Radiological biomarkers for diagnosis in PSP: Where are we and where do we need to be?

PSP is a pathologically defined neurodegenerative tauopathy with a variety of clinical presentations including typical Richardson's syndrome and other variant PSP syndromes. A large body of neuroimaging research has been conducted over the past two decades, with many studies proposing different structural MRI and molecular PET/SPECT biomarkers for PSP. These include measures of brainstem, cortical and striatal atrophy, diffusion weighted and diffusion tensor imaging abnormalities, [18F] fluorodeoxyglucose PET hypometabolism, reductions in striatal dopamine imaging and, most recently, PET imaging with ligands that bind to tau. Our aim was to critically evaluate the degree to which structural and molecular neuroimaging metrics fulfill criteria for diagnostic biomarkers of PSP. We queried the PubMed, Cochrane, Medline, and PSYCInfo databases for original research articles published in English over the past 20 years using postmortem diagnosis or the NINDS‐SPSP criteria as the diagnostic standard from 1996 to 2016. We define a five‐level theoretical construct for the utility of neuroimaging biomarkers in PSP, with level 1 representing group‐level findings, level 2 representing biomarkers with demonstrable individual‐level diagnostic utility, level 3 representing biomarkers for early disease, level 4 representing surrogate biomarkers of PSP pathology, and level 5 representing definitive PSP biomarkers of PSP pathology. We discuss the degree to which each of the currently available biomarkers fit into this theoretical construct, consider the role of biomarkers in the diagnosis of Richardson's syndrome, variant PSP syndromes and autopsy confirmed PSP, and emphasize current shortfalls in the field. © 2017 International Parkinson and Movement Disorder Society.

senting biomarkers for early disease, level 4 representing surrogate biomarkers of PSP pathology, and level 5 representing definitive PSP biomarkers of PSP pathology. We discuss the degree to which each of the currently available biomarkers fit into this theoretical construct, consider the role of biomarkers in the diagnosis of Richardson's syndrome, variant PSP syndromes and autopsy confirmed PSP, and emphasize current shortfalls in the field. V C 2017 International Parkinson and Movement Disorder Society.
K e y W ord s: progressive supranuclear palsy; diagnosis; magnetic resonance imaging; positron emission tomography; single-photon emission computed tomography Progressive supranuclear palsy (PSP) is a pathologic diagnosis with neurodegeneration characterized by abnormal tau pathology in the form of globose neurofibrillary tangles, tufted astrocytes, coiled bodies, and threads, 1 with a predominance of 4-repeat (4R) tau isoforms. 2 Tau pathology is typically observed in the brain stem, basal ganglia, diencephalon, and temporal, motor, and premotor cortices, 1,2 although distribution can vary. 3,4 The most commonly recognized clinical presentation of PSP is Richardson's syndrome (PSP-RS), in which patients have early and notable gait and postural instability, frequent falls, and abnormal vertical eye movements (supranuclear gaze palsy). 5,6 However, a number of other clinical presentations of PSP have been increasingly recognized, including but not limited to PSP with predominant parkinsonism (PSP-P), 6 PSP with progressive gait freezing (PSP-PGF), 7 PSP with predominant frontal presentation (PSP-F), 8 PSP with a predominant speech/language disorder (PSP-SL), 9 and PSP with predominant corticobasal syndrome (PSP-CBS). 