FDG PET in the differential diagnosis of degenerative parkinsonian disorders: usefulness of voxel-based analysis in clinical practice

Atypical Parkinsonian disorders (APD) is a group of heterogeneous neurodegenerative diseases combining parkinsonian key features (bradykinesia, rigidity, and rest tremor) to dementia and to other clinical signs and symptoms related to the cortex or cerebellum [1]. Differently from Parkinson’s disease (PD), these disorders are characterized not only by cell loss in the substantia nigra, but also by degeneration in added nervous system areas that normally contain dopamine receptors, known as the striatum, and usually do not respond well to drug treatment as levodopa [2]. The more severe evolution and prognosis require different therapeutic approaches. Therefore, the early differential diagnosis among APD, often complicated by overlapping clinical presentations, becomes essential to set up the correct clinical-therapeutic approach [3]. Brain imaging with positron emission tomography (PET) and [¹⁸F]fluoro-deoxy-glucose (FDG) has been widely used in clinical practice as an important tool for revealing in vivo changes at different stages [4] and could represent a valid aid for the differential diagnosis of APD [5, 6]. In literature, many studies have described and considered frequent disease-related metabolic brain patterns. In PD, functional imaging has shown heterogeneous FDG PET patterns of hypometabolism, ranging from no cortical hypometabolism to diffuse parieto-temporo-occipital hypometabolism, which has been often associated with relative hypermetabolism of the striatal nuclei, globus pallidus, midbrain, thalamus, and cerebellum, as well as at the primary sensory-motor cortex [7, 8]. However, the origin of such pattern would be consistent across studies using global brain activity as a reference for activity scaling [9]. In Lewy Body Dementia (LBD), subcortical structures are relatively spared, as in PD. However, the reduction of cortical metabolic activity is widespread and more marked in occipital and temporo-parietal cortices, while a less relevant involvement is demonstrated in parietal and frontal cortices [10, 11]. On the other hand, in patients with progressive supranuclear palsy (PSP), a widespread cortical hypometabolism has been discovered, in particular in the frontal cortex, mainly in medial and premotor areas, as well as in subcortical structures, like the basal ganglia, especially caudate nucleus, and the midbrain [8, 12]. Concerning to multiple system atrophy (MSA), studies with FDG PET revealed significant hypometabolism in the striatum and in particular in the putamen, mainly in the parkinsonian form, as well as in the midbrain and cerebellum, prevalent in the cerebellar form [8, 12].

Notoriously, cortico-basal syndrome (CBS) is a very heterogeneous syndrome referring to gradual neurodegeneration in several cerebral cortical areas as well as deep structures, such as the basal ganglia. However, the underlying pathologies may be different, including CBD, AD, and others, which result in slightly different clinical phenotypes and in different FDG PET patterns [13]. In this context, FDG PET could also represent a valid tool to support the differential diagnosis within the possible etiologies underlying CBS. The typical pattern of hypometabolism of CBS due to CBD (CBS_CBD) is characterized by an asymmetrical distribution, involving frontal, parietal, temporal cortical areas, basal ganglia and thalamus of the same hemisphere [8, 12, 14, 15].

With the evaluation of such metabolic patterns, FDG PET imaging has obtained an important role as supportive feature in the differential diagnosis of APD subtypes [16‐19]. Moreover, some studies described the increase of diagnostic accuracy through the implementation of semiquantitative analysis, in support of qualitative evaluation [8, 20].

In the context of dementing neurodegenerative diseases, the European Association of Nuclear Medicine (EANM) and the European Academy of Neurology (EAN) have already recommended the use of semi-automated analyses as a valid helpful tool to improve the visual reading of FDG PET images [21].

However, both the increased utilization of FDG PET in APD disorders and the pressure for cost-effectiveness request a systematic evaluation and a validation of its utility in clinical practice. An optimized statistical parametric mapping (SPM) routine for FDG PET data analysis in single subjects has already been validated for the differential diagnosis of dementia [22]. Some authors have also evaluated its possible supportive role in a cohort of patients with APD, observing a significant impact on the clinical work-up and prognosis, particularly at early stages when the presentation is uncertain [1, 23]. However, to our knowledge, in literature, it has never been evaluated the impact of the use of FDG PET semi-automated analyses in a clinical setting, neither the use of both hypo- and hypermetabolism maps, as a way to detect both regions of decreased activity and of preserved neuronal activity, which could be especially useful in parkinsonian disorders. Therefore, the purpose of our study is to evaluate the contribution, in terms of increasing accuracy and increasing diagnostic confidence, of voxel-based FDG PET analyses in the differential diagnosis of neurodegenerative parkinsonian disorders, including PD, MSA, PSP, and CBS_CBD. In particular, it will be evaluated the increasing accuracy and confidence of different nuclear medicine physicians in detecting the specific metabolic patterns after visual reading of FDG PET images, with the aid of statistical maps of hypometabolism and also of relative hypermetabolism.

Our study takes part in the well-known debate about the utility of metabolic biomarkers in the diagnostic work-up of APD, in order to obtain benefits for treatment options and support to the patients [1, 23]. In particular, we aim to underline the clinical practice role of FDG PET/CT imaging in the differential diagnosis among neurodegenerative parkinsonian disorders. Our specific focus was to evaluate the additional value, over qualitative PET images assessment, of not only maps of hypometabolism, as a mean to evaluate regions of neuronal dysfunction and loss, but also maps of relative hypermetabolism, as a way to evaluate cerebral regions of preserved synaptic activity.

