Neuroimaging studies further support the importance of structural changes in limbic and cortical areas in the progression of cognitive impairment in PD. Grey matter loss in the limbic/paralimbic system (amygdala, anterior cingulate cortex, hippocampus and parahippocampal cortex) clearly characterizes dementia in PD [10], [11], and some studies have also found grey matter loss in occipito-parietal areas in PDD [12], [13]. Using voxel-based morphometry (VBM), prior studies analyzing the structural changes present in earlier stages of cognitive deterioration have begun to disclose a pattern of cortical involvement in the temporal, parietal and occipital cortex in PD-MCI patients [14], [15], [16], [17]. Further knowledge of the structural changes of PD-MCI patients would aid in the early detection of patients at risk of developing dementia. Recent advances in image processing and acquisition has allowed the automatic extraction of cortical thickness from MRI [18]. In non-demented PD patients, cortical thickness seems more sensitive than VBM to identify regional grey matter changes [19] and, compared to VBM, provides the most plausible measure of change over time [20].

To our knowledge, this is the first study to analyze changes in cortical thickness in PD patients with normal cognition, mild cognitive impairment and dementia. The changes observed in our sample are very similar to those described in three recently published papers that utilized MRI and voxel-based morphometry searching for changes in grey matter volume across these stages of cognitive decline [16], [17], [38]. In the Weintraub et al. study [16], decreased volumes in the hippocampus and parietal–temporal cortex were observed in PD-MCI patients, and is more evident in the PDD group. Melzer et al. [17] showed also a linear progression in grey matter volume loss across cognitive stages, with initial atrophy of the temporal, parietal, frontal and caudal hippocampus in PD-MCI, that became more extensive in PDD patients, and additional volume decreases in the parahippocampus, lingual gyrus, and posterior cingulate gyrus. Interestingly, cortical changes correlated with the degree of global cognitive dysfunction, but did not correlate with motor impairment. Our findings of a significant relationship between delayed verbal memory and cortical thinning in the bilateral anteromedial temporal cortex support previous studies in which memory impairments in PD related to hippocampal volume loss [11], [16].

In these studies, however, no cortical differences were found between HC and PD with normal cognition. Song et al. found loss of gray matter in PD-NC patients in the left occipital area, extending to the bilateral temporal, left prefrontal, and right occipital areas in PD-MCI patients [15]. Yet, the authors did not report correlations of cortical changes with neuropsychology.

In our sample, we observed very early cortical thinning in the PD-NC group in the middle temporal gyrus bilaterally, left superior frontal gyrus, right isthmus (posterior) cingulate cortex, right inferior parietal lobule, and lateral occipital cortex. These findings may be explained by the greater sensitivity of imaging/postprocessing techniques [19], [20], and seem to indicate that some structural changes can be detected by cortical thickness well before the cortical atrophy associated with cognitive impairment. In fact, these changes did not correlate with any neuropsychological task, and PD-NC patients did not differ significantly from the HC group in any cognitive item. In the PD-MCI and PDD groups, correlations found between neuropsychology and decreased cortical thickness indicate the importance of cortical thinning of temporal and parietal regions in the performance of attentional, mnesic and visuoperceptive tasks. In particular, sustained attention and alternating verbal fluency, both of which have been consistently reported as two of the earlier tasks to become impaired in PD [7], [39], correlated with thinning in the cuneus, bSTS, fusiform gyrus, and lingual gyrus in both PD-MCI and PDD patients. Immediate free-recall memory, that relies more on attentional resources than in storage capacities, also correlated to posterior thinnig (cuneus) in PD-MCI, while in PDD delayed free-recall memory correlated with thinning in the right anteromedial temporal cortex, right parahippocampus, and left lingual gyrus. Regarding the two posterior cortical tasks of the PD-CRS, confrontation naming correlated with thinning in the fusiform gyrus in PD-MCI, and with thinning in the fusiform gyrus, temporal pole, and parahippocampus in PDD patients. Clock copying correlated with decreased cortical thickness in the precuneus, parahippocampus, and lingual gyrus in PDD.