10 We have recently developed the Movement Disorder Society-endorsed PSP clinical diagnostic criteria that recognize this heterogeneity and provide criteria for the different clinical variants of PSP. 11 A major challenge faced during the revision of the diagnostic criteria was to determine whether there was enough evidence to support the inclusion of neuroimaging biomarkers in the diagnosis of PSP-RS, the other variant syndromes of PSP (vPSP), or in the diagnosis of pathological PSP, and what role they should play in the diagnostic criteria. Table 1 provides a theoretical construct to judge the utility of diagnostic neuroimaging biomarkers in PSP. The first step is to demonstrate abnormalities in the group of interest compared with matched healthy controls and other clinically overlapping disease groups (level 1). In the context of PSP, this typically means demonstrating abnormalities in PSP-RS compared with other parkinsonian disorders, such as Parkinson's disease (PD), multiple system atrophy with predominant parkinsonism (MSA-P), and CBS. However, if one wishes to ultimately develop a diagnostic biomarker for PSP pathology, it is also important not to ignore vPSP, for which neuroimaging signatures may differ from PSP-RS. A biomarker differentiating PSP-F, PSP-SL, and PSP-CBS from other frontotemporal lobar degeneration spectrum disorders may also be valuable. For these group-level findings to translate into useful biomarkers, the next step is to demonstrate useful sensitivity and specificity (>80%) for the clinical diagnosis at the individual patient level (level 2). Biomarkers that perform well at this level could be valuable to support the clinical diagnosis. However, because these analyses are based on comparison with clinical diagnosis rather than the gold standard of neuropathology, there is still no evidence at this point that the biomarker adds anything to clinical diagnosis, other than to increase confidence. A biomarker could surpass clinical diagnosis if one can demonstrate utility for early clinical diagnosis, when patients have mild or nonspecific symptoms and signs before they meet clinical criteria for the disease (level 3), or if one can demonstrate that a biomarker has a strong relationship with the presence of PSP pathology regardless of clinical phenotype (level 4). The latter will ideally require the demonstration that a biomarker is highly associated with PSP pathology, not only in patients diagnosed with PSP-RS but also in vPSP, thus representing utility for the entire clinical spectrum of PSP. However, neuroimaging biomarkers that satisfy level 4 may still be considered only a surrogate marker of pathology, meaning that they correlate well with pathology but do not directly measure pathology. Thus, the holy grail in neuroimaging is to identify a biomarker that directly measures underlying pathology and hence could be considered a definitive pathological biomarker (level 5). We are getting closer to this goal with the development of PET ligands that can bind to abnormal tau in the brain, and current knowledge of these biomarkers will be discussed. At levels 4 and 5, the ideal biomarker would be one that is specific to PSP pathology, although biomarkers that could identify a 4R tauopathy could also be diagnostically useful. Another issue to consider when assessing the value of neuroimaging biomarkers is how well the proposed measures would translate into clinical practice; ideally they should be relatively inexpensive, convenient, safe, widely available, and comparable across different centers.
This review will utilize the theoretical construct outlined in Table 1 to evaluate the degree to which different proposed structural and functional neuroimaging metrics fulfill criteria as diagnostic biomarkers in PSP. As part of our efforts to develop the new diagnostic criteria, a detailed literature search and content evaluation was performed that formed the basis of this review (Supplemental Data).

Structural MRI Brain Stem Measures
Striking midbrain atrophy is typically observed in PSP-RS, and a number of midbrain metrics have been proposed as potential biomarkers. These metrics include visual assessment of midbrain atrophy, midbrain profile or the presence of specific morphological markers such as the "hummingbird" sign (atrophy of dorsal midbrain resembles hummingbird's head and bill in midsagittal plane 12 ), "Mickey Mouse" sign (rounded rather than rectangular midbrain peduncles in axial planes), 13 and "morning glory" sign (concavity of the lateral margin of the midbrain tegmentum in axial planes 14 ); see Figure 1. Quantitative measures of midbrain anterior-posterior diameter and midsagittal area or volume have also been assessed. Studies are in general agreement that midbrain