Generalized SNR: Considerations for Lesion Detectability for Modern Ultrasound Beamformers
Sunday, April 7, 2024
11:25am – 11:50am
Location: Governor's A - 4th Floor
Historically, image quality metrics have been designed with the expectation that they correlate with an expert's subjective opinion of image quality. However, image quality metrics that were designed assuming DAS beamformed ultrasound data are often not optimal for modern adaptive beamformers due to the breakdown of this assumption. This work investigates lesion detectability in ultrasound from the perspective of what makes an ideal objective image quality metric, and proposes a generalized signal-to-noise ratio that correlates with subjective assessment, is robust in cases where the data is not DAS, and incorporates both the lesion contrast and size and system resolution, all components that are generally understood to impact lesion detection. This work is meant to encourage others to consider the potential consequences of using older image quality metrics on newer beamforming approaches and the complications that could arise.