Theoretical predictions of the generalized contrast-to-noise ratio for photoacoustic images

• Published in: 2020 IEEE International Ultrasonics Symposium (IUS) pp. 1 - 4
• Main Authors: Gubbi, Mardava R., Bell, Muyinatu A. Lediju
• Format: Conference Proceeding
• Language: English
• Published: IEEE 07.09.2020
• Subjects: image processing; In vivo; Laser theory; Measurement by laser beam; Photoacoustic imaging; Signal to noise ratio; Ultrasonic variables measurement; Upper bound
• ISSN: 1948-5727
• DOI: 10.1109/IUS46767.2020.9251750

Target detectability in photoacoustic imaging applications is traditionally analyzed using image quality metrics, such as signal-to-noise ratio, contrast, and contrast-to-noise ratio. These metrics are difficult to interpret in the context of target detectability due to their lack of an upper bound and their sensitivity to image manipulation techniques such as thresholding. The generalized contrast-to-noise ratio (gCNR) is a recently introduced metric designed to assess the probability of lesion detection in ultrasound images. Previous work used empirical models of target and background signals to demonstrate the applicability of gCNR to analyzing target detectability in photoacoustic images. In this work, a theoretical framework for gCNR prediction is developed and validated using simulated, experimental, and in vivo data. We compare predicted and measured gCNR values across variations in channel SNR, laser energy, and frame averaging. Mean absolute errors between predicted and measured gCNR values were 0.032 ± 0.052, 0.057 ± 0.127, and 0.023 ± 0.033 for 1,215 simulated images, 3,888 experimental images, and 810 in vivo delay-and-sum images, respectively, with channel SNRs ranging −40 dB to 40 dB. In addition, we explored relationships among gCNR, laser energy, and channel data frame averaging. Results have the potential to improve our understanding of minimum energy requirements when designing photoacoustic imaging systems. In addition, the theoretical gCNR prediction framework provides a promising foundation to improve the efficiency of presurgical tasks such as energy selection for photoacoustic-guided surgeries.