measurements are smaller in PSP-RS compared with MSA and PD, 14-31 although overlap can occur at the individual level, particularly between PSP-RS and MSA. 15,16,28,30 Diagnostic sensitivity and specificity values ( Table 2) are typically high (>90%) for differentiating PSP-RS from controls and from MSA and PD using midbrain area, 15-17,32 although midbrain diameter 15,18-20,22,23,32 and volume 15,24 and visual assessments 13,14,20,21,25,33,34 have been more variable, not always meeting the 80% cut point required for a level 2 biomarker. Visual assessments of the midbrain can be particularly problematic because they are not quantitative, lack objectivity, and can be highly dependent on image acquisition and patient positioning. 35,36 A ratio of midbrain to pons area ( Fig. 1) in the midsagittal plane has been proposed as a biomarker to differentiate PSP-RS from MSA-P, given that MSA-P is associated with atrophy of the pons and sparing of the midbrain, the opposite pattern to PSP-RS. 15 Some studies have found high sensitivity and specificity for the midbrain-pons area ratio in differentiating PSP-RS from MSA-P and from PD, 15 Table 2. A number of studies have found that the MRPI was superior or equivalent to the midbrain-pons ratio in differentiating PSP-RS from MSA-P and PD 17,37,38,40,41 (Table 2). Fewer data are available to assess how well midbrain measures could differentiate PSP-RS from CBS. 13,45 Therefore, there is plenty of evidence to support brain stem measurements as level 2 diagnostic biomarkers in PSP-RS (Table 3). However, proposing one specific measure for the purposes of diagnostic criteria is challenging because centers differ in how they perform these measurements, and specific cut points vary and will likely be cohort-and acquisition-specific. The MRPI appears to be less affected by aging compared with the midbrain-pons ratio 46 but requires detailed measurement of a number of structures that may be difficult to standardize. Indeed, 1 multicenter study found that the MRPI did not perform as well as the midbrain-to-pons ratio in differentiating PSP-RS from PD and MSA-P. 32 However, another multicenter study showed high sensitivity/specificity for the MRPI in differentiating PSP-RS and PD and showed that an automated MRPI measurement that does not rely on rater reliability performs as well as a manual MRPI measurement. 43 Automated methods for measuring midbrain volume are also now available 47 and may improve standardization. There is evidence that these biomarkers could reach level 3 and show diagnostic value in early PSP-RS (Table 3). Abnormal MRPI and midbrain-pons ratios have been shown to predate and predict the development of PSP-RS in patients with clinically unclassifiable parkinsonism at baseline in a retrospective 23 and prospective study, 48 with abnormalities detected 15 months before patients fulfill criteria for PSP-RS in the retrospective study. 23 Given that the clinical diagnosis of PSP-RS has high sensitivity and specificity for pathological PSP, 13,49,50 the midbrain-based measures discussed above also tend to perform well in autopsy-confirmed studies. 19,29 However, it is less clear whether these measures add anything to the clinical diagnosis of PSP-RS in predicting pathology and hence could be level 4 biomarkers. 13 Group-level studies have failed to find midbrain atrophy in patients with PSP pathology who presented with clinical syndromes other than PSP-RS, 51 including patients presenting with CBS. 52 Conversely, reduced midbrain areas were identified in PSP-RS that had underlying corticobasal degeneration pathology. 51 It therefore appears that in many instances midbrain atrophy is related to the PSP-RS clinical was only present in 68.4%, even after a disease duration of 4.8 years. 13 However, midbrain atrophy has been observed in speech and language disorders that are confirmed or suspected of having PSP pathology, 53-56 as well as in PSP-F 8 and PSP-P. 39,57-59 Midbrain atrophy in vPSP is typically less severe than in PSP-RS, 39,56-58 although there is some suggestion that abnormalities on the MRPI could be an early feature in PSP-P 59 and have some value as a level 3 biomarker.