The discrete changes that we have found in neocortical regions may suggest that structural changes in PD play only an additive dysfunctional effect on cortical areas already impaired by metabolic defects and presynaptic dysfunction [40]. However, the relatively limited atrophic changes observed are centered in regions representing important nodes of information integration [37], [41], therefore apparently minor structural changes may entail important functional impairments in distant but functionally related areas. The medial temporal lobe (including the entorhinal cortex and anteriomedial temporal cortex) provides information from prior experiences in the form of memories and associations between stimuli from different sensory modalities [42]. In both Alzheimer’s disease (AD) and dementia with Lewy bodies, concurrent atrophy of the medial temporal lobe, precuneus, and temporo-parietal cortices appears to play a major role in the development of dementia [43]. Cortical thickness analysis of normal brains has revealed the existence of a set of posterior medial cortical regions that form a densely interconnected network [44]. Key components of this network are the cuneus, precuneus, posterior aspects (isthmus) of the cingulate cortex, inferior parietal lobule, and the bSTS. These heteromodal cortical areas have a central role in integrating information across functionally segregated brain regions [45], and seem particularly vulnerable to atrophy in different neurodegenerative diseases [43], [45]. Noteworthy, these posterior cortical regions are densely interconnected with the medial temporal lobe [46]. In our sample, the areas with the more robust projections to the entorhinal cortex, namely, the parahippocampus, lingual gyrus, and bSTS, were precisely those to become progressively more atrophic in the transition from normal cognition to dementia. These regions respond to multiple sensory stimuli and provide multimodal sensory information to the hippocampal formation [47], [48].

There is no clear data on the potential underlying mechanisms of cortical thinning as detected by Freesurfer on MRI. However, literature based on histology suggests that cortical thinning is unlikely to originate predominantly from neuronal death. Rather, cellular shrinkage and reduction in dendritic arborization are more likely to account for cortical thinning. [49].

The increased cortical thickness observed in the caudal (PD-MCI vs PD-NC) and rostral (PDD vs. PD-MCI) regions of the anterior cingulate cortex (ACC), the correlation between thicker caudal ACC and impaired sustained attention, and the posterior thinning of the caudal ACC in the PDD group, are in line with previous findings of progressive thickness changes in neurodegenerative diseases analyzed over time. For instance, in the preclinical and early phases of AD the evolution of cortical thickness appears to present an inverted-U shape model, with initial increased cortical thickness in the ACC and temporoparietal areas, and posterior thinning of these same structures. [50] In healthy ApoE-4 carriers, increased thickness has been associated with impaired selective attention [51] and, interestingly, neuropathological studies of healthy elderly subjects compared to asymptomatic AD subjects and patients with AD-MCI have shown the existence of a phase of hypertrophy of the neuronal cell bodies in preclinical AD that precedes the atrophy of these areas in symptomatic patients. [52] Similarly, patients with early Huntington’s disease present specific increased thickness in the ACC that progresses to neuronal atrophy in more advanced stages. [53].

We acknowledge some limitations of our study. First, lack of significance at the corrected level relies mostly on a limited sample size. All p-values reported in relation to cortical thinning in this sample are uncorrected, and therefore the results of this study need to be replicated in future studies. Second, there are some technical limitations in cortical thickness determinations on MRI due to artefacts, such as susceptibility or motion, which can make the brain surfaces difficult to be defined. In this regard, the method that we have implemented with Freesurfer, which uses an automatic approach, and a manual correction afterwards, seems appropiate to minimize this limitation. Thirdly, we used a hypothesis-driven approach to search for correlations between cortical thinning and neuropsychological performance. This approach is based on the changes observed between cognitive groups, looking for those cognitive domains that characterize better patients with PD-MCI or PDD. Since causality cannot be inferred by correlational findings alone, and larger samples are needed to look for correlations that survive multiple comparisons correction, future studies are warranted to verify our results. Finally, the presence of hallucinations in the PDD group may confound the results of the study. Imaging studies have associated the presence of visual hallucinations with grey matter volume reductions in the orbitofrontal cortex, parahippocampal area, lingual gyrus, and superior parietal lobe. [54], [55] Studies analyzing PDD patients with and without hallucinations are then needed to disentangle the differential involvement of these areas on cognitive deterioration and hallucinatory phenomena.