Deep learning enabled real-time photoacoustic tomography system via single data acquisition channel.

Area of Science:

• Biomedical Imaging
• Medical Technology
• Optical Physics

Background:

• Photoacoustic computed tomography (PACT) integrates optical contrast with ultrasound penetrability.
• Traditional PACT systems often require expensive multi-channel data acquisition (DAQ) hardware.
• There is a need for cost-effective and real-time PACT solutions.

Purpose of the Study:

• To develop a novel PACT system enabling real-time imaging with a significantly reduced DAQ channel count.
• To demonstrate the feasibility of a cost-effective, single-channel DAQ PACT system.
• To validate the system's performance using phantom and in-vivo experiments.

Main Methods:

• A 120-element ultrasound array was employed, with signals superimposed in the analog domain to reduce channels from 120 to 4.
• A four-to-one delay-line module combined the 4 channels into a single channel for acquisition.
• A deep learning model was trained for image reconstruction from the superimposed signals.
• Signal decoupling was performed post-acquisition.

Main Results:

• The developed PACT system successfully achieved real-time imaging capabilities.
• Preliminary phantom and in-vivo experiments demonstrated robust imaging performance.
• The system drastically reduced the DAQ module cost by using only one channel instead of 120.

Conclusions:

• The novel PACT system offers a significant reduction in hardware cost, moving towards portable and affordable real-time medical imaging.
• This approach democratizes access to PACT technology.
• The integration of analog signal superposition and deep learning reconstruction is effective for simplified PACT.