Schlüsselwörter Dopamintransporter-Szintigrafie, 123 I-Ioflupan, FP-CIT, DaTSCAN, semi-quantitative Analyse, Referenzregion, globale Skalierung

Transcription

1 Original article 1 Global scaling for semi-quantitative analysis in FP-CIT SPECT D. Kupitz 1 *; I. Apostolova 1 *; C. Lange 2 ; G. Ulrich 1 ; H. Amthauer 1 ; W. Brenner 2 ; R. Buchert 2 1 Radiologie und Nuklearmedizin, Otto-von-Guericke-Universität, Magdeburg, Germany; 2 Nuklearmedizin, Universitätsmedizin Charité Berlin, Germany Keywords Dopamine transporter scintigraphy, 123 I- ioflupane, FP-CIT, DaTSCAN, semi-quantitative analysis, reference region, global scaling Summary Semi-quantitative characterization of dopamine transporter availability from single photon emission computed tomography (SPECT) with 123 I-ioflupane (FP-CIT) is based on uptake ratios relative to a reference region. The aim of this study was to evaluate the whole brain as reference region for semiquantitative analysis of FP-CIT SPECT. The rationale was that this might reduce statistical noise associated with the estimation of non-displaceable FP-CIT uptake. Patients, methods: 150 FP-CIT SPECTs were categorized as neuro degenerative or non-neurodegenerative by an expert. Semi-quantitative analysis of specific binding ratios (SBR) was performed with a custom-made tool based on the Statistical Parametric Mapping software package using predefined regions of interest (ROIs) in the anatomical space of the Montreal Neurological Institute. The following reference regions were compared: predefined ROIs for frontal and and whole brain (without striata, thalamus and brainstem). Tracer uptake in the reference region was characterized by the mean, median or 75th percentile of its voxel intensities. The area (AUC) under the receiver operating Correspondence to: Ralph Buchert, PhD Charité Universitätsmedizin Berlin Department of Nuclear Medicine Charitéplatz 1, Berlin, Germany Tel. +49/(0)30/ Fax +49/(0)30/ * These authors contributed equally. characteristic curve was used as performance measure. Results: The highest AUC of was achieved by the SBR of the putamen with the 75 th percentile in the whole brain as reference. The lowest AUC for the putamen SBR of was obtained with the mean in the as reference. Conclusion: We recommend the 75 th percentile in the whole brain as reference for semi-quantitative analysis in FP-CIT SPECT. This combination provided the best agreement of the semi-quantitative analysis with visual evaluation of the SPECT images by an expert and, therefore, is appropriate to support less experienced physicians. Schlüsselwörter Dopamintransporter-Szintigrafie, 123 I-Ioflupan, FP-CIT, DaTSCAN, semi-quantitative Analyse, Referenzregion, globale Skalierung Zusammenfassung Die semi-quantitative Charakterisierung der Verfügbarkeit von Dopamintransportern bei der Einzel-Photonen-Emissions-Tomographie (SPECT) mit 123 I-Ioflupan (FP-CIT) basiert auf dem Vergleich des Uptakes in striatalen Regionen mit einer Referenzregion. Ziel dieser Studie war die Evaluierung des gesamten Gehirns als Referenzregion für die semi-quantitative Analyse bei der FP-CIT SPECT. Die Rationale hierfür war, dass dies zu einer Reduktion des Globale Skalierung des Tracer-Uptakes für die semi- quantitative Analyse in der FP-CIT SPECT Nuklearmedizin 2014;53:- received: April 4, 2014 accepted in revised form: August 25, 2014 epub ahead of print: September 11, 2014 statistischen Fehlers bei der Bestimmung der unspezifischen FP-CIT-Aufnahme führen könnte. Patienten, Methoden: 150 FP-CIT SPECTs wurden retrospektiv eingeschlossen und mittels visueller Beurteilung durch einen erfahrenen Befunder als neurodegenerativ oder nicht neurodegenerativ kategorisiert. Für den benutzerunabhängigen Vergleich der verschiedenen Referenzregionen wurden semi-quantitative Analysen von SBR (specific binding ratios) vollautomatisch mit einem eigenen Tool durchgeführt, welches Routinen des Statistical Parametric Mapping Software- Pakets und im anatomischen Raum des Montreal Neurological Institutes vordefinierte ROIs (regions of interest) verwendet. Es wurden folgende Referenzregionen miteinander verglichen: ROIs für den Frontal- und Okzipitallappen, sowie das gesamte Gehirn (ohne Striatum, Thalamus und Hirnstamm). Der Tracer Uptake in der Referenzregion wurde durch den Mittelwert, Median oder das 75. Perzentil der Voxelintensitäten in der Referenzregion charakterisiert. Die Fläche (AUC) unter der Receiver Operating Characteristic Kurve für die Differenzierung von neurodegenerativ und nicht neurodegenerativ wurde als Maß für das diagnostische Potential der semi-quantitativen Analyse verwendet. Ergebnisse: Die größte AUC (0,973) erzielte das SBR des Putamens mit dem 75. Perzentil im gesamten Gehirn als Referenzwert. Die kleinste AUC (0,937) für das Putamen SBR ergab sich mit dem Mittelwert des Frontallappens als Referenz. Schlussfolgerung: Wir empfehlen die Verwendung des 75. Perzentils im gesamten Gehirn als Referenz für die semi-quantitative Analyse von FP-CIT SPECTs. Dieser Referenzwert scheint zur besten Übereinstimmung von semi-quantitativer Analyse mit visueller Beurteilung der SPECT Bilder durch einen erfahrenen Experten zu führen und ist daher für die Unterstützung weniger erfahrener Befunder besonders geeignet. Schattauer 2014 Nuklearmedizin 6/2014

