Development of 3-D photoacoustic image reconstruction algorithms for real-time operation on GPU

Abstract

Acoustic resolution photoacoustic microscopy (AR-PAM) is a type of photoacoustic microscopy (PAM) that has an imaging depth of several centimeters and can provide information such as deep blood vessel imaging and oxygen saturation. To increase the resolution of AR-PAM, various signal processing methods are used, and the synthetic aperture focusing technique (SAFT) is one of these methods. However, high computation cost is a major obstacle to the general use of SAFT for high-quality AR-PAM imaging. Note that the amount of computation involved in 2D-SAFT is equal to the square of the 1D-SAFT. In this study, to reduce the computation time of the SAFT, the change in resolution according to the synthesis range was evaluated and the optimal synthesis range was confirmed because the synthesis range mainly determines the computation time of the SAFT. For graphics processing unit (GPU)-based acceleration of the computation for 2D-SAFT using the obtained optimal synthesis range, the SAFT algorithm was modified

data collection and computation can be performed simultaneously, thus significantly reducing the time to acquire the image data. From the experiments, it was confirmed that the computation speed was increased by 140 times, while maintaining the resolution. In conclusion, the proposed 2D-SAFT on GPU facilitates real-time operation as well as spatial resolution enhancement. |AR-PAM(Acoustic resolution photoacoustic microscopy)은 광음향 현미경(PAM)의 일종으로 영상 깊이가 수 센티미터이고 심부 혈관 영상 및 산소 포화도와 같은 정보를 제공할 수 있습니다. AR-PAM의 해상도를 높이기 위해 다양한 신호 처리 방법이 사용되며 그 중 하나가 합성 조리개 알고리즘입니다. 그러나 2D 합성 조리개 포커싱 기법은 1D-SAFT의 제곱에 해당하는 연산량을 가지므로 처리에 많은 시간이 소요된다. 본 연구에서는 합성 조리개 포커싱 기법의 계산 시간을 줄이기 위해 합성 조리개 포커싱 기법에서 사용하는 합성 범위에 따른 해상도 변화를 확인하고 최적 합성 범위를 확인하였다. 획득한 최적의 합성 범위를 GPU를 통해 합성 조리개 알고리즘의 연산을 가속화하여 데이터 수집과 연산을 동시에 수행할 수 있도록 알고리즘을 수정하여 영상 획득 시간을 크게 단축하였다.