Dynamic Filtering of Adherent and Non-adherent Microbubble Signals Using Singular Value Thresholding and Normalized...

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Signal Processing

Background:

• Ultrasound molecular imaging requires distinguishing static tissue, adherent, and non-adherent microbubbles.
• Accurate microbubble signal identification is crucial for effective molecular imaging.

Purpose of the Study:

• To combine Singular Value Thresholding (SVT) and Normalized Singular Spectrum Area (NSSA) for microbubble signal isolation and classification.
• To evaluate the performance of the combined SVT+NSSA method against differential targeted enhancement in a mouse tumor model.

Main Methods:

• Utilized a Verasonics Vantage 256 system with a custom pulse inversion sequence for contrast image acquisition.
• Applied SVT to suppress static tissue signals while preserving microbubble signals.
• Employed NSSA to classify microbubble signals as adherent or non-adherent, achieving high accuracy (ROC AUC = 0.97).

Main Results:

• The combined SVT+NSSA method accurately differentiated microbubble signals from other signals (ROC AUC = 0.89).
• This approach matched the classification performance of differential targeted enhancement without contrast agent destruction.
• SVT suppressed static tissue signals by 9.6 dB.

Conclusions:

• SVT and NSSA provide an automated method for segmenting and classifying contrast signals in ultrasound molecular imaging.
• This filtering technique has potential for real-time application, improving workflow and accelerating clinical adoption.
• The SVT+NSSA method offers a non-destructive approach to enhance diagnostic capabilities.