Spatiotemporal tensor reconstruction for echocardiography: Effects of multidimensional structure and motion

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Signal Processing

Background:

• Cardiac ultrasound imaging faces limitations due to sampling requirements.
• Accelerating ultrasound acquisition is desirable for capturing faster dynamics and larger fields of view.
• The spatial and temporal properties of ultrasound images enabling acceleration are not well understood.

Purpose of the Study:

• To fundamentally study ultrasound image structure for accelerated sampling.
• To quantify tensor rank and spatial/temporal roughness in cardiac ultrasound data.
• To explore the implications of these properties for tensor completion algorithms.

Main Methods:

• Utilized simulations, phantoms, and in vivo cardiac data.
• Quantified tensor rank and spatial/temporal roughness.
• Evaluated inverse distance-weighted (IDW) interpolation and fast multiway delay-embedding transform for reconstruction.
• Assessed the impact of motion compensation.

Main Results:

• Ultrasound data were not sufficiently low-rank for high-quality low-rank reconstruction.
• IDW interpolation and fast multiway delay-embedding transform showed superior reconstruction accuracy compared to low-rank methods.
• Spatial and temporal roughness were found to inversely correlate with tensor completion success.
• Motion compensation reduced temporal roughness and rank, improving reconstruction.

Conclusions:

• Advanced statistical algorithms can circumvent traditional sampling limits in cardiac ultrasound.
• Local information-based methods are more effective than low-rank approaches for accelerated ultrasound reconstruction.
• Understanding and mitigating image roughness, alongside motion compensation, is crucial for successful accelerated cardiac ultrasound imaging.