ESTRO 2024 - Abstract Book

S3767

Physics - Image acquisition and processing

ESTRO 2024

A linear relation, with small variation between subjects, was found between the camera angle and retinal location, expressed as an angle with respect to the patient’s second nodal point. At a camera angle of 40°, for example, a retinal angle of 40° with a standard deviation of 0.2° between subjects was found. For other reference points, such as the eye center, larger variations were observed. For the subsequent non-raytracing mapping method, this linear relation was used as a basis to determine the corresponding retinal location for each location on the fundus image. At all camera angles, the systematic and random errors for this second nodal point method were smaller than those for the other methods. At a camera angle of 40°, the difference between real (ray-traced) and predicted camera angle for the second nodal point method was 0.05° (mean, SD = 0.2°), for the EYEPLAN method 9.6° (SD = 0.7°) and for the OPTOS method -2.9° (SD = 2.1°). For the example patient in Figure 2, a clinically significant 1 mm difference was observed when this second nodal point scaling was not applied.

Figure 1: A) Ray tracing result for a single subject. The parallel dashed lines show that the camera angles and second nodal point angles are equal. B) Compared with other methods, the second nodal point-based method provides a more accurate mapping between retinal locations and camera angles. Error bands represent the standard deviation.

Figure 2: Difference between peripheral scaling based on the nodal point method (green circles) and the conventional fundus camera calibration (blue circles) on a fundus photograph of a uveal melanoma patient. The second nodal point