SUFI: an automated approach to spectral unmixing of fluorescent multiplex images captured in mouse and post-mortem human brain tissues

Table 3 Unmixing results using a single image of smFISH data from post-mortem human DLPFC brain tissue sections

Method	DAPI	Opal 520	Opal 570	Opal 620	Opal 690	Lipofuscin	Mean
Root mean squared error (RMSE)
FCLSU	0.0026	0.0045	0.0073	0.0023	0.0046	0.0011	0.0037
ELMM	0.0020	0.0043	0.0066	0.0026	0.0040	0.0011	0.0034
GELMM	0.0022	0.0041	0.0070	0.0028	0.0047	0.0012	0.0037
SØrensen–Dice similarity coefficient
FCLSU	0.9869	0.8398	0.7833	0.9266	0.5987	0.5431	0.7797
ELMM	0.9884	0.8453	0.7882	0.9364	0.7422	0.5008	0.8002
GELMM	0.9556	0.8534	0.7889	0.9037	0.6435	0.4119	0.7595
Structural similarity index (SSIM)
FCLSU	0.9695	0.9931	0.9857	0.9953	0.9328	0.9972	0.9789
ELMM	0.9786	0.9950	0.9899	0.9946	0.9460	0.9974	0.9836
GELMM	0.9695	0.9952	0.9872	0.9950	0.9337	0.9962	0.9795

Here we compare the performance of SUFI against the ZEN unmixed image using three metrics (RMSE, dice similarity, SSIM) for all three methods (FCLSU, ELMM, GELMM). For the three metrics, values range between 0 and 1 with better performance indicated by lower RMSE values and higher SSIM and dice similarity values. Mean over all channels is presented as last column

ISSN: 1471-2202