Predicting Generalized Contrast-to-Noise Ratios in Frame-Averaged Photoacoustic Images

• Published in: 2022 IEEE International Ultrasonics Symposium (IUS) pp. 1 - 4
• Main Authors: Gubbi, Mardava R., Bell, Muyinatu A. Lediju
• Format: Conference Proceeding
• Language: English
• Published: IEEE 10.10.2022
• Subjects: Computer vision; image processing; Image quality; Imaging; Measurement uncertainty; Photoacoustic imaging; Statistical analysis; Ultrasonic imaging
• ISSN: 1948-5727
• DOI: 10.1109/IUS54386.2022.9958666

Our related journal article presents a theoretical framework to relate the performance of photoacoustic-based computer vision systems to the parameters of the underlying photoacoustic imaging systems. However, software-based adjustments were not previously included. In this paper, we extend the previous framework to include software-based noise reduction algorithms (e.g., frame averaging). We derive expressions for generalized contrast-to-noise ratio (gCNR) predictions in frame-averaged photoacoustic images. These gCNR predictions are validated against histogram-based gCNR measurements, with mean absolute errors of 4.4×10 −3 and 7.7×10 −3 in the phantom and ex vivo environments, respectively. This extension reduced the mean absolute prediction error by 7.74% and 27.15% when compared to predictions with the previous photoacoustic framework applied to frame-averaged photoacoustic images from phantom and ex vivo data, respectively. Results demonstrate that our extended framework more accurately predicts target detectability in frame-averaged photoacoustic images, which is beneficial for photoacoustic-based computer vision systems.