Decorrelation-based adherent microbubble identification as a faster alternative to singular spectrum-based targeted molecular (SiSTM) imaging of large blood vessels

Ultrasound-based targeted molecular imaging holds promise for early detection and diagnosis of cardiovascular disease and stroke. Current methods used to separate signal from adherent microbubbles are based on frequency domain filtering. Singular spectrum-based targeted molecular (SiSTM) imaging is a recently proposed technique that employ statistical properties, as quantified by the normalized singular spectrum area (NSSA), to more effectively separate signal components in large blood vessels. However, the computational cost to calculate NSSA is high, and thus real-time implementation is challenging. In this paper, flow phantom experiments demonstrated the NSSA-decorrelation patterns caused by different mechanisms: electronic noise, in-beam and out-of-beam movement of scatterers. Results showed that flow rates had little effect on the NSSA-decorrelation pattern caused by out-of-beam decorrelation. Based on the relationship between NSSA and decorrelation (approximately quadratic, adjusted-R² > 0.86), decorrelation-based adherent microbubble detection was demonstrated to be a faster (2-fold) alternative while maintaining similar performance compared to the NSSA-based method (less than 3% difference).