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

Table 4

Evaluation metrics for multiclass classification in the simulation study.

Method		Acc				F1 score	Runtime (s)
		Cal	FC	LP	MC
HH (5 DGD + GoogLeNet)		0.9243	0.9146	0.9322	0.8801	0.8980	29.1
IRR-Net		0.9191	0.9089	0.9161	0.8702	0.8878	10.7
HH (DGD + U-Net+ResNet)		0.9191	0.9089	0.9161	0.8702	0.8878	33.6
LH	1 DGD + GoogLeNet	0.9002	0.8913	0.8867	0.8561	0.8705	9.8
	TR + GoogLeNet	0.8976	0.8891	0.8555	0.8405	0.8627	3.1
	TR + AlexNet	0.8776	0.8702	0.8415	0.8245	0.8229	2.7
	TR + ResNet	0.8714	0.8649	0.8358	0.8191	0.8203	2.7
HL (5 DGD + SVM)		0.8617	0.8587	0.8302	0.8238	0.8189	28.6
LL (TR + SVM)		0.8575	0.8461	0.8266	0.8212	0.7968	1.4

CAL, FC, LP, and MC stand for calcification, lipid pool, fibrous cap, and mixed calcification, respectively.