2 2 D. Kupitz et al.: Global scaling in FP-CIT SPECT Single photon emission computed tomography (SPECT) with N-ω-fluoro - propyl-2β-carbomethoxy-3β-(4-123 I-iodo - phenyl)nortropane (FP-CIT) is widely used for the differentiation of Parkinsonian syndromes (PS) with presynaptic dopaminergic loss (referred to as neurodegenerative PS) versus PS without such loss (non-neurodegenerative PS) (4, 6, 11, 12, 19, 21, 24). FP-CIT is registered for this indication in both the United States and in Europe. Guidelines for the diagnosis of Parkinson s disease recommend FP-CIT SPECT with the highest evidence level, particularly in the presence of significant diagnostic uncertainty such as parkinsonism associated with neuroleptic exposure and atypical tremor manifestations such as isolated unilateral postural tremor (3). In Europe, FP-CIT is also registered for the differentiation between dementia with Lewy bodies and Alzheimer s disease (27). The interpretation of FP-CIT SPECT is based on visual assessment of the SPECT images. Visual analysis can be supported by semi-quantitative analysis of FP-CIT uptake in the striatum (8, 9, 13, 15, 20, 22, 29). Semi-quantitative analysis can be useful in follow-up studies and in borderline cases with mild reduction of FP-CIT uptake, particularly for less experienced raters. Semiquantitative analysis of FP-CIT SPECT usually is performed as pseudo-equilibrium analysis, to be more specific, by determination of the so-called specific binding ratio SBR according to the following formula SBR = (C ROI : C REF ) 1 (1), where C ROI and C REF and are the total tracer concentration in the region of interest and the reference region, respectively, during pseudo-equilibrium (in case of FP-CIT 3 6 h post injection). It is evident from this formula that the choice of an adequate reference region is important for reliable determination of the SBR. A pivotal requirement for the reference region is that it is void of the specific binding site or shows only very low concentrations. Considering the dopamine transporter (DAT), Hall and co-workers, using autoradiography with the highly selective DAT ligand 125 I-PE2I in normal post mortem human brains, found high DAT density in the striatum (10). Distinct DAT density was also found in the substantia nigra in the brainstem (about 50% of the level in the striatum). Other brain structures including thalamus, hippocampus, cerebellum and neocortical regions showed no or very low DAT density. According to these findings, all brain regions except striatum and brainstem (substantia nigra) appear appropriate as reference region in FP-CIT SPECT. However, very often the reference region is restricted to the (8, 9) or, less often, to the or the cerebellum (15). This restriction to a smaller than necessary reference region might result in increased test-retest and inter-subject variability of the SBR, not only due to increased statistical noise of C REF, but also due to variability in the identification (segmentation) of the reference region in each individual scan. In the, inter-subject variability might be further increased by particularly large unspecific uptake in white matter (10). Increased inter-subject variability reduces the statistical power of the SBR to detect alterations of DAT availability. This loss of diagnostic power might be reduced by increasing the spatial extent of the reference region. In the present study we propose and evaluate the largest possible reference region for FP-CIT SPECT, the whole brain without the striatum and the brainstem. The thalamus was also excluded from the reference region, because it shows rather high density of serotonin transporters (14) to which FP-CIT also binds with non-negligible affinity (28). A first step towards the use of the whole brain without striatum as reference region in FP-CIT SPECT has been described by Tossici-Bolt and co-workers who used a 44 mm thick transaxial slice centred at the highest striatal FP-CIT uptake for semiquantitative analyses (23). The whole brain (with exclusion of the striata) enclosed in this slab was used as reference region. The present study extends the so-called 2-dimensional method of these authors to a fully 3-dimensional approach by using a 3-dimensional region of interest (ROI) for the whole brain without striatum, thalamus and brainstem as reference region. In the following we abbreviate whole brain without striatum, thalamus and brainstem by (= all regions of the whole brain which show only non-displaceable FP-CIT uptake). 