Cortical Measures
A number of group-level studies have demonstrated cortical atrophy in PSP-RS, typically involving the frontal lobes. 33,60-74 The focus of atrophy appears to be the premotor cortex, but atrophy also spreads into the prefrontal cortex. Studies have demonstrated that whole-brain and frontal atrophy are greater in PSP-RS than in PD 24,64,67,72,75 and MSA-P, 72 although visual assessment of frontal atrophy had poor sensitivity (17% and 57%) and moderate specificity (75% and 83%) in differentiating PSP-RS from MSA 13,20 in 2 studies, reflecting the fact that discernible frontal atrophy is only present in approximately 60% of PSP-RS patients. 13 Frontal atrophy may be more useful if considered in addition to brain stem regions. One study found that adding frontal, third ventricle, and wholebrain volumes to midbrain and SCP volumes improved the differentiation of PSP-RS from PD and MSA (sensitivity, 88.9%; specificity, 97.3%). 24 Another showed that combining frontal, ventricular, and whole-brain volumes could differentiate PSP-RS from PD and controls with 95.2% sensitivity and 90.9% specificity. 64 One caveat to consider, however, is that frontal atrophy is unlikely to differentiate PSP-RS from CBS, given that CBS shows striking frontal atrophy. 60,62,68,76 Quantitative methods for assessing frontal volume or thickness also vary widely across studies and may influence diagnostic utility.
Frontal atrophy also occurs in vPSP, particularly in PSP-F, 8 PSP-SL, 9,54,55 and PSP-CBS 52,77 and can be greater than in PSP-RS, 62 likely reflecting a shift in PSP pathological burden from brain stem to cortex. 78 The degree of frontal atrophy is similar in both PSP-PGF 79 and PSP-P 57 compared with PSP-RS. Although no diagnostic data are available on the value of frontal atrophy in vPSP, the presence of frontal atrophy would be consistent with these diagnoses. Data are needed to determine whether cortical measures could help to differentiate vPSP from other frontotemporal lobar degeneration disorders that are primarily characterized by frontal atrophy.

Other Subcortical Measures
Atrophy of subcortical structures, including the caudate nucleus, putamen, globus pallidus, subthalamus, and thalamus, has also been observed in group-level studies of PSP-RS either using visual assessment or volumetric measurements. 13,28,62,63,74,80-82 There is evidence that volumes of putamen, thalamus, and globus pallidus are smaller in PSP-RS than in PD, 82 with thalamus volumes also being smaller than in MSA-P. 28 However, studies have found that visual assessments of putamen and globus pallidus atrophy are not diagnostically useful in differentiating PSP-RS from MSA or PD. 13,20 The caudate nucleus, putamen, and thalamus have also been reported to be atrophic in CBS 13,62,83 and are unlikely to be diagnostically useful in differentiating PSP-RS and CBS. Basal ganglia structures have been reported to be atrophic in patients with PSP-P, 57 PSP-CBS, 52,77 and PSP-SL, 9 with thalamic atrophy reported in PSP-PGF. 79 However, the diagnostic value of these findings is unclear and limited to level 1 (

Pattern Approaches to Diagnosis
A number of studies have proposed that the assessment of multiple regions of the brain will optimize sensitivity and specificity for PSP-RS. These studies typically develop optimal prediction models 93 or use automated machine-learning techniques to identify diagnostic patterns. 94-100 A number of these studies have found that assessment of multiple regions including the midbrain, basal ganglia, 95,97,98,100 cerebellum, 98,100 or thalamus 99 provided excellent sensitivity and specificity to differentiate PSP-RS from PD and MSA-P. One study found that a prediction model using midbrain, putamen, and cerebellar gray-matter volumes could differentiate PSP-RS from MSA and PD with 90% sensitivity and 100% specificity in an early stage of the disease when not all patients had yet fulfilled clinical diagnostic criteria for these diseases. 100 It has also been suggested that volumetric white-matter measurements may show greater diagnostic utility than gray-matter measures. 94,96 There is also some evidence that a pattern-based approach using brain stem and cortical gray-and white-matter measures could be used in the differential diagnosis of autopsyconfirmed PSP and CBD. 93 Generally, assessing the pattern of atrophy, rather than focusing on specific regions, appears to be a sensible and sensitive and specific approach to differential diagnosis, although there is currently a lack of agreement across studies on which specific regions should be used, and further validation of these results in independent cohorts is necessary. In addition, no data are yet available on how well these approaches perform in vPSP. Further work is needed before these approaches can be incorporated into clinical criteria.

