Fig. 1.

DCBFC Sketch. Step i: (a) Raw time-series fluorescent calcium images are acquired and processed with a 0.1- to 4-Hz bandpass. Then, (b) the preprocessed images are obtained by global signal regression. Step ii: (c) The similarity matrix using Pearson’s correlation between the time series is calculated. Then, a dataset $\{D\}=\{1,\dots ,N\}$ is initialized. Step iii: The composite index $\gamma$ values of all pixel are calculated, and (d) these values are sorted in descending order. The red-dotted line corresponds to the threshold ${\gamma }_{\mathrm{thre}}$. The pixels whose $\gamma$ values are greater than ${\gamma }_{\mathrm{thre}}$ are chosen as candidate central pixels. Step iv: The central pixels (e, black stars) of the current loop are screened out. Step v: Those pixels that are similar to the central pixels are removed from $\{D\}$, and the similarity matrix of the remaining pixels is used for the next loop. Steps iii–v: are repeated until no central pixels are available to select. Step vi: (f) All central pixels are obtained. Step vii: (g) All the other pixels in the image are assembled to the clusters corresponding to their central pixels. Step viii: The seed pixel functional connection maps corresponding to all pixels in the same cluster are averaged, and (h) the RSFC patterns of all clusters are obtained.