Tissue typing using ultrasound RF time series: experiments with animal tissue samples

Area of Science:

• Medical Imaging
• Biophysics
• Ultrasound Technology

Background:

• Tissue typing is crucial for various medical applications.
• Traditional ultrasound imaging relies on B-mode images, which may not capture all tissue characteristics.
• Radiofrequency (RF) echo signals contain rich information about tissue microstructure.

Purpose of the Study:

• To investigate the potential of radiofrequency (RF) ultrasound time series data for tissue typing.
• To explore the tissue typing information present in RF time series.
• To compare the efficacy of low-frequency (6.6 MHz) and high-frequency (55 MHz) ultrasound for tissue classification.

Main Methods:

• RF echo signals were recorded from animal tissues (bovine liver, pig liver, bovine muscle, chicken breast) with stationary probes.
• Time series of RF echoes were generated for each spatial sample.
• Spectral and fractal features of RF time series were extracted and analyzed using feedforward neural networks.
• Experiments were conducted at 6.6 MHz and 55 MHz, with variations in acoustic power and frame rate.

Main Results:

• High classification accuracies were achieved: 95.1% (6.6 MHz) and 98.1% (55 MHz) for two-class problems.
• Four-class classification yielded accuracies of 78.6% (6.6 MHz) and 86.5% (55 MHz).
• RF time series analysis outperformed traditional texture features from B-mode images (77.5% accuracy).
• Increased acoustic power and frame rate improved classification performance.

Conclusions:

• RF ultrasound time series data provide a viable method for ultrasound-based tissue typing.
• High-frequency ultrasound demonstrates superior performance for tissue classification using RF time series.
• Further research is needed to elucidate the underlying physical mechanisms.