mode ultrasound images, as well as to see if they can distinguish focal liver lesions from healthy tissue. This research is conducted with patients who have a clinical indication for elastography or suspected liver lesions and involves collecting detailed ultrasound data for analysis. Participants scheduled for elastography will have both b-mode images and radiofrequency data collected during their ultrasound scans. For those with suspected focal liver lesions, ultrasound data is collected along with a definitive diagnosis obtained through standard clinical procedures such as contrast-enhanced ultrasound, biopsy, MRI, or CT, depending on what is normally done at the participating center. The study includes two groups: one focused on elastography data collection and another on focal lesion evaluation, all without randomization or masking. During the study, participants undergo clinical ultrasound examinations to capture both b-mode images and corresponding radiofrequency data. Additional tests or procedures may be performed to confirm diagnoses for focal liver lesions. Researchers will analyze the performance of the trained neural networks after the study concludes, approximately one year later. The total participation time varies, with assessments aligned to routine clinical care. Safety monitoring includes exclusion of recent liver interventions to avoid confounding results.

Assessment of Liver Diseases Using a Deep-Learning Approach Based on Ultrasound RF-Data