Fig. 1. Schematic of optically based reproduction of color information of soft tissues

Fig. 2. Key implementation steps

Fig. 3. Reconstruct the endoscopic surgery scene into a composite radiation field, and use the SASM to divide the endoscopic surgery scene into three regions: dynamic region, static region, and deformable region

Fig. 4. Overview of the general framework

Fig. 5. By promoting a more concentrated density distribution along each camera ray in the static component, the recovered background contains more highlights to enhance the realism of the image

Fig. 6. Experimental setup and robotic surgical simulation system

Fig. 7. Visualization results of segmentation in real surgical scenes

Fig. 8. Results of each soft tissue reconstruction in the endoscopic scenario of robotic surgery simulation (qualitative)

Fig. 9. Line graphs of color difference of each organ. (a) Lung color difference line graph; (b) kidney color difference line graph; (c) liver color difference line graph

Fig. 10. Different color difference representations indicate that the color of the reconstructed organ is close to the true color of biological soft tissue. (a) Photographed original kidney map; (b) 3D scatter plot of positional color difference; (c) contour map of magnification chromatic aberration; (d) 3D reconstructed kidney map; (e) vector chromatic aberration map in color coded map of R channel; (f) vector chromatic aberration map in color coded map of G channel

Fig. 11. Performance of reconstructed biological soft tissue improves over time (represented by increasing number of iterations)

Fig. 12. Influence of hyperparameters (the number of input images) on reconstruction performance. (a) PSNR; (b) SSIM; (c) MSSIM

Fig. 13. Influence of hyperparameters (number of hidden layers, regularization weights) in hash grids on reconstruction quality. (a) PSNR; (b) MAE; (c) NRMSE

Table 1. Dataset reconstructed by the introduction of supervised loss after each soft tissue structure is trained for 50000 times in each robotic surgery simulation scenario, and the dataset is compared with the ground truth