A Self-supervised Learning Reconstruction Algorithm with an Encoder-Decoder Architecture for Diffuse Optical Tomography

Diffuse optical tomography (DOT) is an emerging non-invasive optical imaging technique, which has a promising application in breast cancer detection and diagnosis. However, the conventional image reconstruction algorithm in DOT is time-consuming and easy to error when recovering the distribution of optical parameters within the complete tissue. In this paper, we present an end-to-end reconstruction algorithm for DOT based on a deep convolutional encoder-decoder architecture, which consists of a data processing part and a convolutional encoder-decoder net. Its effectiveness was evaluated using simulation data. The results show that the overall quality of our method is significantly improved compared with the traditional algorithm based on the FEM method, the single inclusion deviation is reduced by 150% compared with the traditional algorithm, the standard deviation is reduced by 50%; multiple inclusions deviation is reduced by 100% and the standard deviation by 38.7%.