Cognitive-behavioral therapy (CBT) is thought to be useful for chronic pain, with the pathology of the latter being closely associated with cognitive-emotional components. However, there are few resting-state functional magnetic resonance imaging (R-fMRI) studies. We used the independent component analysis method to examine neural changes after CBT and to assess whether brain regions predict treatment response. We performed R-fMRI on a group of 29 chronic pain (somatoform pain disorder) patients and 30 age-matched healthy controls (T1). Patients were enrolled in a weekly 12-session group CBT (T2). We assessed selected regions of interest that exhibited differences in intrinsic connectivity network (ICN) connectivity strength between the patients and controls at T1, and compared T1 and T2. We also examined the correlations between treatment effects and rs-fMRI data. Abnormal ICN connectivity of the orbitofrontal cortex (OFC) and inferior parietal lobule within the dorsal attention network (DAN) and of the paracentral lobule within the sensorimotor network in patients with chronic pain normalized after CBT. Higher ICN connectivity strength in the OFC indicated greater improvements in pain intensity. Furthermore, ICN connectivity strength in the dorsal posterior cingulate cortex (PCC) within the DAN at T1 was negatively correlated with CBT-related clinical improvements. We conclude that the OFC is crucial for CBT-related improvement of pain intensity, and that the dorsal PCC activation at pretreatment also plays an important role in improvement of clinical symptoms via CBT.