An adaptive multi-modal hybrid model for classifying thyroid nodules by combining ultrasound and infrared thermal images

Table 3 Comparison of methods for classification of thyroid nodules

Row	Modality	Model	ACC	PRE	SEN	SPE	F1	F2
1	US	VGG	0.6838	0.7410	0.4881	0.8527	0.5885	0.7419
2	US	ResNet	0.7570	0.8316	0.5962	0.8958	0.6945	0.8140
3	US	Inception	0.7298	0.6744	0.8056	0.6644	0.7342	0.6885
4	US	ViT	0.7395	0.7710	0.6226	0.8404	0.6889	0.7854
5	IRT	VGG	0.7188	0.7705	0.5595	0.8562	0.6483	0.7741
6	IRT	ResNet	0.7098	0.7463	0.5660	0.8339	0.6438	0.7618
7	IRT	Inception	0.7482	0.7366	0.7103	0.7808	0.7232	0.7154
8	IRT	ViT	0.7378	0.7138	0.7245	0.7492	0.7191	0.7441
9(ours)	US + IRT	AmmH	0.9738	0.9699	0.9736	0.9739	0.9717	0.9738

ISSN: 1471-2105