Deep-Learning-Based Scattering-Enhanced Dynamic Microwave-Induced Thermoacoustic Tomography Applying Single Measurement

The rapid advancement of microwave-induced thermoacoustic tomography (MITAT) technology has shown significant potential in many biomedical application disciplines. However, conventional MITAT modalities struggle to achieve an optimal balance among quantity of measurement data, sampling speed, and system cost. This renders it difficult to apply MITAT to dynamic imaging, which is a highly active frontier in biomedical imaging. This work proposes a novel MITAT modality leveraging artificially introduced wave scattering and deep learning (DL) to perform dynamic imaging based on only one measurement. We name this technique as DL-based scattering-enhanced MITAT, that is, DL-SE-MITAT. The scattering induced by many scatterers can efficiently increase the amount of received information of the sample being imaged and improve image quality. The DL technique can facilitate the imaging and enhance image quality applying scattered signals using only one measurement. We perform proof-of-concept validation of the DL-SE-MITAT method for dynamic imaging by investigating saline solution flowing in tube-shaped blood vessel phantoms. Experimental results, including both images and videos, clearly demonstrate the efficacy of the DL-SE-MITAT method in faithfully reconstructing the time-varying sample at a frame rate of 1 frames/s. It is also found that deploying more scatterers can notably enhance the quality of images. This work presents a new avenue for dynamic imaging with single measurement and contributes to the potential applications of MITAT in various dynamic imaging fields as well as other microwave-related imaging methods.