150 consecutive patients from clinical routine were included retrospectively for comparison of with occipital or as reference region in semiquantitative analysis. All patients had been referred to FP-CIT SPECT because of a clinically uncertain Parkinsonian syndrome. The mean, the median and the 75 th percentile of all voxel values within the reference region were compared for characterization of the FP-CIT uptake in the reference region. The rationale for the median instead of the mean was to exclude voxels in cerebrospinal fluid, and that the median might be less sensitive to moderate (disease-related) intensity changes. The rationale for the 75 th percentile was that it might be more representative of grey matter voxels than the median (50 th percentile). The latter is expected to represent both grey and white matter. In order to avoid operator-dependence and potential bias in the semi-quantitative analyses, a custom made, SPM-based software tool for fully automated analysis of FP-CIT SPECTs was used, similar to a tool previously developed by our group for fully automated analysis of dopamine receptor SPECT with 123 I-IBZM (7). Patients, material, methods Patients 150 consecutive patients who had been referred to the Department of Nuclear Medicine of the university medical centre Magdeburg, Germany, between August 2012 and September 2013 for FP-CIT SPECT because of an unclear Parkinsonian syndrome (PS) were recruited retrospectively. No further inclusion or exclusion criteria were applied. Patients were categorized as either neurodegenerative or non-neurodegenerative PS by an experienced physician (I.A.) based on visual inspection of the SPECT images using the rating scale described by Benamer and co-workers (2). Nuklearmedizin 6/2014 Schattauer 2014

3 D. Kupitz et al.: Global scaling in FP-CIT SPECT 3 Fig. 1 Predefined standard masks for the regions of interest (caudate, putamen) and the different reference regions in the anatomical space of the Montreal Neurological Institute. The whole brain mask excludes caudate, putamen, thalamus and brainstem (it includes the frontal and the ) fro: ; cau: caudate; put: putamen; wb: whole brain; occ: ; tha: thamalus The visual evaluation was performed blinded for both clinical information and the results of the semi-quantitative analysis. 66 patients were categorized as nonneurodegenerative PS (36 women, age 67.8 ± 12.8 years, years), 84 patients as neurodegenerative PS (29 women, age 68.1 ± 11.8 years, years). This categorization was used as ground truth. The study protocol had been approved by the local Clinical Institutional Review Board (reference number 02/14, RAD241) and complied with the Declaration of Helsinki. SPECT acquisition SPECT data had been acquired with a dual head gamma camera E.CAM 7957 (Siemens AG, Erlangen, Germany) equipped with low-energy, high-resolution fanbeam collimators (focal distance 417 mm). The acquisition had started 3 hours after intravenous injection of MBq FP-CIT following blocking of the thyroid gland by oral administration of perchlorate. A matrix was used and an energy window with a width of ± 10% was centred at the photopeak of 123 I at 159 kev. The collimator-patient distance ranged between 16.5 and 18.0 cm. Each camera head covered 180 of a circular orbit in step and shoot mode with 60 angular views at steps of 3. The duration of the acquisition at each step was 40 s resulting in a total acquisition time of about 40 min. Transversal SPECT images were reconstructed by the filtered backprojection algorithm implemented in the scanner software (esoft 3.5) with a Butterworth filter (cutoff cm -1 ). Attenuation correction was performed by Changs method with a linear attenuation coefficient of 0.11 cm -1. No scatter correction was applied. The size of the reconstructed voxels was mm3. Semi-quantitative analyses Semi-quantitative analyses were performed fully automatically using a custom made, SPM-based software tool for the analysis of FP-CIT SPECT. The tool is similar to a tool previously developed by our group for fully automated analysis of dopamine receptor SPECT with 123 I-IBZM (7). The SPM tool first stereotactically normalizes the patient s FP-CIT SPECT to a custom made FP-CIT template in the anatomical space of the Montreal Neurological Institute (MNI) ( Fig. 1). The normalization routine of SPM8 is used for this purpose, including both affine transformation and warping (parameter settings: source image smoothing: 16 mm, source weighting image: none, template image smoothing: 0 mm, template weighting image: none, affine regularization: MNI