RT Journal Article SR Electronic T1 Impact of reference databases on evaluation of 123-I-FP-CIT striatal binding: a comparison between Hermes EARL-BRASS and GE DaTQuant JF Journal of Nuclear Medicine JO J Nucl Med FD Society of Nuclear Medicine SP 70 OP 70 VO 61 IS supplement 1 A1 Hans-Georg Buchholz A1 Florian Rosar A1 Manuela Hoffmann A1 Mathias Schreckenberger YR 2020 UL http://jnm.snmjournals.org/content/61/supplement_1/70.abstract AB 70Aim: Dopamine Transporter SPECT with 123-I-FP-CIT is widely used to detect nigrostriatal dopaminergic neurodegeneration in patients with parkinsonian syndromes. In addition to the subjective visual rating, a semiquantitative analysis which calculates specific binding ratios (SBR) is used. Moreover, an automated single subject analysis with comparison to an implemented reference database can be performed in order to classify into normal or abnormal binding. In this study, two software packages, the Hermes EARL-BRASS (BR) and the GE DaTQuant (DQ), were investigated. Aim was to assess differences between BRASS and DaTQuant results when using the same 123-I-FP-CIT raw datasets. A second goal was to test the performance of BRASS and DaTQuant after exchanging their normal databases with an own camera-specific reference database. Methods: Retrospectively, 123-I-FP-CIT studies performed on Philips IRIX SPECT camera were randomly selected from clinical routine. Based on the visual inspection as well as on the semiquantitative VOI analysis subjects were classified as normal or abnormal. In equivocal cases the clinical reports on outcome were used for classification. In the first step 30 patients were analyzed: SBR of nucleus caudatus, anterior and posterior putamen were calculated with both, BRASS and DaTQuant, using vendor recommended settings for reconstruction and analysis (i.e. VOI template for quantitation, age-correction and implemented reference database). For each VOI the deviation to the reference database was calculated and expressed as z-score (= difference between patients’ SBR and reference SBR related to the standard deviation of the reference SBR). If z-score was < -2 at least in one region the subject was classified as abnormal. In the second step own camera-specific normal databases for the use in BRASS and in DaTQuant were constructed using 56 datasets of as normal classified 123-I-FP-CIT studies. Analyses of the same 30 datasets were repeated in BRASS and DaTQuant using own databases. Regional z-scores of the two BRASS and DaTQuant evaluations were compared and concordance of classification in normal/abnormal binding was assessed. Results: Calculated SBR of the 30 studies with DaTQuant were systematically higher than with BRASS (17.4% in nucleus caudatus and 6.5% in putamen). Performance of BRASS and DaTQuant normal databases was very different: mean z-score (=SBR deviation of all regions) was -0.97 in DaTQuant compared to -2.77 in BRASS. A different classification (normal vs. abnormal) with BRASS and DaTQuant was found in 10 cases. Using own camera-specific reference databases in BRASS and DaTQuant led to better comparability of the results: the mean z-score of all regions now was -2.5 (DQ) and -2.7 (BR). Between BR and DQ, the strongest correlation of z-scores was found in the posterior putamen VOI (r=0.94). Classifying with BR and DQ showed better concordance (27 of 30 cases). Conclusions: The performance of EARL-BRASS and DaTQuant can be substantially improved using own camera-specific reference databases. Differences in image reconstruction, VOI template and age-correction had only a minor effect on classification with BRASS and DaTQuant.