Temporal SNR Guided Adaptive B-mode and Color Doppler Transmit Parameter Selection

Abstract

Ultrasound is a dominant modality in medical imaging. The use of ultrasound continues to increase, and improved methods and new applications continue to be developed. Given the prevalence of ultrasound and its increasing use, streamlining scans, making imaging more straightforward for operators, and improving the quality of the acquisitions that are performed would have significant impact. One task currently standing in contrast to these goals is the adjustment of imaging parameters. Ideally, imaging parameters should be adjusted with each patient and imaging view to optimize quality, yielding the greatest likelihood for successful diagnosis. However, instead there is a reliance on preset imaging parameter settings that perform well on average, but do not necessarily optimize an individual scan. This work explores a method for automated adjustment of two imaging parameters, frequency and acoustic output, achieving optimal settings for one or both parameters during B-mode and color Doppler imaging without needing operator intervention. Frequency presents a tradeoff between data quality and imaging resolution. Acoustic output settings have a tradeoff between data quality and acoustic exposure levels. Acoustic output can be described by the Mechanical Index (MI), which is based on the pulse pressure, or Thermal Index (TI), a measure of tissue heating. Following the ALARA (As Low As Reasonably Achievable) principle, acoustic output should be kept as low as possible to minimize any risks associated with imaging. Implementing the adaptive parameter selection methods explored here in clinical settings could make ultrasound more accessible for operators, increase its availability to patients, and allow for the continued growth of this modality.The method explored here for frequency and acoustic output adjustment relies on real-time assessment of temporal signal-to-noise ratio (SNR). Temporal SNR describes the quality of the received data and is directly related to frequency and acoustic output levels. Higher frequencies experience greater attenuation leading to lower signal levels, while decreasing acoustic output leads to less intense receive echo signals. A human observer study was performed with simulated ultrasound data to assess at what temporal SNR noise becomes imperceptible during B-mode imaging. This threshold SNR was determined to be 28 dB for typical imaging conditions. An imaging sequence was implemented on a clinical ultrasound system that measured SNR during live imaging and performed real-time acoustic output adjustment, targeting the 28 dB SNR threshold. Adjustments minimally impacted the system operator, while automatically enforcing the ALARA principle. The recommended acoustic output level ranged from 0.1-0.4 MI, significantly lower than the 0.8-1.4 MI typically used by default on clinical systems. To extend the work done in B-mode imaging, similar adaptive adjustment methods were evaluated during color Doppler imaging. Both frequency and acoustic output were considered for adaptive adjustment in color Doppler. A simulation study was performed to assess velocity estimation accuracy under a variety of imaging settings and temporal SNR levels. At low SNR levels, velocity estimation performance was poor. Increasing SNR to 5 dB led to low levels of bias in the velocity estimate. With further SNR increases to 15-20 dB, the standard deviation in the velocity estimates across the vessel were near the noise-free minimum. These results were validated in phantom imaging experiments. In an in vivo clinical test, sweeps of acoustic output and frequency were performed in five volunteers. The SNR and color Doppler velocity estimates were calculated for each frequency and acoustic output combination. The highest frequency achieving SNR at or above the target level was recommended for imaging. From there, acoustic output was reduced to meet the target SNR threshold. Relative to their values for a potential default condition of mid-level (2.5 MHz) frequency imaging with 100% transmit power, MI was reduced 24-45% and TI reduced 68-69%. These results further supported the simulation and phantom findings, indicating SNR could guide the in vivo adjustment of frequency and acoustic output for color Doppler. The results of these studies demonstrate the promise of adaptive frequency and acoustic output adjustment and encourage the exploration of additional adaptive imaging methods. Demonstrating that these techniques could be implemented on existing clinical systems makes clear that temporal SNR guided adaptive imaging has the potential to assist operators in the short-term, while offering opportunity for further expanded benefits with continued development.