space, non-linear frequency cut-off: 25, non-linear iterations: 12, non-linear regularization: 1, intensity modulation: preserve intensities of original image, voxel size: mm 3, bounding box: [ ; ], interpolation: tri-linear, wrapping: none). The fully automated ROI analysis performed by the tool is based on standard ROIs predefined as masks in MNI space ( Fig. 1). The whole brain was defined by the standard SPM8 brain mask. Thalamus and brainstem were removed from the whole brain mask using the ROIs for thalamus and brainstem (union of midbrain, pons and medulla) defined in the Wake Forest University (WFU) pickatlas (TDLobes, TDLabels) (17). The masks for occipital and were created from the corresponding ROIs of the WFU pickatlas (TDLobes), too. ROIs for caudate and putamen were defined manually in one hemisphere and then mirrored to the other. These ROIs are much bigger than caudate and putamen in order to guarantee that the actual caudate and putamen are completely included in the standard ROIs in each individual patient, independent of some residual anatomical inter-subject variability after stereotactical normalization. Mean FP-CIT uptake in the caudate and in the putamen was obtained by hottest voxel analysis within the large ROIs, separately in left and right hemisphere. The number of hottest voxels to be averaged was fixed to a total volume of 5 ml for the caudate and 10 ml for the putamen. The total volume of 15 ml for caudate and putamen is compatible with the normal range of the striatum volume in healthy subjects (1). In case of severely reduced FP-CIT uptake in the putamen, for example, manual definition of the putamen ROI tends to de- Schattauer 2014 Nuklearmedizin 6/2014

4 4 D. Kupitz et al.: Global scaling in FP-CIT SPECT Tab. 1 Results of the semi-quantitative analysis: specific binding ratio in caudate and putamen, putamen-to-caudate ratio and left-right asymmetry (in %) in caudate and putamen for all combinations of reference region and characteristic value. Given are the mean value ± 1 standard deviation in the group of patients with neurodegenerative Parkinsonian syndrome (PS) and the group with non-neurodegenerative PS. For the specific binding ratio and the putamento-caudate ratio the minimum over both hemispheres was used. Parkinsonian syndrome reference region specific binding ratio caudate putamen characteristic putamen-to-caudate ratio left/right asymmetry caudate putamen neuro - degenerative (n = 84) mean BRASS 1.52 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± th percentile 1.49 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 15.9 median 1.86 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 13.3 non-neuro - degenerative (n = 66) mean BRASS 2.53 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± th percentile 2.18 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 5.2 median 2.62 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 4.7 : all regions of the whole brain which show only non-displaceable FP-CIT uptake lineate the regions with more or less conserved FP-CIT uptake. This results in overestimation of the FP-CIT uptake in the whole putamen (underestimation of the reduction) which in turn reduces the sensitivity for detection of reduced uptake. This limitation of manual ROI definition is avoided by the fully automated hottest voxel approach by using a fixed number of hottest voxels (stereotactical normalization is expected to eliminate inter-subject differences in the volume of the striatum and its subregions to a large extent, justifying the use of the same fixed striatal volume in all patients). FP-CIT uptake in the reference region was characterized either by the mean or by the median or by the 75 th percentile of the voxel values within the reference region. Specific binding ratios (SBR) were computed as the ratio of FP-CIT uptake in the region of interest (caudate, putamen) to the characteristic uptake in the reference region minus 1 (equation 1). This was done for each combination of reference region ( or or ) and characteristic value (mean or median or 75 th percentile), resulting in a total of nine SBRs for each region of interest. These nine SBRs for each region of interest were computed for all 150 patients fully automatically within a single batch processing session within a few minutes on a standard personal computer. The cerebellum, which is also used as reference region by some groups, was not considered in the present study, because it was not completely within the field-of-view of SPECT imaging in some patients (with short neck). For the as reference region, incomplete coverage of the brain by the SPECT field-of-view was taken into account by restricting the computation of the characteristic value (mean, median, 75 th percentile) to voxels with non-zero intensity. Semi-quantitative analysis was also performed with the commercially available brain analysis software BRASS (version 3.5, Hermes Medical Solutions, Stockholm, Sweden) (25). SPECT raw data were reconstructed iteratively on a Scivis workstation (Scivis-GmbH) with ReSPECT 2.5 software. BRASS stereotactically normalizes the patient s FP-CIT SPECT to an integrated template and then performes the ROI analyses. FP-CIT uptake is characterized by the mean value both in the region of interest (caudate, putamen) and in the reference region (). SBRs are Nuklearmedizin 6/2014 Schattauer 2014

