Skip to main content
. 2022 Jun 28;12(7):973. doi: 10.3390/life12070973

Table A4.

Effect of different amounts of training data on the performance (mean ± standard deviation) of U-Nets trained using different color channels for segmenting CRBVs. Note that the CLAHE is applied in the pre-processing stage.

CHASE_DB1
Data Split Color Precision Recall AUC MIoU
25% Training,
20% Validation,
55% Test		 RGB 0.569 ± 0.203 0.448 ± 0.041 0.729 ± 0.059 0.537 ± 0.046
GRAY 0.615 ± 0.081 0.412 ± 0.041 0.735 ± 0.024 0.503 ± 0.051
RED 0.230 ± 0.030 0.332 ± 0.053 0.613 ± 0.010 0.474 ± 0.006
GREEN 0.782 ± 0.026 0.526 ± 0.020 0.792 ± 0.007 0.606 ± 0.045
BLUE 0.451 ± 0.114 0.370 ± 0.018 0.683 ± 0.032 0.485 ± 0.008
25% Training,
20% Validation,
25% Test		 RGB 0.571 ± 0.207 0.441 ± 0.045 0.724 ± 0.062 0.538 ± 0.048
GRAY 0.624 ± 0.080 0.407 ± 0.036 0.731 ± 0.023 0.502 ± 0.050
RED 0.244 ± 0.037 0.342 ± 0.046 0.619 ± 0.009 0.474 ± 0.007
GREEN 0.791 ± 0.026 0.515 ± 0.022 0.787 ± 0.008 0.602 ± 0.044
BLUE 0.449 ± 0.116 0.362 ± 0.017 0.677 ± 0.033 0.484 ± 0.008
55% Training,
20% Validation,
25% Test		 RGB 0.816 ± 0.012 0.541 ± 0.024 0.784 ± 0.012 0.684 ± 0.018
GRAY 0.803 ± 0.002 0.515 ± 0.026 0.775 ± 0.010 0.671 ± 0.016
RED 0.389 ± 0.039 0.363 ± 0.027 0.680 ± 0.021 0.504 ± 0.028
GREEN 0.838 ± 0.005 0.583 ± 0.017 0.806 ± 0.009 0.687 ± 0.038
BLUE 0.648 ± 0.019 0.383 ± 0.012 0.698 ± 0.006 0.601 ± 0.010
DRIVE
Data Split Color Precision Recall AUC MIoU
25% Training,
20% Validation,
55% Test		 RGB 0.796 ± 0.036 0.443 ± 0.065 0.749 ± 0.028 0.622 ± 0.072
GRAY 0.835 ± 0.016 0.419 ± 0.022 0.739 ± 0.009 0.590 ± 0.066
RED 0.362 ± 0.098 0.342 ± 0.072 0.628 ± 0.015 0.476 ± 0.007
GREEN 0.846 ± 0.010 0.463 ± 0.025 0.758 ± 0.009 0.671 ± 0.027
BLUE 0.537 ± 0.078 0.297 ± 0.028 0.660 ± 0.022 0.512 ± 0.026
25% Training,
20% Validation,
25% Test		 RGB 0.839 ± 0.035 0.442 ± 0.068 0.749 ± 0.030 0.626 ± 0.073
GRAY 0.874 ± 0.018 0.413 ± 0.023 0.737 ± 0.009 0.592 ± 0.068
RED 0.400 ± 0.108 0.352 ± 0.073 0.637 ± 0.014 0.476 ± 0.009
GREEN 0.896 ± 0.009 0.462 ± 0.025 0.760 ± 0.009 0.676 ± 0.028
BLUE 0.575 ± 0.080 0.300 ± 0.024 0.663 ± 0.020 0.512 ± 0.027
55% Training,
20% Validation,
25% Test		 RGB 0.896 ± 0.005 0.539 ± 0.010 0.787 ± 0.006 0.732 ± 0.014
GRAY 0.895 ± 0.004 0.528 ± 0.012 0.781 ± 0.005 0.731 ± 0.006
RED 0.660 ± 0.085 0.316 ± 0.037 0.674 ± 0.017 0.520 ± 0.038
GREEN 0.904 ± 0.003 0.533 ± 0.008 0.786 ± 0.003 0.718 ± 0.024
BLUE 0.783 ± 0.042 0.386 ± 0.044 0.705 ± 0.021 0.645 ± 0.037
HRF
Data Split Color Precision Recall AUC MIoU
25% Training,
20% Validation,
55% Test		 RGB 0.792 ± 0.006 0.537 ± 0.021 0.799 ± 0.013 0.597 ± 0.024
GRAY 0.776 ± 0.004 0.497 ± 0.017 0.781 ± 0.011 0.579 ± 0.025
RED 0.204 ± 0.024 0.258 ± 0.017 0.591 ± 0.014 0.467 ± 0.002
GREEN 0.821 ± 0.013 0.578 ± 0.012 0.824 ± 0.006 0.624 ± 0.037
BLUE 0.155 ± 0.002 0.361 ± 0.010 0.580 ± 0.001 0.482 ± 0.008
