Search

The first stage of testing towards the further development of a full-ring SPECT/MR system has been evaluated. Performance of the cadmium zinc telluride (CZT) detectors were found to be unaffected by the static magnetic field, however they were found to react unreliably during simultaneous imaging when the MR sequence repetition time (TR) was less than 400ms and were influenced significantly by the imaging gradients. In an effort to better understand this phenomenon, component testing was performed by subjecting the CZT detector modules and electronics to various gradient sequences in a stand-alone gradient testing apparatus. We exposed the modules to 31 unique gradient sequence shapes of varying power and duration, typical of those used in clinical MR devices while simultaneously operating the CZT modules in event readout mode. Results indicate that certain detector pixels were preferentially affected (typically around the periphery of the detector array), leading to induced currents in the detector material and significant increases in erroneous event rates during the gradient ramp-up and ramp-down phases. The shape of the gradient waveform was also seen to have an effect on the detector performance with higher slew rates leading to impaired performance. Hardware disabling of the most severely affected pixels led to some reduction in erroneous background events, however, it also resulted in reduced detector resolution and sensitivity. Subsequent work led to the development of a gradient triggering circuit to momentarily suspend CZT event detection during the MR gradient ramp up and ramp down intervals. Subsequent testing of this circuit in a clinical 3T MR was performed. When gradient triggering was enabled, erroneous event data was reduced to near zero levels and resultant images show no structural differences when compared to control acquisitions. During gradient application, the CZT modules are suspended, thus yielding a 60% drop in sensitivity due to the additional deadtime of the triggering. The produced MR image showed a reduction in SNR of 92%. Co-registration of the MR and gamma camera images was successfully performed, showing both the structural and activity makeup of the phantom.

Machine learning (ML) is increasingly used in medical imaging. This paper provides pilot data of decision trees and random forests (RFs) to predict if a 18F-florbetapir brain positron emission tomography (PET) is positive or negative for amyloid deposition based on quantitative data analysis. The dataset included 55 18F-florbetapir brain PETs in participants with severe white matter disease and mild cognitive impairment (MCI), early Alzheimer's disease (AD) or transient ischemic events. The Montreal Cognitive Assessment (MoCA) score was known for each participant. All PET images were processed using the MINC toolkit to extract standardized uptake value ratios (SUVRs) for 59 regions of interest (features). Each PET was clinically read by 2 dual certified radiology/nuclear medicine physicians with final interpretation based on consensus. An initial study of RFs using conventional binary decision trees and PET quantitation suggests this is a powerful algorithm for PET classification as positive or negative for amyloid deposition. Preliminary data did not show improved results when a ternary RF approach was used. Finally, a soft-decision approach may be helpful to predict the $\mathbf{MoCA}$ score.

There has been significant interest in the use of semiconductor detectors such as cadmium zinc telluride (CZT) for SPECT/MRI and PET/MRI due to their ability to operate in high magnetic fields. However, the effect of these materials on magnetic field homogeneity must be considered when designing an MR-compatible gamma camera. Field maps were generated for several materials considered for an MR-compatible gamma camera to determine the extent of the shift in magnetic field. The materials used were CZT, Al, carbon fiber, printed circuit boards, thermoelectric cooler, Pb and a composite material consisting largely of tungsten. A lipid phantom filled with a mixture of 50% canola oil and 50% water (∼9 cm × ∼9 cm × 6 cm) was imaged with and without the materials placed adjacent to the phantom/foot ankle coil. Data were acquired with the materials individually and as a combined system with and without power. MRI phase images, at TE = 5 and 8 ms (using a 3T MRI), were used to calculate magnetic field (B0) homogeneity. Tungsten and the thermoelectric cooler resulted in a significant B0 shift near the boundary of the phantom where the materials were placed, whereas the aluminum had the least effect on the field homogeneity. Based on a linear approximation, the material with the largest effect on the magnetic field, tungsten, would need to be placed approximately 0.6 cm outside of the RF receiver coil, making it 3.4 cm in total from the phantom to have negligible effect on the field homogeneity. There was negligible effect on magnetic field homogeneity when high or low voltage power were supplied to the CZT detector system.

Research on microwave imaging for medical diagnostics has recently expanded significantly. There is substantial need for physical phantoms in order to conduct experiments. These phantoms should mimic the electrical properties of the target tissues. They must also have certain mechanical properties. This paper proposes recipes for breast tissue and malignant tumor phantoms that have been tested and used in experiments for over two years now. These phantoms mimic breast tissues in the ultra-wideband (UWB) frequency range from 3 GHz to 10 GHz. The recipes presented here use non-biological materials.

A new texture feature-based seeded region growing algorithm is proposed for automated segmentation of organs in abdominal MR images. 2D Co-occurrence texture feature, Gabor texture feature, and both 2D and 3D Semi- variogram texture features are extracted from the image and a seeded region growing algorithm is run on these feature spaces. With a given Region of Interest (ROI), a seed point is automatically se-lected based on three homogeneity criteria. A threshold is then obtained by taking a lower value just before the one causing ‘explosion’. This algorithm is tested on 12 series of 3D ab-dominal MR images.