5 D. Kupitz et al.: Global scaling in FP-CIT SPECT 5 Tab. 2 Area under the receiver operator characteristic curve (AUC) for the differentiation of neurodegenerative versus non-neurodegenerative Parkinsonian syndrome by the semi-quantitative measures of FP-CIT uptake as function of the reference region and the value used to characterize FP-CIT uptake in the reference region. The 95%-confidence interval of the AUC is given in square brackets. characteristic reference region specific binding ratio caudate putamen putamen-to-caudate ratio left/right asymmetry caudate putamen BRASS [ ] [ ] [ ] [ ] [ ] mean [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] 75 th percentile [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] median [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] : all regions of the whole brain which show only non-displaceable FP-CIT uptake computed as described above. It should be noted that, because of the use of a different reconstruction algorithm by the BRASS software, observed differences between results from BRASS and results obtained with the custom made SPM tool described above are a mixed effect of image reconstruction, ROI definition and characteristic value. The putamen-to-caudate ratio in both hemispheres and left/right asymmetry were considered in addition to SBRs: asym(%) = 200 abs[(sbr left SBR right ) : (SBR left + SBR right ) Statistical analyses Statistical analyses used the minimum over both hemispheres for SBRs and the putamen-to-caudate ratio in each patient. Univariate analysis of variance was performed with the SBR as dependent variable and the following variables as fixed factors: group (neurodegenerative vs non-neurodegenerative PS), region of interest (caudate, putamen), reference region (, occipital lobe, ) and type of characteristic value within the reference region (mean, median, 75 th percentile). Receiver operator characteristic (ROC) analysis was performed to evaluate the impact of the reference region and the characteristic value on the power of the semiquantitative measures for differentiation between non-neurodegenerative and neuro degenerative PS. The area under the ROC curve (AUC) was used as performance measure. (Maximum) diagnostic accuracy was determined by the point on the ROC curve closest to the upper left corner. Univariate and ROC analyses were performed with SPSS (version 19, IBM Corp., Armonk, NY, USA). Results The processing quality of the fully automatic analysis of the FP-CIT SPECTs with the custom-made SPM tool was supervised by visual inspection of a 12 mm thick transversal slab through the striatum after stereotactical normalization (7). The latter is the critical processing step. However, stereotactical normalization worked properly in all cases, although no patient was excluded based on technical constraints such as poor quality of the SPECT image. Results of the semi-quantitative analysis are summarized ( Tab. 1). Univariate analysis of variance revealed a highly significant effect (p < ) on the SBR for all considered factors, i.e. group (lower SBR in neurodegenerative than in non-neurodegenerative PS), region of interest (lower SBR in putamen than in caudate), reference region (lowest SBR for ) and type of characteristic value within the reference region (lowest SBR for 75 th percentile). In addition, the following inter - actions had a significant effect on the SBR group: region of interest (p < , smaller putamen-to-caudate ratio in neuro - degenerative PS), reference region (p = 0.038, smallest group difference with as reference region) and reference region characteristic (p = 0.018, reduction of SBR with occipital lobe as reference region compared to and more pronounced for mean and median than for 75 th percentile). The results of the ROC analyses for the differentiation between neurodegenerative and non-neurodegenerative PS by the semi-quantitative measures are summarized ( Tab. 2). ROC curves for the SBR in caudate and putamen are shown Fig. 2. The semi-quantitative measures obtained by the custom-made SPM tool provided very similar diagnostic accuracy as the semiquantitative values provided by the BRASS software ( Tab. 2). Concerning the impact of the reference region and the characteristic value, the largest AUC of was achieved by the putamen SBR (minimum Schattauer 2014 Nuklearmedizin 6/2014

