Table 3.

Quantitative results of various model variants evaluated on STARmap and MERFISH datasets, focusing on performance metrics including PCC, SPCC, SSIM, RMSE, and COSSIM. These metrics provide a comprehensive assessment of the models’ ability to accurately reconstruct spatial gene expression patterns.

Variants	STARmap					MERFISH
	PCC	SPCC	SSIM	RMSE	COSSIM	PCC	SPCC	SSIM	RMSE	COSSIM
Module-Variant-1	0.2099	0.1845	0.0842	1.2722	0.4015	0.3125	0.2514	0.2015	1.1615	0.4215
Module-Variant-2	0.2155	0.1884	0.0812	1.2451	0.4154	0.3348	0.2678	0.2248	1.1443	0.4412
Loss-Variant-1	0.1654	0.1654	0.0611	1.2978	0.3889	0.2484	0.1814	0.1872	1.2311	0.3548
Loss-Variant-2	0.1933	0.1789	0.0074	1.2845	0.3945	0.2778	0.2019	0.0187	1.2102	0.3784
SELF-Former	0.2359	0.2330	0.1016	1.2327	0.4387	0.3566	0.2780	0.2493	1.1270	0.4692