Exploring the Role of Ultrasound RF Envelope Features in Bone Health Monitoring

Hossam H. Sultan, Enrico Grisan, Sevan Harput

Abstract: Ultrasound imaging has emerged as a pivotal tool in bone assessment, offering a non-invasive, safe, and practical approach that does not involve ionizing radiation. However, accurate bone properties classification from ultrasound data remains a challenging task due to inherent complexities in the bone internal structure. Traditional methods often overlook slight variations in bone structure and composition, delaying the precision of diagnosis and treatment planning. In response to this, our study investigates the potential of envelope features in selecting specific frequency bandwidths to improve the monitoring of bone health. By extracting and identifying features, we aim to develop models suitable for machine learning/artificial intelligence applications, as well as for point-of-care diagnostics and remote care. We demonstrate that multi-frequency ultrasound imaging is particularly helpful in identifying the most effective frequency for bone assessment. Additionally, we find that standard deviation emerges as the most discriminative feature, exhibiting 2.5 times greater discriminatory power compared to other secondary features.