D2-RecST: Dual-domain joint reconstruction strategy for fluorescence molecular tomography based on image domain and perception domain

•We propose an effective adversarial learning strategy in the image domain to enhance the reconstruction robustness.•We propose a novel transfer learning strategy in the perceptual domain to improve the reconstruction accuracy.•The adversarial learning strategy is combined with the transfer learning strategy to form the dual-domain joint reconstruction strategy to jointly improve the FMT reconstruction performance.•Prior knowledge of reconstructed fluorescent sources could be transferred to the FMT reconstruction to optimize the structural edge details.•The perceptual loss works together with the adversarial loss to mitigate the severely ill-posed nature of the FMT inverse problem.