Comprehensive framework of GPU-accelerated image reconstruction for photoacoustic computed tomography

Area of Science:

• Medical Imaging
• Computational Imaging
• Biomedical Engineering

Background:

• Photoacoustic computed tomography (PACT) is a vital non-invasive imaging modality for life sciences and clinical use.
• Increasingly large ultrasound transducer (UST) arrays in PACT systems generate massive data, challenging fast image reconstruction.
• Existing GPU-accelerated PACT methods lack explicit, platform-agnostic implementation guidance.

Purpose of the Study:

• To present a comprehensive framework for developing GPU-accelerated PACT image reconstruction.
• To provide the research community with a clear understanding of advanced PACT image reconstruction techniques.
• To facilitate the adoption of efficient PACT reconstruction methods across various hardware platforms.

Main Methods:

• Leveraging open-source parallel computing tools: Python multiprocessing, Taichi Lang, and CUDA.
• Implementing parallel computing strategies for multicore CPUs, single GPUs, and multi-GPU setups.
• Developing a framework adaptable to diverse hardware configurations for PACT reconstruction.

Main Results:

• Achieved significant performance enhancement in PACT reconstruction using the proposed framework.
• Demonstrated effective acceleration on a dual-GPU platform compared to a high-core-count CPU for large-scale 3D PACT images.
• Shared example code on GitHub to promote community adoption and adaptation.

Conclusions:

• The developed framework significantly accelerates PACT image reconstruction, enabling faster analysis and real-time applications.
• The approach is designed for easy adoption, fostering broader implementation of PACT in life science and medicine.
• This work empowers researchers to harness advanced computational techniques for improved PACT imaging.