Low-frequency Functional Connectivity in fMRI - Biomedical Signal Analysis: Contemporary Methods and Applications - page 270

tance matrix where the rows and columns appear in an ordered fashion.
The emerging block-diagonal structure reflects the characteristic of the
TMP algorithm to cluster PTCs based on their mutual dependency (i.e.
their pairwise distance).
For each resting-state fMRI data set, the position of the motor cortex
is determined based on the segmentation provided by the pixel-specific
stimulus-correlation map obtained in the motor task fMRI experiment
of the same subject. That is, a PTC whose correlation coecient in
the motor stimulation experiment is above a defined threshold of Δ
(e.g., Δ = 0 . 6) is considered as belonging to the motor cortex. This
segmentation approach is referred to as the cc-cluster method.
For the clustering methods, the segmentation of the motor cortex is
obtained by merging single clusters. The identification of such clusters
is determined by the similarity index ( SI ) [300]. The SI index is defined
as

SI =2 |

A 1 ∩ A 2 |

(9.1)

| A 1 | + | A 2 |

and gives a measure of the agreement of the two binary segmentations
A 1 and A 2 . It is defined as the ratio of twice the common area to the sum
of the individual areas. An excellent agreement is given for SI > 0 . 7,
according to [300]. Although the absolute value of SI is dicult to
interpret, it gives a quantitative comparison between measurement pairs.
The cluster identification works as follows. First, the cluster showing
the largest SI value with the reference segmentation is selected. Then
this cluster is combined with the remaining cluster, if the SI value of
the two merged clusters is increased. This procedure continues until no
increase in the SI value is observed.
Figure 9.5 shows a comparison between the segmentation results
obtained by the unsupervised clustering methods for subject #1 in
the resting-state. By taking the average value of all PTCs belonging
to a certain determined segmentation, a representative PTC for each
segmentation is obtained. The figure shows that both the topographic
mapping of proximity data and the classical clustering techniques are
able to detect low-frequency connectivity associated with the motor
cortex.
The resulting values for the SI index for the proposed methods