Diffusion Imaging
Measurements of microstructural damage using diffusion-weighted imaging (DWI) show some promise as biomarkers of PSP-RS. Apparent diffusion coefficient (ADC) measurements from DWI have been assessed in gray-and white-matter structures in PSP-RS, showing elevated ADC values in putamen, caudate, globus pallidus, midbrain, SCP, and prefrontal and precentral white matter. 101-107 PSP-RS patients typically show higher ADC values in the putamen, caudate nucleus, globus pallidus, SCP, and midbrain compared with PD, 102,106-108 with 1 study obtaining high sensitivity (90%) and specificity (100%) to differentiate PSP-RS from PD using values from the putamen 107 and another obtaining 100% sensitivity and specificity using the SCP. 103 Compared with MSA-P, PSP-RS has higher ADC values in the caudate nucleus 106 and SCP 103,106 but lower values in the MCP, 105,109 cerebellum, 110 and putamen. 101 Sensitivity and specificity values for differentiating PSP-RS from MSA-P are high using DWI of the SCP (sensitivity, 96.4%; specificity, 93.3% 103 ). Therefore, the diagnostic performance of DWI measurements is excellent, supporting these measurements as level 2 biomarkers (Table 3). There is no consensus regarding the best structure to assess, although the SCP appears promising.
Diffusion tensor imaging (DTI) allows for the assessment of directional water diffusion and the interrogation of specific white-matter tracts. White-matter tract degeneration has been demonstrated to be a striking feature of PSP-RS, with abnormalities observed predominantly in the SCP, cerebellum, body of the corpus callosum, cingulum, white-matter laminar of the thalamus, and premotor aspects of the superior longitudinal fasciculus. 63, [111][112][113][114][115][116][117][118][119][120][121][122][123][124] The majority of these white-matter tracts show greater degeneration in PSP-RS compared with PD 112,118,122,[125][126][127] and MSA-P. 72,118,126 Little data are currently available on the diagnostic utility of DTI measures, although the corpus callosum 113 and SCP 125 show high sensitivity and specificity in differentiating PSP-RS and PD. There is also evidence that adding DTI measures to the MRPI may help in the differentiation of PSP-RS from controls. 128 The diagnostic value of DTI measures to differentiate PSP-RS and MSA-P is unclear. It is also unclear whether DTI measures could differentiate PSP-RS and CBS, particularly given that patterns of DTI abnormalities overlap to a large degree between these 2 syndromes. 112,[129][130][131] A few studies have assessed DTI measures in PSP-P, which appears to show similar although slightly less severe patterns of tract abnormalities compared with PSP-RS. 128 Some studies have found regions with greater abnormalities in PSP-P compared with PSP-RS, although the results have not been consistent across studies. 117,120,128 In summary, DTI abnormalities are striking in PSP-RS and have the potential to be useful diagnostic biomarkers (Table 3). However, data are needed on the utility of both DWI and DTI measures in vPSP and autopsy-confirmed PSP. The issue of whether DWI and DTI measurements can be translated into clinical practice is also unclear, because there is little standardization of methods across studies and no established diagnostic cut points for these measurements.

