Gamma-band rhythmic abnormalities have been of significant interests in autism spectrum disorders (ASD). Most studies used magnetoencephalography (MEG) due to its advantage in measuring weak gamma signals as compared to electroencephalography (EEG). However, EEG is more accessible, portable, and importantly, more sensitive to cortical sources located at the crowns of gyri, than MEG. Therefore, it is extremely valuable if EEG can be used to detect gamma-band abnormalities in ASD, which could provide complementary insights on pathology of ASD. One challenge in detecting gamma-band neural activities is to remove muscular artifacts, which share the same frequency band. In the present study, we used a previously developed time-frequency independent component analysis (ICA)approach to probe EEG gamma-band abnormalities in ASD. We examined functional connectivity (FC) patterns on intrinsic connectivity networks (ICNs), i.e., the ICs representing distributed neural activities obtained from ICA, using the metrics of spectral power of individual ICNs and coherence between different ICNs. Seven ICNs that reassembled ICNs obtained from EEG data in the band of 2-30 Hz, were successfully identified in the gamma-band (31-50 Hz) data by the approach. Local over-connectivity in the bilateral frontal and left parietal ICNs, as well as long-range under-connectivity between left and right motor ICNs, were observed in ASD. In addition, the age-related effect was identified in the left motor and left parietal ICNs in healthy control, but not in ASD. These findings demonstrated a mixed pattern of gamma-band FC changes in ASD. It further indicated that the developed approach is promising in reconstructing gamma-band patterns from resting-state EEG signals.