Table 5.

Classification results of each model on our dataset. “M” means $\times 10^6$. Batch time means the runtime of each batch in the model testing.

Method	Modality	Accuracy	Precision	Sensitivity	Specificity	F1-score	AUC	Parameters (a.k.a., model size)	Batch time
CAAD [35]	XR	0.7394	0.7430	0.7659	0.7791	0.7543	0.7488	138.3M	2.06 s
COVID-CAPS [82]	XR	0.7413	0.7450	0.7689	0.7844	0.7568	0.7540	23M	2.09 s
COVID-DA [83]	XR	0.7446	0.7471	0.7692	0.7844	0.7580	0.7593	33M	2.32 s
COVID-ResNet [84]	XR	0.7466	0.7497	0.7747	0.7856	0.7620	0.7702	60.2M	2.31 s
DCSL [85]	XR	0.7486	0.7510	0.7777	0.7870	0.7641	0.7709	125M	2.54 s
COVNet [86]	CT	0.7518	0.7512	0.7805	0.7880	0.7656	0.7812	130.8M	2.18 s
AnoDet [87]	CT	0.7552	0.7518	0.7853	0.7907	0.7682	0.7843	125.8M	2.19 s
DLS [88]	CT	0.7607	0.7535	0.7878	0.7958	0.7703	0.7878	179.1M	2.18 s
DeCovNet [89]	CT	0.7613	0.7573	0.7947	0.8022	0.7755	0.7885	93M	2.12 s
DLQCTM [90]	CT	0.7663	0.7648	0.7989	0.8081	0.7815	0.7896	103M	2.26 s
DenseNet-161[91]	UL	0.7029	0.7229	0.7394	0.7517	0.7311	0.7285	18.8M	2.12 s
ResNet-34[70]	UL	0.7154	0.7288	0.7514	0.7553	0.7400	0.7287	63.6M	0.80 s
VGG-19[66]	UL	0.7214	0.7367	0.7525	0.7605	0.7445	0.7307	144M	2.82 s
ResNet-18[70]	UL	0.7278	0.7395	0.7575	0.7708	0.7484	0.7434	33.3M	0.56 s
VGG-16[66]	UL	0.7365	0.7420	0.7633	0.7777	0.7525	0.7474	140M	2.48 s
DGLM [92]	XRCTUL	0.8070	0.7842	0.8322	0.8591	0.8075	0.8067	130.2M	2.08 s
LM3FT [93]	XRCTUL	0.8110	0.7984	0.8323	0.8676	0.8150	0.8120	126M	2.16 s
MMCL [94]	XRCTUL	0.8120	0.8010	0.8351	0.8682	0.8177	0.8154	129.3M	2.34 s
MEC [95]	XRCTUL	0.8122	0.8079	0.8370	0.8716	0.8222	0.8186	131.7M	2.22 s
GMMF [96]	XRCTUL	0.8127	0.8082	0.8378	0.8741	0.8227	0.8202	130.3M	2.53 s
Ours w/o knowledge	XRCTUL	0.8160	0.8108	0.8378	0.8741	0.8413	0.8234	71.2M	1.74 s
Ours w/o TSAM	XRCTUL	0.8308	0.8121	0.8494	0.8898	0.8491	0.8293	70M	1.12 s
Ours w/o knowledge	XRCTULT	0.8389	0.8293	0.8297	0.8292	0.8292	0.8319	73M	1.77 s
Ours w/o TSAM	XRCTULT	0.8679	0.8688	0.8641	0.8909	0.8664	0.8675	72M	1.14 s
Ours	XRCTUL	0.9371	0.9209	0.8884	0.9805	0.9498	0.9171	75.6M	1.12 s
Ours	XRCTULT	0.9810	0.9889	0.9861	0.9859	0.9875	0.9908	76.4M	1.14 s