6 6 D. Kupitz et al.: Global scaling in FP-CIT SPECT Fig. 2 Receiver operator characteristic (ROC) curves for the differentiation of neurodegenerative versus non-neurodegenerative Parkinsonian syndrome by the SBR of the caudate (left) and the putamen (right). Each coloured line represents a given combination of reference region and characteristic value of FP-CIT uptake in the reference region as specified in the legend. : whole brain (wb) without striatum, thalamus and brainstem; occ: : fro: ; 75 th perc.: 75 th percentile of left and right hemisphere) with the 75 th percentile in the whole brain as reference (maximum accuracy: 133 of 150 patients classified correctly, i.e. 88.7%). Using the occipital or as reference region (with the 75 th percentile) resulted in small reduction of the diagnostic accuracy of the putamen SBR (accuracy 87.3% and 86.7%). The effect of the characteristic value, i.e. mean vs. median vs. 75 th percentile, was also small (with whole brain as reference 88.7% accuracy for 75 th percentile and median, 85.3% for mean). The AUC was much smaller for the putamen-to-caudate ratio and the left/right asymmetries than for the SBRs (Tab. 2). Discussion The aim of our study was to evaluate the whole brain without striatum, thalamus and brainstem () as reference region for semi-quantitative analysis of FP-CIT SPECT. The choice of the reference region might be considered a technical aspect, but, as is evident from equation (1), the computation of the SBR is as sensitive to errors in the estimation of the FP-CIT uptake in the reference region (C REF ) as it is sensitive to errors in the estimation of the FP-CIT uptake in the region of interest (C ROI ). Thus, the choice of the reference region can have a significant impact on the error of the SBR, and, as direct consequence, on the diagnostic value of semiquantitative analysis. Increased variability of the SBR results in decreased statistical power for the detection of disease-related alterations of DAT availability. The is the largest possible reference region for FP-CIT SPECT, because the reference region must not contain the binding site of interest (the dopamine transporter) or other specific binding sites (the serotonin transporter), or at very low concentrations only. The rationale for choosing the largest possible reference region is that this might reduce the error of the estimation of the FP-CIT uptake in the reference region by decreased statistical noise of the mean (or other characteristic values) in an spatially increased reference region (larger sample size), and decreased error in the identification of the reference region (placement of the ROI) in each individual patient. In this study, the use of as reference region resulted in only a small improvement of the agreement between semiquantitative analysis and visual expert reading compared to the, although the effect on the SBRs itself was much larger ( Tab. 1) and highly significant statistically. The fact that the impact on the agreement was small suggests that neither the statistical error nor the error of the ROI placement was strongly reduced by the whole brain compared to the occipital lobe as reference. The lack of an effect on the statistical error might in part be explained by the rather large FP-CIT doses of MBq used in the patients included in this study. The effect of the reference region on the statistical quality of the estimate of the FP-CIT uptake in the reference region might be larger at lower FP-CIT doses, particularly in combination with the use of single-head SPECT systems. The guidelines of the (American) Society of Nuclear Medicine for FP-CIT SPECT recommend a dosage of MBq (9), in agreement with the prescribing information of the manufacturer. The European Association of Nuclear Medicine recommends MBq (9). The lack of an effect of the reference region on the error of the placement of the ROI might be explained by the fact that the semi-quantitative analysis was performed fully automatically using stereotactical normalization of each individual patient s SPECT image and predefined standard masks for the considered reference regions. Stereotactical normalization worked properly in all 150 cases, independent of the FP- CIT uptake in the striatum, although there were no inclusion or exclusion criteria with respect to the image quality of the SPECT images. This indicates the robustness of the method, which is an important prerequisite for its routine use in everyday clinical patient care. Successful stereotactical normalization and the use of predefined ROI masks for the reference regions eliminate errors of the identification of the reference region by manual placement (both intraand inter-rater errors) (26). The benefit of the whole brain as reference region, therefore, might be larger in case of manual ROI placement. Concerning the characteristic value used for the description of FP-CIT uptake in the reference region, a percentile might be less sensitive to some disease-related (or physiological) local alterations of tracer uptake in the reference region. This effect may have been observed in the present study to some extent with the whole brain as reference region: the AUC under the Nuklearmedizin 6/2014 Schattauer 2014

