With the increasing number of diabetic patients, diabetic retinopathy has become a leading cause of vision loss worldwide. Currently, 532 nm panretinal laser photocoagulation is a primary treatment method for diabetic retinopathy. Because of the difficulties with existing technology in directly measuring the temperature distribution of the fundus during treatment, improper selection of laser parameters may damage normal fundus tissue and affect vision. Accordingly, a real three-dimensional (3D) entire-eye model is constructed using fundus optical coherence tomography (OCT) images, and the effects of the structural differences in fundus tissue on the fundus temperature under different laser incidence angles are studied via numerical simulation to provide a reference for the selection of laser parameters in actual treatment.

OCT images of the retina in the OCTA-500 dataset are segmented to obtain the key tissue layers for photocoagulation therapy, and a 3D voxel model of a real fundus is constructed. Anterior tissue is then added to establish a complete 3D voxel model of the eyeball. A 3D voxelized Monte Carlo simulation (MCVM) is next performed to obtain the propagation path and absorption distribution of the 532 nm laser in the eye. Based on the simulated absorption distribution results, the Pennes biological heat transfer equation is used to calculate the fundus temperature under a 532 nm laser. We are able to study the effects of the incidence angle and fundus structural differences on the fundus temperature distribution by changing both the photon incidence angle in the MCVM and the OCT image data of the fundus.

This study conducts a 3D MCVM to analyze the photon propagation characteristics derived from an eye model. The results indicate that the photon escape and absorption rates within the ocular tissues are 8.9%±0.06% and 91%± 0.05%, respectively. Notably, the energy absorbed by the retinal tissues accounts for approximately 94% of the total absorption, with the anterior segment of the eye contributing only 6% (Fig. 4). In addition, variations in the photon incident angle are found to significantly affect the retinal temperature. Specifically, for every 1° deviation, the peak retinal temperature decreases by an average of 1.8, 2.5, and 3.2 ℃, with a corresponding relative error in temperature rise of 11% (Fig. 7). In addition, under different OCT imaging conditions, distinct peak temperatures are observed across different individuals at the same laser incident angle. The temperature reduction per degree of deviation also exhibits individual variability. Of the studied retinal tissues, the maximum difference in temperature rise observed is 1.5 ℃, resulting in a 13% error relative to the corresponding retinal temperature rise (Fig. 9).

We construct a realistic 3D eyeball model utilizing OCT images and employ a 3D MCVM to simulate the distribution of laser energy within fundus tissue. The simulation results enable us to calculate the temperature variations under different laser incidence angles and fundus structures. Our findings show that the laser propagation trajectory in the anterior segment of the eye remains largely unchanged, with only 6% of the total light being absorbed by this region. By contrast, 94% of light absorption occurs in the fundus tissue, predominantly within the retinal pigment epithelial layer. This aligns with the mechanisms observed in actual photocoagulation treatments. In addition, we find that the refractive and scattering effects of ocular tissues significantly influence laser behavior. Specifically, a mere 1° deviation in the laser incidence angle can result in an error of approximately 11% in the peak temperature rise in the fundus. This underscores the importance of considering the incidence angle to ensure that simulation results accurately reflect the dynamics of actual photocoagulation procedures. Finally, variations in fundus structures are found to substantially affect the temperature simulation outcomes, with potential errors in temperature rise of up to 13%. Thus, the development of 3D fundus modeling based on OCT images is critical for enhancing the adaptability of numerical simulations to diverse patient anatomies. The methodologies and findings of this study can serve as valuable references for optimizing laser parameter selection in panretinal photocoagulation therapy.