PET/SPECT [ 18 F]FDG-PET
Studies of [ 18 F]-fluorodeoxyglucose PET (FDG-PET) have shown hypometabolism in the midbrain, basal ganglia, thalamus, and frontal lobes in PSP-RS, [132][133][134][135][136][137][138][139][140][141][142][143][144][145] with frontal involvement particularly targeting premotor, precentral, and prefrontal regions 134 and anterior cingulate 146 (Fig. 2A). In an autopsy cohort including 7 PSP patients (all PSP-RS), the most common FDG-PET findings were hypometabolism of the thalamus (100%), caudate (86%), midbrain (86%), and frontal lobes (71%). 145 PSP-RS tends to show greater frontal hypometabolism than PD and MSA, 146 with visual assessments of frontal hypometabolism producing good sensitivity (76%) and specificity (98%) for PSP-RS in 1 study. 147 Visual assessments of midbrain hypometabolism have performed modestly, with 1 study finding 79% sensitivity and 69% specificity in differentiating PSP-RS from MSA and CBS. 144 Consideration of the pattern of hypometabolism may hold more diagnostic promise. Visual assessment of the pattern of hypometabolism associated with PSP-RS (eg, anterior cingulate, midbrain, basal ganglia) gave 93% sensitivity and 90% specificity to differentiate PSP-RS from PD, MSA, and CBS in 1 study. 147 Automated pattern detection techniques have given mixed results. [148][149][150][151][152] Differentiating PSP-RS from CBS can be challenging, given that patterns of hypometabolism overlap between these 2 syndromes to a large degree 138,145,152 although there is some suggestion that PSP-RS may have greater hypometabolism in midbrain and thalamus, 136,153 and CBS patients have greater hypometabolism in parietal lobes. 135,138,153 The presence of hemispheric asymmetry in CBS may further help to differentiate it from PSP-RS. 145,152 Therefore, current evidence provides some support for frontal and midbrain hypometabolism or the combination of both as potential level 2 biomarkers of PSP-RS (Table 3). There is some evidence that hypometabolism in the striatum and cortex can be present before the development of clinical PSP-RS (level 3 biomarker), although this has only been observed in familial PSP. 154 Some FDG-PET findings have been reported in vPSP. One study found that PSP-P was associated with slightly greater hypometabolism of the putamen than PSP-RS, with less severe involvement of the thalamus, and that a putamen-to-thalamus ratio differentiated PSP-RS from PSP-P and PD with 100% sensitivity and 75% specificity. 155 The PSP-P patients in that study did not show much frontal hypometabolism. 155 Frontal hypometabolism has also not been observed in PSP-PGF, with midbrain hypometabolism only observed in 25% of patients. 156 Patients with PSP-SL have shown frontal, basal ganglia, and midbrain hypometabolism, 157,158 although these studies did not have autopsy confirmation. Taken together, these studies show that neither frontal nor midbrain hypometabolism is present consistently across the vPSP syndromes. Therefore, the presence of these features could be supportive of PSP, but the absence does not preclude underlying PSP.
However, there is a lack of standardization in the quantitative methods used across FDG-PET studies, particularly in regard to the choice of reference regions used to standardize regional uptake values, which vary across studies, including cerebellum, pons, cortical regions, or global mean values, each of which may have different limitations in PSP.

Dopamine Imaging
Striatal presynaptic dopamine binding, measured using dopamine active transporter (DAT) imaging using [123I]-FP-CIT SPECT or [18F]FP-CIT-PET, is consistently decreased in PSP-RS compared with controls 159 (Fig. 2B). However, decreased DAT binding has also been observed in PD, MSA-P, and CBS, [160][161][162][163][164] without differences in the degree of general striatal binding observed across groups. 160,162,165 However, studies have found that the caudate nucleus is affected to a greater degree in PSP-RS than in PD 161,163,166,167 and that regional patterns of binding, such as ratio of caudate to ventral striatum (sensitivity, 94%; specificity, 92%), 163 ratio of caudate to putamen, 166 or ratio of anterior-posterior putamen, 167 could help to differentiate PSP-RS from other parkinsonian disorders; however, diagnostic performance has not always been consistent with these measures. 164,167 It has also been shown that PSP-RS shows more symmetric striatal binding than PD, 168 although the diagnostic value of this finding is unclear. Overall the finding of reduced striatal DAT binding is highly supportive and sensitive for a diagnosis of PSP-RS, but heterogeneity across studies and lack of diagnostic data limit its value in differentiating across parkinsonian disorders (Table 3). Midbrain DAT binding is also decreased in PSP-RS, with lower binding than in PD but a degree of binding similar to in MSA. 160,169 Brain stem DAT levels could differentiate PSP-RS and MSA from PD with 89.7% sensitivity and 94.1% specificity in 1 study. 169 Little is currently known about the diagnostic utility of DAT findings in vPSP, although there is evidence from a few studies that both PSP-PGF and PSP-P are associated with striatal DAT reductions similar to those in PSP-RS, 156,[170][171][172] with similar putamen-to-caudate ratios. 171 Imaging using D2 receptor ligands, most commonly [123I]-IZBM SPECT, to assess postsynaptic dopaminergic function also appears to be sensitive in PSP-RS, with the majority of patients showing striatal reductions. [172][173][174][175][176] However, the value of D2 receptor ligand imaging in the differential diagnosis from other parkinsonian disorders is unclear. 162,175,176 In addition, there is some evidence that striatal uptake may not be reduced in PSP-P. 172