7 D. Kupitz et al.: Global scaling in FP-CIT SPECT 7 ROC curve for the differentiation between neurodegenerative and non-neurodegenerative PS by the putamen SBR increased from for the mean value to for the median and for the 75 th percentile. Although the impact of reference region and characteristic value separately appears rather small in the present study, the total difference between the best combination (75 th percentile in the whole brain) and the worst (mean or median in ) was already rather sizable: AUC versus Diagnostic accuracy was 88.7% versus 84.7%, i. e. the best combination provided 4 % better accuracy than the worst. This appears a relevant improvement, particularly when taking into account that the diagnostic accuracy of 84.7% for the worst combination was already rather good, i. e. there was not very much space for improvement. The diagnostic value of the SBR was slightly smaller for the caudate than for the putamen ( Tab. 2). This is in agreement with the temporal progress of neurodegeneration in Parkinson s disease which affects the (posterior) putamen first (5, 13, 18). The diagnostic value of the secondary semi-quantitative measures putamen-tocaudate ratio and left/right asymmetries was considerably smaller than that of the primary SBR measures ( Tab. 2). Combining the putamen SBR with the putamen-to-caudate ratio and the left/right asymmetries by binary linear regression also did not improve the diagnostic power compared to the putamen SBR alone. This suggests that putamen-to-caudate ratio and left/right asymmetries provide no or very little additional information to improve the power of semi-quantitative analysis in FP-CIT SPECT in clinical routine. In this study, visual analysis of the SPECT images by an expert was used as gold standard for the evaluation of semiquantitative analysis. The rationale for this was that semi-quantitative analysis might support less experienced readers of FP-CIT SPECT so that they might achieve similar diagnostic performance as experts. However, best agreement between semi-quantitative analysis and the visual expert rating was 88.7%, i. e. there was disagreement in 11.3% of the cases. Most of these were borderline cases with rather mild reduction of striatal FP-CIT uptake. It would be very interesting to clarify these cases, i. e. to check whether visual or semi-quantitative analysis was correct. However, this requires a reliable ground truth, clinical follow-up over a period of 2 years, for example, which was not available. Conclusion We recommend the 75 th percentile in the whole brain (without striatum, thalamus and brainstem) as reference for semiquantitative analysis in FP-CIT SPECT. This combination provides slightly better agreement of the semi-quantitative analysis with visual evaluation of the SPECT images by an expert compared to the mean FP- CIT uptake in a smaller reference region such as the. We hypothesize that the improvement by the 75 th percentile in the whole brain is particularly large in case of low statistical image quality (low FP-CIT dose) and/or manual placement of ROIs. However, it should be mentioned that the impact of the different semi-quantitative analysis methods tested in the present study on the diagnostic accuracy was rather small, which underlines the robustness of FP-CIT SPECT. It also suggests that it probably is more important that the analysis is always done the same way than using a particular method. Conflict of interest The authors declare that that they have no conflicts of interest. References 1. Aylward EH, Li Q, Habbak QR et al. Basal ganglia volume in adults with Down syndrome. Psychiatry Res 1997; 74: Benamer TS, Patterson J, Grosset DG et al. Accurate differentiation of parkinsonism and essential tremor using visual assessment of [ 123 I]-FP-CIT SPECT imaging: the [ 123 I]-FP-CIT study group. Movement disorders 2000; 15: Berardelli A, Wenning GK, Antonini A et al. EFNS/MDS-ES/ENS [corrected] recommendations for the diagnosis of Parkinson s disease. Eur J Neurol 2013; 20: Berding G, Brucke T, Odin P et al. [ 123 I]beta-CIT SPECT imaging of dopamine and serotonin transporters in Parkinson s disease and multiple system atrophy. Nuklearmedizin 2003; 42: Bernheimer