In 2000, Gorman et al. published a widely acknowledged revised version of their 1989 neuroanatomical hypothesis of panic disorder (PD). Herein, a 'fear network' was suggested to mediate fear- and anxiety-related responses: panic attacks result from a dysfunctional coordination of 'upstream' (cortical) and 'downstream' (brainstem) sensory information leading to heightened amygdala activity with subsequent behavioral, autonomic and neuroendocrine activation. Given the emergence of novel imaging methods such as fMRI and the publication of numerous neuroimaging studies regarding PD since 2000, a comprehensive literature search was performed regarding structural (CT, MRI), metabolic (PET, SPECT, MRS) and functional (fMRI, NIRS, EEG) studies on PD, which will be reviewed and critically discussed in relation to the neuroanatomical hypothesis of PD. Recent findings support structural and functional alterations in limbic and cortical structures in PD. Novel insights regarding structural volume increase or reduction, hyper- or hypoactivity, laterality and task-specificity of neural activation patterns emerged. The assumption of a generally hyperactive amygdala in PD seems to apply more to state than trait characteristics of PD, and involvement of further areas in the fear circuit, such as anterior cingulate and insula, is suggested. Furthermore, genetic risk variants have been proposed to partly drive fear network activity. Thus, the present state of knowledge generally supports limbic and cortical prefrontal involvement as originally proposed in the neuroanatomical hypothesis. Some modifications might be suggested regarding a potential extension of the fear circuit, genetic factors shaping neural network activity and neuroanatomically informed clinical subtypes of PD potentially guiding future treatment decisions.