|
SN
|
Symbol
|
Description of the Symbols
|
| 1 |
|
Cross Entropy-loss |
| 2 |
m |
Model used for segmentation in the total number of models M |
| 3 |
n |
Image scan number in total number N
|
| 4 |
|
Mean estimated lung area for all images using AI model ‘m’ |
| 5 |
|
Estimated Lung Area using AI model ‘m’ and image ‘n’ |
| 6 |
|
GT lung area for image ‘n’ |
| 7 |
|
Mean ground truth area for all images N in the database |
| 8 |
|
Mean estimated lung long axis for all images using AI model ‘m’ |
| 9 |
|
Estimated lung long axis using AI model ‘m’ and image ‘n’ |
| 10 |
|
GT lung long axis for image ‘n’ |
| 11 |
|
Mean ground truth long axis for all images N in the database |
| 12 |
|
Figure-of-Merit for segmentation model ‘m’ |
| 14 |
|
Figure-of-Merit for long axis for model ‘m’ |
| 15 |
JI |
Jaccard Index for a specific segmentation model |
| 16 |
DSC |
Dice Similarity Coefficient for a specific segmentation model |
| 17 |
TP, TN |
True Positive and True Negative |
| 18 |
FP, FN |
False Positive and False Negative |
| 19 |
x
i
|
GT label |
| 20 |
pi
|
SoftMax classifier probability |
| 21 |
Yp
|
Ground truth image |
| 22 |
|
Estimated image |
| 23 |
P
|
Total no of pixels in an image in x, y-direction |
| 24 |
K5 |
Cross-validation protocol with 80% training and 20% testing (5 folds) |
|
Deep Learning Segmentation Architectures
|
| 25 |
PSP Net |
SDL model for lung segmentation with pyramidal feature extraction |
| 26 |
VGG-SegNet |
HDL model designed by fusion of VGG-19 and SegNet architecture |
| 27 |
ResNet-SegNet |
HDL model designed by fusion of ResNet-50 and SegNet architecture |