Several studies in the literature have demonstrated the significant value of voxel-based analyses as an aid in brain FDG PET assessment in patients with cognitive decline [16‐19, 21, 22]. Similar results are also expected in the field of APD, although there are no studies that have ever formalized this result. First of all, the results of our study show that the diagnostic accuracy values are higher in expert evaluators than in not-experts, especially regarding the qualitative assessment of PET images. This result was obviously expected and it emphasizes the complexity of brain FDG PET evaluation, which requires considerable work experience. For expert evaluators, overall high accuracy just after qualitative assessment increases with the use of hypometabolism maps and concurrent increase also in confidence levels and reaches stability after the use of hypermetabolism maps. It suggests that they could not be completely useless in this group of evaluators. For not-expert readers, overall accuracy after qualitative assessment is suboptimal and the use of both hypometabolism and hypermetabolism maps let them arrive at levels of accuracy similar to expert readers, consequently increasing their levels of confidence. Regarding the different diseases considered in this study, we could express the following considerations.

In the cases of PD disease, experts reach the highest accuracy, suggesting that this pattern is well recognizable even if very heterogeneous actually including both forms of PD patients, either without cognitive deficits or with mild cognitive impairment, ranging from no cortical hypometabolism to some areas of posterior hypometabolism. Accuracy levels still improve after the use of hypometabolism maps but do not change after the use of hypermetabolism maps, suggesting that the typical relative hypermetabolism in striata is well recognizable by the experts. Instead, not experts’ accuracy is quite low after qualitative assessment, suggesting that the widespread hypometabolism is difficult to recognize by them. However, with the use of both hypometabolism and hypermetabolism maps, accuracy increases of around 10% each time, reaching good levels, suggesting that voxel-based assessment is fundamental for not-expert readers. However, they did not reach the levels of accuracy of experts. Regarding confidence, it is quite high and in both groups increases at each evaluation step. In the cases of MSA disease, accuracy after the qualitative assessment is one of the lowest for the experts; however, it reaches 100% after the use of hypometabolism maps. This suggests that hypometabolism maps are really relevant even in experts’ readers to identify the MSA patterns. In not experts, it happens almost the same thing. After hypometabolism maps, they reached the highest accuracy among diseases, and it did not improve after hypermetabolism maps. In both groups, however, the use of hypermetabolism maps increases confidence, maybe because of the confirmation that there is a metabolic sparing of cortical regions [24, 25]. In the cases of PSP disease, one of the lowest accuracy levels after qualitative assessment happens in both experts and not experts, improving after hypometabolism maps in both groups and after hypermetabolism maps only in not experts. This could suggest the difficulty in evaluating this pathology, which could present with different patterns in the different variants [26]. This lower accuracy level as compared to the other pathologies is reflected by the confidence, which is the lowest among pathologies. In the cases of CBS disease, diagnostic accuracy values are rather low in both expert and not-expert readers. This is in line with the literature since the typical CBS-specific metabolic pattern is actually present only in the subgroup of patients with cortico-basal degeneration (CBD) pathology, while others could show metabolic patterns mimicking other diseases [13, 27]. It also seems that when the specific pattern is present (cortical and subcortical hypometabolism with frank asymmetry), it is well recognizable even only with qualitative assessment, and even by not-expert evaluators, since the percentages of correctly classified subjects did not improve after the use of any statistical maps. For expert evaluators, it can be noted an increase in accuracy only with the help of the hypometabolism maps, while it remains at the same level after the evaluation of the relative hypermetabolism maps. This result suggests that expert readers probably are able to evaluate which regions are spared by disease already from the qualitative assessment, and do not need further analysis. Moreover, the accuracy and the confidence values increase significantly with the help of the hypometabolism maps, so we derived that experts find valuable help in this type of analysis. On the other hand, even if relative hypermetabolism maps did not increase accuracy values for experts, diagnostic confidence seems to grow up. It can be deduced that this type of analysis is not completely useless, but contributes to increase the certainty with which the diagnosis is proposed. With regard to not-expert evaluators, the use of both hypo- and hypermetabolism maps proves to be very useful. In fact these evaluators, starting from rather low levels of accuracy at qualitative assessment alone, reaches much better levels, not equal to those of the experts but however very close to them and clinically acceptable. It is inferred that the not-expert evaluators need to make use of further analysis to better understand where both the disease-affected and disease-spared regions are located. It is also interesting to notice that the diagnostic confidence in such evaluations increases significantly as accuracy increases.

Early diagnosis of neurodegenerative parkinsonian disorders is a crucial point of interest because both prognosis in terms of disease duration and clinical progression, and treatment options can vary considerably among the various subtypes. In this study, we demonstrated the additional value of combining voxel-based analyses with a qualitative assessment of brain PET images. It is shown that metabolic maps are very useful in supporting this type of evaluation, especially for less experienced nuclear physicians, but also for experienced ones. In not-expert evaluators, the support results in a significant increase in diagnostic accuracy as well as clinical confidence. In expert evaluators, the increase in accuracy is supported by hypometabolism maps alone, and there is still an increase in diagnostic confidence. Finally, in this particular group of diseases, it was shown that even maps of relative hypermetabolism can make their contribution to clinical practice, particularly for less experienced evaluators.