IRR-Net: A Joint Learning Framework for Image Reconstruction and Recognition of Photoacoustic Tomography

Table 7

Evaluation metrics for binary classification of experimental phantom images.

Method		Acc	Sens	Spec	F1 score	Runtime (s)
HH (5 DGD + GoogLeNet)		0.9211	0.9338	0.8879	0.9502	23.6
IRR-Net		0.9108	0.9237	0.8721	0.9395	11.8
HH (DGD + U-Net+ResNet)		0.9108	0.9237	0.8721	0.9395	30.2
LH	1 DGD + GoogLeNet	0.8907	0.9123	0.8516	0.9206	10.3
	TR + GoogLeNet	0.8743	0.8873	0.8354	0.9137	3.1
	TR + AlexNet	0.8562	0.8694	0.8166	0.9007	2.6
	TR + ResNet	0.8511	0.8647	0.8132	0.8916	2.6
HL (5 DGD + SVM)		0.8461	0.8573	0.8094	0.8812	22.7
LL (TR + SVM)		0.8088	0.8221	0.7689	0.8658	1.5