Table 1.
Nerve Segmentation DSC | Bright lighting 0.421 (± 0.054, n = 155) |
Medium lighting 0.573 (± 0.079, n = 70) |
Dim lighting 0.395 (± 0.097, n = 52) |
---|---|---|---|
Far-away 0.360 (± 0.058, n = 138) |
0.343 (± 0.077, n = 78) | 0.395 (± 0.136, n = 30) | 0.369 (± 0.129, n = 30) |
Close-up 0.548 (± 0.054, n = 139) |
0.500 (± 0.074, n = 77) | 0.707 (± 0.075, n = 40) | 0.430 (± 0.159, n = 22) |
Quantitative results of micro-averaged segmentation dice similarity coefficient (DSC) across all 5 cross-validation test folds. Evaluation examines all combinations of image conditions (picture distance and brightness) with 95% confidence intervals. Confidence intervals are calculated using the standard error of the n averaged dice scores. Best-performing condition is bolded. The algorithm had the best performance on images that were close-up to the surgical anatomy and had medium lighting.