Tau-PET Imaging
The development of PET ligands that can bind to aggregated tau inclusions in the brain has been an exciting recent advance in the field with the potential of becoming a biomarker of tau pathology. A number of tau-PET ligands have been developed, 177 but the [ 18 F]AV-1451 (previously known as T807) ligand 178,179 has been the most widely used to date. Studies have demonstrated relatively consistent patterns of increased [ 18 F]AV-1451 uptake in PSP-RS compared with controls in the globus pallidus, putamen, caudate nucleus, thalamus, subthalamic nucleus, midbrain, and dentate nucleus of the cerebellum 180-184 ( Fig. 2C and E). The cortex has typically shown less striking uptake in PSP-RS, 183 with measures from subcortical structures showing the most promise as potential diagnostic biomarkers. 180,182,183 Quantification of globus pallidus retention provided sensitivity and specificity of 93% in differentiating PSP-RS from controls and 93% sensitivity and 100% specificity in differentiating PSP-RS and PD in 1 study, 180 although the thalamus provided the best separation between PSP-RS and controls in another. 182 There is also evidence that the pattern of uptake in PSP-RS differs from that in Alzheimer's disease (AD), with many of the PSP-RSrelated regions showing greater uptake in PSP-RS than in AD despite AD showing greater cortical [ 18 F]AV-1451 uptake. 183,184 Therefore, there is some evidence to support [ 18 F]AV-1451 as a level 2 biomarker of PSP-RS. A caveat is that overlap in the [ 18 F]AV-1451 signal is observed between PSP-RS and controls, 182 with 1 study failing to observe differences between PSP-RS and controls. 185 Standardization of methods will also be required, including optimizing scan time and quantitative outcomes. Current studies have analyzed standard uptake value ratio (SUVR) values referenced to cerebellar gray-matter 180,182,183 or binding potentials. 184 Although early studies are certainly encouraging, several limitations of [ 18 F]AV-1451 need to be considered. One caveat is that regions that show [ 18 F]AV-1451 uptake in PSP, including the basal ganglia, thalamus, midbrain, and dorsal cerebellum, also show some degree of "off-target" binding in normal subjects, which increases with age. 186,187 The nature of this binding is unclear. Although age correction in quantitative studies may go some ways to correct for this off-target binding, it will likely limit the value of [ 18 F]AV-1451 in the differential diagnosis of individual patients. Furthermore, it is unknown whether the off-target signal may also be altered by the disease in PSP, confounding any potential true signal of tau. Another caveat comes from an apparent disconnect between in-vivo and ex vivo studies. Although regions that show elevated binding typically show tau deposition at autopsy, autoradiographic studies have found little or no binding of [ 18 F]AV-1451 to tau in autopsied brains of PSP patients, 182,[187][188][189][190][191][192] casting doubt on whether the signal identified by [ 18 F]AV-1451 reflects tau pathology and whether it could be considered a level 5 biomarker of tau. This kind of disconnect is not uncommon for PET tracers, and the utility of such in vitro studies has been questioned. 193 However, a recent article found that tau pathology discovered postmortem in a patient with PSP correlated with antemortem FDG-PET but not with [ 18 F]AV-1451 signal. 190 Another caveat is that elevated [ 18 F]AV-1451 uptake has also been observed in nontau diseases, 187,194 which again questions the specificity of the ligand to 4R tau. Another chemically distinct tau PET ligand, THK-5351, 195 was found to have high affinity for PSP tau lesions in an autoradiographic study 196 and has shown uptake in the globus pallidus and midbrain 196 in patients with PSP-RS ( Fig. 2D and  F). However, the degree of off-target THK-5351 binding in PSP-related regions is at least as high, if not higher, than that observed with [ 18 F]AV-1451. 197 Overall, much more work needs to be done to evaluate these PET tracers. It is likely that different tau-PET ligands may bind to tau conformers with differing sensitivity and specificity and show different off-target binding, and hence head-to-head and indirect comparisons of the currently available tau imaging agents are needed.

