Development of a PET/EPRI combined imaging system for assessing tumor hypoxia
Heejong Kim, Boris Epel, Subramanian Sundramoorthy, Hsiu-Ming Tsai, Eugene Barth, Inna Gertsenshteyn, Howard Halpern, Yuexuan Hua, Qingguo Xie, Chin-Tu Chen, Chien-Min Kao
aa r X i v : . [ phy s i c s . m e d - ph ] J a n Prepared for submission to JINST
Development of a PET/EPRI combined imaging system forassessing tumor hypoxia
H. Kim, š B. Epel š,š
S. Sundramoorthy š,š
H.-M. Tsai š E. Barth š,š
I. Gertsenshteyn š,š,š
H.Halpern š,š
Y. Hua š Q. Xie š C.-T. Chen š C.-M. Kao š š Department of Radiology, University of Chicago, Chicago, IL 60637 š Department of Radiation and Cellular Oncology, University of Chicago, Chicago, IL 60637 š Center for EPR Imaging In Vivo Physiology, University of Chicago, Chicago, IL 60637 š Raycan Technology Co, Ltd., Suzhou, Jiangsu, China š Huazhong University of Science and Technology, Biomedical Engineering Department, Wuhan, Hubei,China
E-mail: [email protected]
Abstract:
Precise quantitative delineation of tumor hypoxia is essential in radiation therapy treat-ment planning to improve the treatment eļ¬cacy by targeting hypoxic sub-volumes. We developeda combined imaging system of positron emission tomography (PET) and electron para-magneticresonance imaging (EPRI) of molecular oxygen to investigate the accuracy of PET imaging inassessing tumor hypoxia. The PET/EPRI combined imaging system aims to use EPRI to preciselymeasure the oxygen partial pressure in tissues. This will evaluate the validity of PET hypoxic tumorimaging by (near) simultaneously acquired EPRI as ground truth. The combined imaging systemwas constructed by integrating a small animal PET scanner (inner ring diameter 62 mm and axialļ¬eld of view 25.6 mm) and an EPRI subsystem (ļ¬eld strength 25 mT and resonant frequency 700MHz). The compatibility between the PET and EPRI subsystems were tested with both phantomand animal imaging. Hypoxic imaging on a tumor mouse model using F-ļ¬uoromisonidazoleradio-tracer was conducted with the developed PET/EPRI system. We report the development andinitial imaging results obtained from the PET/EPRI combined imaging system.
Keywords:
Multi-modality systems, Positron Emission Tomography, Electron-Paramagnetic Res-onance Imaging, Tumor Hypoxia Corresponding author. ontents : animal imaging 6 Because of the rapid proliferation of cancer cells outgrowing blood supply, the amount of oxygenavailable is limited in some tumor regions distant from blood vessels. Tumor hypoxia [1, 2] refersto the lower oxygen concentration in tumor tissue, typically below 10 torr, and it is a characteristicfeature of human and animal solid tumors. Hypoxia is known to promote tumor progression andmigration by changing its metabolism to adapt to oxygen deprived environment [3, 4]. It has beenknown that hypoxic tumors are more resistant to radiation therapy [5ā7], and require additionaldose delivery for eļ¬ective treatment. Therefore, precise delineation of hypoxic tumor region isessential for targeting only hypoxic sub-volume within the tumor in radiation therapy to improvethe treatment outcome [8ā11].In the clinic, PET imaging with radio-tracers such as F-ļ¬uoromisonidazole (F-MISO) [12, 13]has been used for hypoxic tumor imaging, which exploits the mechanism of misonidazole bindingintracellularly at oxygen-depleted sites 2-4 hours post-injection. However, F-MISO PET imaginghas not yet beneļ¬cially guided radiation therapy treatment for hypoxic tumors. For example, arecent phase-II clinical study [14] showed that dose painting based on F-MISO PET imaging didnot signiļ¬cantly improve the treatment outcomes.Currently, the F-MISO uptake and binding mechanisms in hypoxic tumor are not clearlyunderstood [15]. Other factors, such as the vascular structure and pH in the micro-environment,have been observed to aļ¬ect the F-MISO absorption within a hypoxic region [16, 17]. These factorssubstantially complicate the correlation between the degree of hypoxia and F-MISO uptake. On theother hand, in preclinical studies, the eļ¬ectiveness of oxygen imaging by electron para-magneticā 1 āesonance imaging (EPRI) for targeting hypoxic tumors has been demonstrated [11, 18, 19]. EPRIis a non-invasive imaging technology based on the detection of unpaired electron spins of injectablespin probe subjected to a magnetic ļ¬eld, and is capable of quantitative measurements of the absolutepartial pressure of oxygen (pO ) in tissue [20ā22]. However, EPRI is not available for clinical usepresently.We recently initiated a project to investigate the potential of PET imaging in assessing tumorhypoxia in small-animal settings. In the project, we developed a combined small-animal PET andEPRI imaging system to allow simultaneous or rapid succession imaging so that we can use EPRoxygen imaging as a gold standard for in vivo measurement of tissue pO to evaluate the accuracyof PET hypoxia imaging, and to develop correction methods if necessary. The (near) simultaneousPET/EPR operation ensures the recording of the same physiological and biochemical changes in thetissue, and accordingly reļ¬ects the correlations between PET and oxygen images more accurately.PET-EPR image registration is not required with the combined system which is not the case forseparate PET and EPR scanners.In this article, we present the development of the combined PET/EPRI system and the initialimaging results obtained by using the system. Figure 1 . (a) The PET detector before integration. (b) A loop-gap type EPRI RF resonator. A phantom withcapillary tubes is positioned at the center of the resonator for imaging. (c) A PET/EPRI combined system.The PET detector is installed between two permanent EPRI magnets.
