Comparative study on similarity metrics for seed-based analysis of functional connectivity photoacoustic tomography images

Seed-based correlation analysis is one of the most popular methods to explore the functional connectivity in the brain. Based on the time series of a seed, i.e., small regions of interest, connectivity is computed as the correlation of time series for all other pixels in the brain. Similarity metric to measure the similarity between time courses of different seeds plays an important role in the detection of functional connectivity maps. In this study, we investigate the performance of six similarity metrics including Pearson correlation, Kendall, Spearman, Goodman-Kruskal Gamma, normalized cross correlation and coherence analysis to determine their performance for the functional connectivity photoacoustic tomography (fcPAT) signals/images. The methods are implemented and applied on the fcPAT data of a mouse brain. We also add noise to the fcPAT data and explore the noise tolerance of these metrics.