Yoshiyuki Yamakawa

Performance Evaluation of a New Dedicated Breast PET Scanner Using NEMA NU4-2008 Standards

The aim of this work was to evaluate the performance characteristics of a newly developed dedicated breast PET scanner, according to National Electrical Manufacturers Association (NEMA) NU 4-2008 standards. Methods: The dedicated breast PET scanner consists of 4 layers of a 32 × 32 lutetium oxyorthosilicate–based crystal array, a light guide, and a 64-channel position-sensitive photomultiplier tube. The size of a crystal element is 1.44 × 1.44 × 4.5 mm. The detector ring has a large solid angle with a 185-mm aperture and an axial coverage of 155.5 mm. The energy windows at depth of interaction for the first and second layers are 400–800 keV, and those at the third and fourth layers are 100–800 keV. A fixed timing window of 4.5 ns was used for all acquisitions. Spatial resolution, sensitivity, counting rate capabilities, and image quality were evaluated in accordance with NEMA NU 4-2008 standards. Human imaging was performed in addition to the evaluation. Results: Radial, tangential, and axial spatial resolution measured as minimal full width at half maximum approached 1.6, 1.7, and 2.0 mm, respectively, for filtered backprojection reconstruction and 0.8, 0.8, and 0.8 mm, respectively, for dynamic row-action maximum-likelihood algorithm reconstruction. The peak absolute sensitivity of the system was 11.2%. Scatter fraction at the same acquisition settings was 30.1% for the rat-sized phantom. Peak noise-equivalent counting rate and peak true rate for the ratlike phantom was 374 kcps at 25 MBq and 603 kcps at 31 MBq, respectively. In the image-quality phantom study, recovery coefficients and uniformity were 0.04–0.82 and 1.9%, respectively, for standard reconstruction mode and 0.09–0.97 and 4.5%, respectively, for enhanced-resolution mode. Human imaging provided high-contrast images with restricted background noise for standard reconstruction mode and high-resolution images for enhanced-resolution mode. Conclusion: The dedicated breast PET scanner has excellent spatial resolution and high sensitivity. The performance of the dedicated breast PET scanner is considered to be reasonable enough to support its use in breast cancer imaging.

Reconstruction of 4-Layer DOI detector equipped C-shaped PEM via list-mode iterative algorithm

This C-shaped PEM unit currently under development is designed for the very early stage detection of very small lesions, such as breast cancers. This PEM units scanner is equipped with 4-layer Depth of Interaction (DOI) detectors, each of which contains small scintillator crystals. The scanner is positioned closely around the breast, providing both high resolution and sensitivity. The open end of the detector unit allows the arm to be placed there and the C-shaped design of the scanner accommodates a variety of patient physiques, ensuring inclusion of the entire breast into the effective field of view (FOV). In this work, list-mode iterative algorithms can be utilized for image reconstruction using incomplete acquisition datasets, due to this scanners module gap and large number of LORs. To evaluate the effect of using DOI detectors on C-shaped PEM images, and the resulting effective FOV of this PEM, Monte- Carlo simulations of the acquisitions from this PEM scanner were used for image reconstruction via list-mode DRAMA algorithms. Results indicate that while non-DOI detector contrast near the detector gap deteriorates substantially, use of a DOI detector system preserves a high level of contrast.

Development of a C-shaped breast PET scanner equipped with four-layer DOI detectors

For diagnosis of very small lesions of breast cancer on very early stage, a dedicated breast positron emission tomography (PET) scanner consisting of four-layer depth of interaction (DOI) detectors is now under development. We are aiming for the spatial resolution of less than 1 mm in this scanner and acquisition time is less than 5 minutes by one breast and 10 minutes in total. The “C” shape of this scanner allows it to be positioned closely around the breast, effectively increasing both resolution and sensitivity. The open end of the detector unit allows the arm to be placed there and the C-shaped design of the scanner accommodates a variety of patient physiques, ensuring inclusion of the entire breast into the effective field of view (FOV).