Other Biomarkers
There are a number of other neuroimaging biomarkers that have been assessed in PSP-RS with fewer data available to assess diagnostic value. MR modalities that demonstrate abnormalities in PSP-RS include magnetic resonance spectroscopy and magnetization transfer imaging, 73,[198][199][200][201][202][203][204][205] although the ability of these modalities to differentiate PSP-RS from other parkinsonian disorders is unclear. 181,184,185,205 Resting-state (task-free) functional MRI has also been used to demonstrate abnormalities in functional connectivity in PSP-RS across the network of PSP-RS-associated regions, 63,206,207 but the loss of cortical connectivity is not specific to PSP-RS versus PD. 208 Longitudinal MR studies have shown increased rates of whole-brain, cortical, and midbrain atrophy and SCP diffusivity in PSP-RS compared with controls, [209][210][211][212][213][214][215][216][217][218] with some evidence for greater rates than in PD, but similar rates of whole-brain and midbrain atrophy as in MSA-P. 212,215 However, cortical and whole-brain rates of atrophy are greater in CBS than in PSP-RS. 209,213 Cerebral blood flow single-photon emission computed tomography studies have demonstrated frontal [219][220][221][222][223][224][225] and, less commonly, thalamic 220 and striatal 222 hypoperfusion in PSP-RS. 221,226 Findings concerning differentiating PSP-RS from other parkinsonian disorders are lacking here, although PSP-RS may show greater frontal hypoperfusion than PD. 224,227 Abnormalities in other neurotransmitter systems, such as the cholinergic [228][229][230] and serotoninergic 231 systems, have also been demonstrated in PSP-RS.

Conclusions
Neuroimaging research over the last several decades has improved our understanding of the neurobiology of PSP but has not yielded many confirmed diagnostic biomarkers ( Table 3). The most mature research area is the assessment of midbrain measurements, which has yielded a number of measures that have good sensitivity and specificity for PSP-RS versus other parkinsonian disorders, such as midbrain-pons area and the MRPI, which appear to be the most reliable biomarkers for the diagnosis of PSP-RS. The presence of frontal atrophy and hypometabolism are also prominent features of PSP-RS and may improve diagnosis when considered together with midbrain atrophy. It is clear that PSP-RS is associated with striking damage to the white matter, with DWI measures of the SCP providing good sensitivity and specificity for PSP-RS diagnosis, although data supporting this measure come from only a couple of studies. DTI measures could prove to be very valuable, although more work is needed to provide and validate standardized measures of the kind that could be used in diagnostic criteria. Measures of dopamine function are highly sensitive to PSP-RS and many of the vPSP syndromes, but specificity is low, and thus they are less useful in ruling out other parkinsonian syndromes. Data so far only support neuroimaging biomarkers as level 2 biomarkers for PSP-RS. Only a handful of studies have assessed patients early in the disease course to suggest level 3 biomarkers. More work is needed to assess the value of these measures in vPSP and in autopsyconfirmed cases to determine whether they could be useful level 4 biomarkers. Capturing the disease in its earliest phase will also be critical for developing wellvalidated level 3 biomarkers. Last, tau-PET imaging techniques are exciting, but more work is needed to truly understand the biological underpinnings of the tau-PET signal in PSP. However, these are early days in tau-PET imaging, and we expect our understanding of these biomarkers to increase exponentially over the coming years.