SmartPhantom--an fMRI simulator

Abstract

Many informatics tools have emerged to process the voluminous and complex data generated by functional magnetic resonance imaging (fMRI). The interpretation of fMRI exams is largely determined by these tools. However, their performance is hard to evaluate because there is no independent means of calibration. A novel fMRI calibration system called SmartPhantom has been developed to simulate functional blood oxygen level dependent (BOLD) imaging. SmartPhantom contains a quadrature radio frequency coil, comprising two perpendicular planar loops that can be externally activated or deactivated. The system is able to produce reasonably uniform signal enhancements in a calibration sample surrounded by the two loops during an MRI scan. The enhancement is controlled well in both magnitude and predefined timing and produces BOLD-like signals. Characteristics of SmartPhantom are discussed in detail, followed by a comparison of fMRI informatics tools. Two fMRI data sets are acquired with the SmartPhantom. One with high signal-to-noise ratio provides the calibration. Another with lower SNR is input into three software packages (BrainVoyager, FSL and Statistical Parametric Mapping 2) for data preprocessing and statistical analysis. Results from the three packages are compared in both sensitivity of detecting the activation and correlation between the predicted activation and calibration.

Figure 1

Simulation results of the signal from enhancement coils. a) Signal difference at the central axial slice; b) signal difference along the middle line across the image shown in a, only pixel number from 28 to 38 is shown; c) signal difference at the central sagittal slice; d) signal difference along the middle line across the image shown in c, only pixel number from 28 to 38 is shown.

Figure 2

Schematic circuit designs for SmartPhantom control with a serial resistor array circuit.

Figure 3

a) SmartPhantom inserted into head coil; b) Axial MR image of SmartPhantom (FOV 24 cm); c) Sagittal MR image of SmartPhantom (FOV 24 cm).

Figure 4

Enhancement as a function of resistance, curve fitted to equation 9.

Figure 5

Enhancement levels for three adjacent slices as a function of circuit resistance. The slice thickness is 4 mm and space between the slices is 1 mm.

Figure 6

BOLD images and time course characteristics. a) Mean ‘off’ image of 90° flip angle; b) Mean difference image of 90° flip angle (average ‘on’ images minus average ‘off’ images); c) Time course of the average signal in the ROI outlined in a).

Figure 7

t-maps and difference map. a) BOLD difference signal; b) t-map from BrainVoyager; c) t-map from FSL; d) t-map from SPM.

Figure 8

Sensitivities of detecting activation under different t-thresholds for Brainvoyager (BV), FSL and SPM2 (SPM). Activation is defined as enhancement between 0.8% and 1.2%.

Figure 9

a) Correlation between calibration and t-map from GLM analysis of BrainVoyager (green), FSL (red) and SPM (blue). Correlation is plotted as a function of the threshold of percentage of the maximum contrast. Only those pixels with enhancement (positive or negative) less than the threshold are included in the correlation computation. b) Difference of the correlation curve between BrainVoyager and SPM (red), BrainVoyager and FSL (green), and FSL and SPM (blue).