Case 3: TF featured and the K-means clustering along with the proposed hard labeling method.
Figure 9 shows the ROC plot of the proposed TF feature extraction and discriminant cluster selection using hard labeling.
The former is performed using the hard labeling clustering method, and the latter three are evaluated employing the fuzzy labeling approach.
From the entire clusters, 25% were assigned as common clusters and the remaining clusters labeled class normal or abnormal as explained in hard labeling scheme.
Two methods are proposed to define the class label of each signal: hard labeling which is based on majority vote, and fuzzy labeling which is based on majority vote weighted by the membership distribution of each cluster.
In hard labeling, clusters with more than 30% Class 1 feature vectors are labeled as Class 1 (i.e., clusters 3, 6, 9, 10, 11, 15, 16, and 17) and the ones with more than 30% Class 2 feature vectors are labeled as Class 2 (i.e., clusters 2, 4, 5, 7, 8, 14, 19, 21, and 22).
Instead we tested two postprocessing methods, an MRF based smoother that uses "soft" (probabilistic) patch labels such as those provided by our LRC/LRC method and a local oversegmentation algorithm that uses "hard" labels such as those provided by our LRC/CRF algorithm (whose usual output is a crisp FastPD segmentation, not soft patchwise marginals).
In hard labeling, discriminant clusters were assigned to one of the possible classes, but in fuzzy labeling, they were associated to each class with a relative membership value ranging from 0 to 1 (with 0 being the least contribution, and 1 being the most).
To determine whether these regions demonstrated any differential response at the time of label onset, we subjected the parameter estimates from each ROI to a t-test comparing activation during the onset of the "EASY" labels to activation at the onset of the "HARD" labels.
