Fig. 1. Setup of training dataset. (a) Two-layer brain-emulating model; (b) task-related HbO and HbR concentration change curves and the absorption perturbation in CC layer when θT=1; (c) time courses of the absorption perturbation induced by the physiological interferences in SS layer when θI=1

Fig. 2. DCNN model structure

Fig. 3. Simulation results of DCNN under weak interference. (a) A comparison of reconstructed absorption efficiency perturbation images in CC-layer; (b) quantitative evaluation of reconstruction

Fig. 4. Simulation results of DCNN under strong interference

Fig. 5. Reconstructed images of double targets

Fig. 6. Setup of phantom experiments. (a) Structural representation of the polyformaldehyde phantom; (b) picture of the polyformaldehyde phantom

Fig. 7. Phantom experimental results

Fig. 8. 3D-DCNN model structure

Fig. 9. Simulation results of 3D-DCNN. (a) A comparison of absorption coefficient perturbation images in CC layer reconstructed at the selected time points; (b) quantitative estimation comparison of reconstruction results at the selected time points; (c) time-courses of average absorption coefficient perturbation in the deep activated region