Table 7.

Comparison state of the art works with our proposed model.

Author, Year	Model	Dataset	Target Output	Evaluation
				Accuracy	Sensitivity	Precision	Specificity	AUC
Štajduhar et al. [23]	HOG+ Linear-kernel SVM (k = 10)	KneeMRI 917 (exams)	partial tear	-	-	-	-	0.894
			ruptured tear	-	-	-	-	0.943
Bien et al., 2018 [27]	AlexNet	MRNet 1370 exams	ACL tear	0.867	0.759	-	0.968	0.965
			abnormal	0.850	0.879	-	0.714	0.937
			meniscus tear	0.725	0.892		0.741	0.847
	Logistic Regression	KneeMRI 917 exam	partial tear, ruptured tear	-	-	-	-	0.911
Chang et al., 2019 [33]	Dynamic patch + ResNet	260 MRI coronal volumes	partial AC, full torn	0.967	1.00	0.938	0.933	-
Liu et al., 2019 [34]	VGG16	sagittal MR 175 (exams)	full thickness ACL tear, Intact ACL	-	0.92	-	0.92	0.95
	DenseNet			-	0.96	-	0.96	0.98
	Alex Net			-	0.89	-	0.88	0.90
Namiri et al., 2019 [39]	2D CNN 3D CNN	NIH MRI 1243 (exams)	Intact ACL	-	0.22 0.89	-	0.90 0.88	-
	2D CNN 3D CNN		Partial tear	-	0.75 0.25	-	1.00 0.92	-
	2D CNN 3D CNN		Full tear	-	0.82 0.76	-	0.94 1.00	-
Zhang et al., 2020 [40]	3D DenseNet	sagittal MR 408 (exams)	ACL tears Intact ACL	0.957 0.943 0.899	0.976 0.952 0.912	0.940 0.952 0.869	0.944 0.909 0.886	0.960 0.946 0.859
	ResNet
	VGG16
Irmakci et al., 2020 [41]	AlexNet	MRNet 1370 exams	abnormal	0.8583	0.978	-	0.400	0.891
			ACL tear	0.833	0.685	-	0.954	0.938
	ResNet-18		abnormal	0.825	0.968	-	0.280	0.811
			ACL tear	0.866	0.777	-	0.939	0.954
	GoogleLeNet		abnormal	0.833	0.978	-	0.280	0.909
			ACL tear	0.808	0.666	-	0.924	0.890
Tsai et al., 2020 [42]	EfficientNet	MRNet 1370	abnormal	0.917	0.968	-	0.72	0.941
			ACL tear	0.904	0.923	-	0.891	0.960
	ELNet 5 -fold	KneeMRI 917 exams	ruptured ACL	-	-	-	-	0.913
Proposed Customized ResNet-14 5-fold cross-validation		KneeMRI 917 exams	ACL Intact	0.92	0.89	0.92	0.93	0.98
			partial tear	0.91	0.87	0.87	0.92	0.97
			ruptured	0.93	0.99	0.96	0.99	0.99