[123I]FP-CIT ENC-DAT normal database: the impact of the reconstruction and quantification methods

Abstract

Background: [¹²³I]FP-CIT is a well-established radiotracer for the diagnosis of dopaminergic degenerative disorders. The European Normal Control Database of DaTSCAN (ENC-DAT) of healthy controls has provided age and gender-specific reference values for the [¹²³I]FP-CIT specific binding ratio (SBR) under optimised protocols for image acquisition and processing. Simpler reconstruction methods, however, are in use in many hospitals, often without implementation of attenuation and scatter corrections. This study investigates the impact on the reference values of simpler approaches using two quantifications methods, BRASS and Southampton, and explores the performance of the striatal phantom calibration in their harmonisation.

Results: BRASS and Southampton databases comprising 123 ENC-DAT subjects, from gamma cameras with parallel collimators, were reconstructed using filtered back projection (FBP) and iterative reconstruction OSEM without corrections (IRNC) and compared against the recommended OSEM with corrections for attenuation and scatter and septal penetration (ACSC), before and after applying phantom calibration. Differences between databases were quantified using the percentage difference of their SBR in the dopamine transporter-rich striatum, with their significance determined by the paired t test with Bonferroni correction. Attenuation and scatter losses, measured from the percentage difference between IRNC and ACSC databases, were of the order of 47% for both BRASS and Southampton quantifications. Phantom corrections were able to recover most of these losses, but the SBRs remained significantly lower than the "true" values (p < 0.001). Calibration provided, in fact, "first order" camera-dependent corrections, but could not include "second order" subject-dependent effects, such as septal penetration from extra-cranial activity. As for the ACSC databases, phantom calibration was instrumental in compensating for partial volume losses in BRASS (~67%, p < 0.001), while for the Southampton method, inherently free from them, it brought no significant changes and solely corrected for residual inter-camera variability (-0.2%, p = 0.44).

Conclusions: The ENC-DAT reference values are significantly dependent on the reconstruction and quantification methods and phantom calibration, while reducing the major part of their differences, is unable to fully harmonize them. Clinical use of any normal database, therefore, requires consistency with the processing methodology. Caution must be exercised when comparing data from different centres, recognising that the SBR may represent an "index" rather than a "true" value.

Keywords: 123I; Calibration; FP-CIT; Quantification; Reconstruction; SPECT; Specific binding ratio.

Fig. 1

The two methods used for measuring the striatal specific binding ratio (SBR), defined as the ratio of specific to non-specific striatal count concentrations, SBR = c _s /c _ns. Top: BRASS quantification method [4]. c _s and c _ns are measured from count concentrations using anatomical VOIs for the sub-striatal structures (caudate and putamen) and the occipital lobes, respectively. The striatal SBR used in this work was obtained by dividing the total counts from these two VOIs by their combined volume. The small volumes of these structures render these concentration measurements susceptible to partial volume losses. Bottom: Southampton quantification method [5]. c _s is derived from a measure of total counts in a geometrical VOI for the striatum. The generous dimensions of this VOI ensure that all counts related to striatal binding are captured, including those detected outside the anatomical boundary, thus averting under-estimations due to partial volume losses. c _ns is also measured from a large VOI, encompassing the whole cortex with the exception of the striata and excluding the outer rim beset by peripheral partial volume losses

Fig. 2

Examples of FBP and iterative reconstructions for a phantom study with highest filling ratio (Left=10:1, Right=8:1,top row) and a human control (bottom row), both acquired on an Infinia Hawkeye camera and reconstructed on the Link Medical workstation. Each image represents one (1 pixel-thick) central slice and is normalised to its own maximum

Fig. 3

ENC-DAT database of normal controls, BRASS quantification. Striatal specific binding ratios (SBR) vs age derived from various reconstructions: FBP (red), IRNC (blue) and ACSC (black), before (top row) and after (bottom row) phantom calibration. Their respective linear fit and the 95% CI (two standard error of the regression) are also shown following the same colour code

Fig. 4

ENC-DAT database of normal controls, Southampton quantification. Striatal specific binding ratios (SBR) vs age derived from various reconstructions, following the same conventions as in Fig. 2: FBP (red), IRNC (blue) and ACSC (black), before (top row), and after (bottom row) phantom calibration. Note the wider y-axis range compared to Fig. 2