H, Birkmayer W, Hornykiewicz O et al. Brain dopamine and the syndromes of Parkinson and Huntington. Clinical, morphological and neurochemical correlations. J Neurol Sci 1973; 20: Booij J, Tissingh G, Boer GJ et al. [ 123 I]FP-CIT SPECT shows a pronounced decline of striatal dopamine transporter labelling in early and advanced Parkinson s disease. J Neurol Neurosurg Psychiat 1997; 62: Buchert R, Berding G, Wilke F et al. IBZM tool: a fully automated expert system for the evaluation of IBZM SPECT studies. Eur J Nucl Med Mol Imaging 2006; 33: Darcourt J, Booij J, Tatsch K et al. EANM procedure guidelines for brain neurotransmission SPECT using 123 I-labelled dopamine transporter ligands, version 2. Eur J Nucl Med Mol Imaging 2010; 37: Djang DS, Janssen MJ, Bohnen N et al. SNM practice guideline for dopamine transporter imaging with 123 I-ioflupane SPECT 1.0. J Nucl Med 2012; 53: Hall H, Halldin C, Guilloteau D et al. Visualization of the dopamine transporter in the human brain postmortem with the new selective ligand [ 125 I]PE2I. NeuroImage 1999; 9: Hesse SOC, Meyer PT, Roessler A et al. Is there a role for I-123-FP-CIT SPECT in the management of suspected Parkinsons disease? J Nucl Med 2003; 44 (5 Suppl): Innis RB, Seibyl JP, Scanley BE et al. Single photon emission computed tomographic imaging demonstrates loss of striatal dopamine transporters in Parkinson disease. Proc Natl Acad Sci USA 1993; 90: Kahraman D, Eggers C, Holstein A et al. 123 I-FP- CIT SPECT imaging of the dopaminergic state. Visual assessment of dopaminergic degeneration patterns reflects quantitative 2D operator-dependent and 3D operator-independent techniques. Nuklearmedizin 2012; 51: Kish SJ, Furukawa Y, Chang LJ et al. Regional distribution of serotonin transporter protein in postmortem human brain: is the cerebellum a SERTfree brain region? Nucl Med Biol 2005; 32: Koch W, Hornung J, Hamann C et al. Equipmentindependent reference values for dopamine transporter imaging with 123 I-FP-CIT. Nuklearmedizin 2007; 46: Koch W, Radau PE, Hamann C, Tatsch K. Clinical testing of an optimized software solution for an automated, observer-independent evaluation of dopamine transporter SPECT studies. J Nucl Med 2005; 46: Maldjian JA, Laurienti PJ, Kraft RA, Burdette JH. An automated method for neuroanatomic and cytoarchitectonic atlas-based interrogation of fmri data sets. NeuroImage 2003; 19: Oh M, Kim JS, Kim JY et al. Subregional patterns of preferential striatal dopamine transporter loss differ in Parkinson disease, progressive supranu- Schattauer 2014 Nuklearmedizin 6/2014

8 8 D. Kupitz et al.: Global scaling in FP-CIT SPECT clear palsy, and multiple-system atrophy. J Nucl Med 2012; 53: Seibyl JP, Marek KL, Quinlan D et al. Decreased single-photon emission computed tomographic [ 123 I]beta-CIT striatal uptake correlates with symptom severity in Parkinson s disease. Ann Neurol 1995; 38: Soderlund TA, Dickson JC, Prvulovich E et al. Value of semiquantitative analysis for clinical reporting of 123 I-2-beta-carbomethoxy-3beta- (4-iodophenyl)-N-(3-fluoropropyl)nortropane SPECT studies. J Nucl Med 2013; 54: Tatsch K, Poepperl G. Nigrostriatal dopamine terminal imaging with dopamine transporter SPECT: an update. J Nucl Med 2013; 54: Tatsch K, Poepperl G. Quantitative approaches to dopaminergic brain imaging. Q J Nucl Med Mol Imaging 2012; 56: Tossici-Bolt L, Hoffmann SM, Kemp PM et al. Quantification of [ 123 I]FP-CIT SPECT brain images: an accurate technique for measurement of the specific binding ratio. Eur J Nucl Med Mol Imaging 2006; 33: Van Laere K, Everaert L, Annemans L et al. The cost effectiveness of 123 I-FP-CIT SPECT imaging in patients with an uncertain clinical diagnosis of parkinsonism. Eur J Nucl Med Mol Imaging 2008; 35: Varrone A, Dickson JC, Tossici-Bolt L et al. European multicentre database of healthy controls for [ 123 I]FP-CIT SPECT (ENC-DAT): age-related effects, gender differences and evaluation of different methods of analysis. Eur J Nucl Med Mol Imaging 2013; 40: Verhoeff NP, Kapucu O, Sokole-Busemann E et al. Estimation of dopamine D2 receptor binding potential in the striatum with iodine-123-ibzm SPECT: technical and interobserver variability. J Nucl Med 1993; 34: Walker Z, Costa DC, Walker RW et al. Differentiation of dementia with Lewy bodies from Alzheimer s disease using a dopaminergic presynaptic ligand. J Neurol Neurosurg Psychiat 2002; 73: Ziebell M, Holm-Hansen S, Thomsen G et al. Serotonin transporters in dopamine transporter imaging: a head-to-head comparison of dopamine transporter SPECT radioligands 123 I-FP-CIT and 123 I-PE2I. J Nucl Med 2010; 51: Zubal IG, Early M, Yuan O et al. Optimized, automated striatal uptake analysis applied to SPECT brain scans of Parkinson s disease patients. J Nucl Med 2007; 48: Nuklearmedizin 6/2014 Schattauer 2014