Development of a prototype DOI-TOF-PET scanner

A prototype depth-of-interaction and time-of-flight positron emission tomography (DOI-TOF-PET) scanner was developed to offer enhanced spatial resolution with high sensitivity and high signal-to-noise-ratio (SNR) in the reconstructed images. The detector ring is 775 mm diameter with 48 mm axial field-of-view (FOV) per ring. The system can be expanded up to three rings. The ring comprises 48 detector modules, each of which consists of four layers of 16 × 16 crystal elements and a 64ch PS-PMT (H8500 position sensitive photomultiplier tube, Hamamatsu Photonics K.K.) optically coupled with silicone resin. The size of the crystal elements are 2.9 mm × 2.9 mm and increase in depth through 5, 6, 7, and 8 mm, from the first to fourth layer, to reduce the sensitivity differences between each layer. The crystal material is used Lu2xGd2(1−x)SiO5 (Hitachi Chemical Co., Ltd.) because of its short decay time, high density and high light yield. A data acquisition board was developed to improve the spatial resolution and the timing resolution of the system. The TDC (time-to-digital Converter) chips (TDC-GPX, Acam Messelectronic) mounted on the board operate in high-resolution mode (R-mode, 27 ps/bin). In addition, a new timing correction method to correct the intrinsic timing difference both each detector and each crystal of the detector by using DOI information was developed. As a result, the average timing resolution of this system was 442 ps (FWHM). Reconstructed image quality with-/without-DOI-TOF technique was evaluated in GATE simulation and a preliminary iamge was obtained with the prototype system.

Random correction using singles count rates for DOI Positron Emission Mammography

With the aim of realizing the early diagnosis of breast cancer, we are promoting the research and development of a high-resolution Positron Emission Mammography (PEM). In order to obtain both high resolution and high sensitivity, this scanner uses for-layer DOI (depth of interaction) detectors consisting of small crystals being positioned close to the breast Since the total number of crystals in the scanner is extremely large and it is predicted that the total number of LORs is greater than the total number of events, a list-mode iterative image reconstruction method is required. Furthermore, a random correction method based on singles counting is beneficial for reducing statistical noise in randoms estimation at each LOR. However, an extremely large number of counters are required to count single events for each crystal. We also developed a random correction method calculating the count for each crystal. This method makes it possible to reduce the number of counters. We evaluated the efficacy of this technique incorporated in a list-mode iterative reconstruction using simulation data. Results demonstrated that random correction using conventional delayed coincidence method does not work well for the list-mode iterative reconstruction having an extremely large number of LORs, whereas random correction using singles count rates is effective for improving the image quality.

Attenuation correction using level set method for application specific PET scanners

Attenuation correction is essential to obtain quantitative positron emission tomography (PET) images. Typically, attenuation map is calculated from corresponding X-ray CT images or by using transmission scan data acquired with an external radionuclide source. However, in a PET scanner dedicated for a specific organ, such as brain or breast, surrounded closely by radiation detector modules, it is difficult to implement transmission scan mechanism. Although some techniques have been developed to calculate uniform attenuation maps from emission sinograms or images based on intensity thresholding, it is generally difficult to decide an optimal threshold value especially due to complicated target shapes, nonuniform activity distribution, or large statistical noise. In this paper, we propose a new method to generate the attenuation map from the attenuation-uncorrected reconstructed image using the level set method, which is one of the active contour models to solve the problem of image segmentation. We evaluated the accuracy of attenuation maps generated from various patient images with our dedicated breast PET scanner. Experimental results showed that the proposed method can provide more accurate and robust attenuation map in comparison with the conventional thresholding method.

Development of a Digital Baseline Restorer for high-resolution PET detectors

For a data processing circuit using Anger positioning method on a high-resolution PET detector, it is difficult to measure precisely the position of the interaction, because a signal baseline fluctuates by causes such as temperature dependence of the detector sensitivity and the circuit drift, pulse pileup at high count rates, and scintillator afterglow. Particularly in the case of the number of scintillator elements much greater than the number of photo detector elements, both high signal precision and stability are required because the positioning error is caused by a slight baseline change. Usually AC-coupled circuit is utilized to reduce the influence of the baseline change, but may not work at high count rates. Then we adopted a DC coupling and have developed a DBLR (Digital Baseline Restorer) method to correct the baseline shift in real time. The amount of baseline shift is periodically estimated during the acquisition based on the shrinkage of a 2D position histogram map of the detector. The ratios of the total count of the outermost region and the inner region for each direction of the 2D map are fed back to correct the position calculation on FPGA. We evaluated the proposed method with a four-layer DOI (depth of interaction) detector that consists of 4,096 LGSO crystal elements and a 64-channel PS-PMT (position sensitive photomultiplier tube). The DBLR can effectively correct the baseline shift and reduce the map shrinkage to less than 0.3% even at a high count rate of 243 kcps.