25% Training,
20% Validation,
25% Test		 RGB 0.759 ± 0.006 0.535 ± 0.023 0.797 ± 0.014 0.593 ± 0.023
GRAY 0.741 ± 0.005 0.503 ± 0.017 0.782 ± 0.011 0.576 ± 0.025
RED 0.197 ± 0.021 0.245 ± 0.017 0.586 ± 0.013 0.467 ± 0.002
GREEN 0.794 ± 0.016 0.581 ± 0.013 0.824 ± 0.006 0.619 ± 0.036
BLUE 0.149 ± 0.004 0.368 ± 0.013 0.578 ± 0.002 0.480 ± 0.007
55% Training,
20% Validation,
25% Test		 RGB 0.781 ± 0.008 0.608 ± 0.005 0.824 ± 0.004 0.693 ± 0.013
GRAY 0.768 ± 0.010 0.573 ± 0.017 0.807 ± 0.009 0.677 ± 0.022
RED 0.512 ± 0.009 0.271 ± 0.021 0.641 ± 0.013 0.536 ± 0.011
GREEN 0.788 ± 0.006 0.647 ± 0.009 0.846 ± 0.003 0.674 ± 0.060
BLUE 0.274 ± 0.110 0.341 ± 0.047 0.620 ± 0.032 0.500 ± 0.019
STARE
Data Split Color Precision Recall AUC MIoU
25% Training,
20% Validation,
55% Test		 RGB 0.556 ± 0.204 0.300 ± 0.073 0.659 ± 0.073 0.478 ± 0.008
GRAY 0.619 ± 0.050 0.283 ± 0.058 0.680 ± 0.033 0.478 ± 0.017
RED 0.148 ± 0.003 0.222 ± 0.033 0.516 ± 0.009 0.468 ± 0.000
GREEN 0.600 ± 0.242 0.351 ± 0.030 0.680 ± 0.082 0.483 ± 0.019
BLUE 0.167 ± 0.036 0.145 ± 0.034 0.518 ± 0.021 0.469 ± 0.001
25% Training,
20% Validation,
25% Test		 RGB 0.531 ± 0.195 0.334 ± 0.082 0.672 ± 0.082 0.482 ± 0.009
GRAY 0.607 ± 0.055 0.314 ± 0.066 0.691 ± 0.039 0.483 ± 0.020
RED 0.143 ± 0.003 0.231 ± 0.038 0.512 ± 0.011 0.471 ± 0.000
GREEN 0.587 ± 0.243 0.376 ± 0.048 0.688 ± 0.092 0.488 ± 0.024
BLUE 0.164 ± 0.032 0.142 ± 0.039 0.517 ± 0.020 0.472 ± 0.001
55% Training,
20% Validation,
25% Test		 RGB 0.756 ± 0.014 0.448 ± 0.031 0.749 ± 0.015 0.610 ± 0.038
GRAY 0.748 ± 0.010 0.504 ± 0.026 0.770 ± 0.010 0.656 ± 0.017
RED 0.181 ± 0.020 0.293 ± 0.069 0.558 ± 0.008 0.474 ± 0.006
GREEN 0.749 ± 0.013 0.550 ± 0.025 0.795 ± 0.012 0.659 ± 0.038
BLUE 0.163 ± 0.007 0.324 ± 0.059 0.547 ± 0.006 0.469 ± 0.004
UoA_DR
Data Split Color Precision Recall AUC MIoU
25% Training,
20% Validation,
55% Test		 RGB 0.320 ± 0.011 0.398 ± 0.008 0.699 ± 0.006 0.541 ± 0.015
GRAY 0.315 ± 0.011 0.353 ± 0.016 0.675 ± 0.007 0.526 ± 0.017
RED 0.203 ± 0.013 0.260 ± 0.016 0.614 ± 0.006 0.516 ± 0.005
GREEN 0.332 ± 0.007 0.415 ± 0.018 0.705 ± 0.009 0.534 ± 0.014
BLUE 0.237 ± 0.012 0.260 ± 0.008 0.620 ± 0.007 0.526 ± 0.006
25% Training,
20% Validation,
25% Test		 RGB 0.313 ± 0.011 0.395 ± 0.008 0.697 ± 0.005 0.540 ± 0.015
GRAY 0.306 ± 0.011 0.350 ± 0.016 0.673 ± 0.008 0.524 ± 0.017
RED 0.201 ± 0.013 0.259 ± 0.015 0.614 ± 0.006 0.516 ± 0.005
GREEN 0.326 ± 0.007 0.412 ± 0.017 0.704 ± 0.009 0.532 ± 0.014
BLUE 0.232 ± 0.011 0.257 ± 0.007 0.618 ± 0.006 0.524 ± 0.005
55% Training,
20% Validation,
25% Test		 RGB 0.333 ± 0.005 0.445 ± 0.012 0.717 ± 0.004 0.557 ± 0.007
GRAY 0.330 ± 0.003 0.413 ± 0.014 0.700 ± 0.006 0.559 ± 0.004
RED 0.289 ± 0.011 0.299 ± 0.007 0.641 ± 0.004 0.543 ± 0.003
GREEN 0.335 ± 0.002 0.470 ± 0.010 0.728 ± 0.004 0.564 ± 0.004
BLUE 0.281 ± 0.012 0.280 ± 0.013 0.630 ± 0.006 0.540 ± 0.004