Skip to main content
. 2021 Nov 24;11:22814. doi: 10.1038/s41598-021-02329-5

Table 3.

The comparison of alpha-shape descriptors and FD obtained from manual annotation and automatic segmentation.

Parameter Healthy (n = 15, DSC = 0.81870 ± 0.02360) DR (n = 15, DSC = 0.58530 ± 0.0003) DR vs Healthy (p-value*)
Manual Automatic MRE (max, p-value+) Manual Automatic MRE (max, p-value+) M A
Opαmin, mean ± SD 41.1 ± 5.6 52.4 ± 8.5 29% (80%, 0.476) 44.0 ± 9.4 58.0 ± 17.1 37% (100%, 0.235) 0.595 0.648
OpA, mean ± SD 1,063,768 ± 127,691 1,034, 863 ± 110,756 10% (24%, 0.144) 1,046,821 ± 88,366 1,004,089 ± 118,917 10% (26%, 0.281) 0.367 0.539
VS, mean ± SD 16,062 ± 3182 20,021 ± 2464 22% (39%, 0.087) 24,516 ± 4270 18,113 ± 3,379 26% (49%, 0.008) 0.305 0.202
Gradα, mean ± SD 366 ± 54 250 ± 57 31% (55%, 0.005) 362 ± 89 213 ± 62 43% (66%, 0.064) 0.967 0.249
FD, mean ± SD 1.47 ± 0.01 1.46 ± 0.01 0.9% (2%, 0.002) 1.47 ± 0.01

1.45

 ± 0.01

1.4% (3%, 0.004) 0.01 0.01

DSC dice similarity coefficient with manual annotation as reference, MRE mean of the relative error with manual annotation as reference, Max maximum relative error with manual annotation as reference. n, number of images.

+Pearson correlation test with null hypothesis that the correlation coefficient is zero.

*Wilcoxon rank sum with null hypothesis that medians are equal between groups.