Learning-based enhancement of limited-view optoacoustic tomography based on image- and time-domain data

Optoacoustic images are often afflicted with distortions and artifacts corresponding to system limitations, including limited-view tomographic data. We developed a convolutional neural network (CNN) approach for optoacoustic image quality enhancement combining training on both time-resolved signals and tomographic reconstructions. Reference human finger data for training the CNN were recorded using a full-ring array system with optimal tomographic coverage. The reconstructions were further refined with a dedicated algorithm that minimizes acoustic reflection artifacts induced by acoustically mismatch structures, such as bones. The combined methodology is shown to outperform other CNN-based methods solely operating on image-domain data.