Effective automatic detection of anterior cruciate ligament injury using convolutional neural network with two attention mechanism modules

Table 2 Evaluation of CNN Models. Our model performed well in accuracy, sensitivity, AUC, F1-score and had lighter parameters

Model	Accuracy	Precision	Sensitivity	Specificity	AUC	F1-score	Parameters
MobileNet	0.6397	0.6975	0.6366	0.6036	0.8335	0.7979	5.35M
EfficientNet-B0	0.7413	0.7500	0.8113	0.6059	0.7820	0.7794	5.22M
EfficientNet-B1	0.7460	0.7481	0.8282	0.6400	0.8126	0.7861	7.72M
VGG	0.7619	0.8213	0.7380	0.7927	0.8523	0.7774	15.29M
ResNet-34	0.7524	0.7386	0.8676	0.6036	0.8335	0.7979	22.36M
ResNet-50	0.7889	0.7921	0.8479	0.7127	0.8627	0.8190	25.32M
Ours	0.8063	0.7741	0.9268	0.6509	0.8886	0.8436	2.23M

ISSN: 1471-2342