Adaptive Models for Multi-Covariate Imaging of Sub-Resolution Targets (MIST).

Area of Science:

• Medical Imaging
• Ultrasound Technology
• Signal Processing

Background:

• Multi-covariate Imaging of Sub-resolution Targets (MIST) is a statistical image formation technique.
• Existing MIST models rely on assumptions about transmit apodization and backscatter covariance.
• Patient-specific factors like nonlinear propagation and wavefront aberration can deviate effective transmit apertures from applied ones.

Purpose of the Study:

• To generalize and extend MIST by developing data-adaptive covariance estimation, parameterization, and model-formation techniques.
• To evaluate the performance gains of adaptive MIST compared to standard MIST and B-mode imaging.

Main Methods:

• Developed data-adaptive covariance estimation and model-formation techniques for MIST.
• Applied adaptive MIST to 152 tissue-harmonic scans of fetal targets from 15 healthy pregnant subjects.
• Utilized a versatile empirical function to parameterize speckle covariance and estimate coherence loss.

Main Results:

• Adaptive MIST improved contrast-to-noise ratio (CNR) by a median of 8.3% and speckle signal-to-noise ratio (SNR) by 9.7% over standard MIST.
• Median CNR and SNR gains over B-mode improved from 29.4% to 40.4% and 24.7% to 38.3%, respectively.
• Demonstrated the ability to estimate effective coherent aperture size and coherence loss.

Conclusions:

• Data-adaptive covariance estimation significantly enhances MIST performance in ultrasound imaging.
• The developed methods offer improved image quality metrics (CNR, SNR) in fetal imaging.
• The proposed adaptive MIST approach shows potential for system-independent implementation.