This increased sensitivity will improve image quality or allow lower radiation dose to the patient. But locating PET detectors near the breast also introduces parallax error, which degrades spatial resolution near the edge of the field-of-view. One way to alleviate this problem is to measure the precise position along the detector crystal length at which the annihilation photon was absorbed, namely the depth-of-interaction (DOI).
At the University of California, Davis, researchers are developing a dedicated breast PET/CT scanner, called the DbPET. In their latest work, they propose and test a novel high-resolution DOI detector for this scanner. Based on a 14 x 14 array of 1.5 x 1.5 x 20 mm3 lutetium orthosilicate (LSO) crystals, the detector uses dual-ended readout, with a large-active-area avalanche photodiode (APD) on the patient end and a position-sensitive photomultiplier tube (PSPMT) on the other.
This hybrid design combines the high timing resolution of the PMT detector with the improved DOI resolution offered by dual-ended DOI encoding. Meanwhile, the compact nature of the APD allows dense detector packing at the patient end, enabling a ring diameter of 20 cm with no gaps between the detector front faces (Phys. Med. Biol. 57 3435).
"The measured DOI information is used to reduce parallax error, resulting in a relatively uniform image spatial resolution across the field-of-view," explained lead author Felipe Godinez. "In breast imaging, this is doubly important since the breast can fill the entire scanner’s field-of-view."
Detector analysis
Godinez and colleagues assembled two detector modules and performed a series of characterization tests. Flood histograms revealed that all 196 crystals could be identified at all depths, except for the one nearest to the APD where the edge and corner crystals were not fully delineated due to decreased light collection. The average timing resolution for the array, measured by acquiring coincidence data from detector modules on either side of a positron-emitting source, was 2.3±0.02 ns.
Next, the researchers measured the DOI resolution, using a single detector module irradiated by a 0.6 mm wide fan-beam of annihilation photons at five depths (2, 6, 10, 14 and 18 mm from the PSPMT-array boundary). The DOI resolution averaged over all depths and crystals was 2.9 mm. The best resolution was seen at the ends of the array nearest to the detectors.
At each sampled depth, the inner 12 x 12 crystals exhibited the most uniform DOI resolution. The DOI resolution averaged across the inner crystals, and all depths, was 2.4 mm. The edge crystals had a poorer average DOI resolution (4.0 mm), while the corner crystals were slightly worse (4.8 mm), attributed to the APD only partially sampling these crystals.
This set-up was also used to measure the energy resolution at each depth for every crystal. This is an important parameter in breast PET, since only a small fraction of the injected activity accumulates in the breast lesion, while the majority lies outside the field-of-view. Good energy resolution allows use of a narrower energy window, which can reduce contributions from random coincidence events.
Averaged across all depths, the energy resolution was 19.1% for the inner crystals, 28.2% for the edge crystals and 32.8 % for the corner crystals (21.6% averaged over all crystals). For a central crystal, the individual energy resolution ranged from 23.2–26.1% for the PSPMT (depending on depth) and from 21.8–32.0% for the APD.
The authors note that this represents an improvement upon UC Davis' current DbPET system, which has an energy resolution ranging from 18–40% and an average of 25%. They also observed that the energy resolution was optimal in the centre and progressively degraded towards either photodetector.
DOI dependence
Finally, the team measured the detectors' intrinsic spatial resolution. They used two detector blocks in three orientations: face-to-face with gamma rays incident at 90°, and off-centre with incident rays at 26° and 51°. At the centre of the field-of-view, the mean intrinsic spatial resolution was 1.2 mm, with a range of 1.1–1.3 mm; this represents the best value possible for the entire system.
Without DOI correction, the intrinsic spatial resolution for a single central crystal at 26° was 4.3 mm. At 51°, where parallax error is more pronounced, it was 6.4 mm. When DOI information was included, these values improved to 1.5 and 1.7 mm, at 26° and 51°, respectively.
The researchers conclude that the hybrid detector design will improve spatial resolution and uniformity across the field-of-view compared with their current prototype breast PET scanner. "We are already building a scanner based on these detectors. The National Institutes of Health has funded this effort," Godinez told medicalphysicsweb.
"Our group is currently working on the mechanical design of the DbPET using the new detectors, and on fast and robust electronic strategies to read out the large number of signal channels arising from these detectors," he added. "At the same time, we're continuing to explore DOI detectors with even higher resolution capabilities."
• Related articles in PMB
Characterization of a high-resolution hybrid DOI detector for a dedicated breast PET/CT scanner
Felipe Godinez et al Phys. Med. Biol. 57 3435
Tapered LSO arrays for small animal PET
Yongfeng Yang et al Phys. Med. Biol. 56 139
PET characteristics of a dedicated breast PET/CT scanner prototype
Yibao Wu et al Phys. Med. Biol. 54 4273