The PET subsystem was originally developed as an insert for a small animal imaging MRI. It consistsof 14 detector modules, which are installed within a cylindrical supporting structure that has a 60mm inner diameter and a 115 mm outer diameter. Each detector module uses an array of 8x4 LYSOscintillators (each LYSO crystal is 3x3x10 mm ) and two Hamamatsu MPPC arrays (S13361-3050-NE or S12642-0404-PA). The scintillators within the array are optically isolated by using enhancedspecular reļ¬ector (ESR, 3M š š ), and are coupled individually to SiPMs (3.2 mm pitch) in the MPPCarray. The axial ļ¬eld of view of the PET detector is 25.6 mm. A strip-line based multiplexingmethod [23] is used for SiPM signal readout. An earlier prototype detector module [23, 24] used 8SiPMs of 4.0x4.4 mm active area (SPM42-75, STM) on a strip-line. In comparison, the detectorā 2 āodule of the reported system used Hammamatsu MPPC arrays instead due to its smaller pixelsize and more uniform gain between SiPM pixels. It also has 16 SiPMs on a strip-line to achieve ahigher multiplexing ratio. The 32 SiPM outputs from a detector module are routed to two strip-linesimplemented in a strip-line board (SLB), which is a 18x110 mm printed circuit board. The SLBoutputs are connected to a multi-voltage threshold (MVT) waveform sampling board [25] placedaway from the MRI magnetic and radio-frequency (RF) ļ¬elds via 5 m long miniature coaxial cablesfor signal digitization. The MVT board implements voltage-based sampling of PET signals asfollowing. The board provides 4 user-deļ¬nable voltage thresholds, which in this study are set to50, 150, 300 and 600 mV. The leading and trailing transitions of a PET signal over and below thesethresholds are obtained by comparators and the transition times are determined by time-to-digitalconverters (TDC) with 95 ps bin width. The known voltage thresholds and the TDC produced timestamps are packed along with the channel number to a data acquisition computer through an Ethernetinterface. The comparators and TDCs are all implemented by using ļ¬eld programmable gate arrays(FPGAs). The performance and MR compatibility of this prototypical PET insert was previouslycharacterized in experiments with a Bruker BioSpec 9.4 T MR scanner (Billerica MA). Figures 1(a)shows a photo of the PET insert. More details on the design and performance characterization ofthe PET insert are described in our previous publication[26]. The EPRI subsystem is a 700 ( Ā±
20) MHz pulse imager/relaxometer utilizing an inversion recoveryelectron spin echo (IRESE) acquisition method [22]. It consists of a parallel face 25 mT permanentmagnet with a 12 cm gap, 2 mT main ļ¬eld oļ¬set coil, three planar orthogonal gradient coils (30mT/m maximum gradient), an RF resonator and RF control electronics. The resonator is locatedinside the PET system while the oļ¬set coils are placed in between the PET system and the magnet.The RF resonator of the EPRI, shown in Figure 1(b), is a loop-gap resonator in reļ¬ection mode forboth exciting and detecting RF signals at 700 MHz. The resonator has a cylindrical 19 mm-diameteraccess that is 15mm in length. It is embedded in a plastic spacer for positioning at the isocenter ofthe gradient ļ¬elds and the PET system. The object to be imaged is placed within the resonator asshown in Figure 1(c). A 4 kW TOMCO RF ampliļ¬er is used to power RF pulses, though only 250W was used for our experiments. The homodyne bridge scheme and control method similar to [27]is used. The instrument is controlled using SpecMan4EPR software [28].
Figure 1(c) shows the combined PET/EPRI system encased in a 40x40x31 cm frame. Both thecenter of the resonator and that of the PET ļ¬eld of view are aligned with the center of the magnets.Data acquisition and processing of the PET and EPRI subsystems are accomplished separatelyby two computers. For EPRI, the IRESE imaging sequence was used with 201 projections anda maximum gradient of 7.5 mT/m. Nine time delays between the inversion and the electronspin echo detection were used to measure relaxation times. Image acquisition time was about 7minutes. The data were reconstructed using a ļ¬ltered back-projection (FBP) algorithm [29]. PETdata were acquired as list-mode data of singles events; post acquisition they were processed toproduce coincidence data that were then reconstructed by using a line-of-response based maximumlikelihood expectation maximization (MLEM) algorithm [30]. The PET image reconstruction codeā 3 āas previously validated in imaging studies of a Na point source, a Ge line source, and acustom-made resolution phantom [26].Alignment between the PET and EPRI system was assessed by imaging a phantom shown in Fig-ure 2(a). The phantom holds 5 capillary tubes; three tubes were ļ¬lled with F-ļ¬uorodeoxyglucose(FDG) for PET, and the other two tubes were ļ¬lled with 1 mM of trityl spin probe solution [31]for EPRI. Figure 2(b) shows a PET/EPRI combined image of the phantom. The PET image of the3 FDG-ļ¬lled capillary tubes is shown in greyscale, and EPR image of the 2 trityl-ļ¬lled capillarytubes is shown in color. Figure 2(c) shows an intensity proļ¬le of the PET image through the centerof the 3 capillary tubes. The PET and EPRI systems were previously evaluated to have an imageresolution of approximately 1.6 mm and 1.0 mm, respectively [26]. The intensity proļ¬les (EPRIintensity proļ¬les not shown) indicate that these resolutions are maintained by the combined system.
Figure 2 . (a) An imaging phantom with 3 capillary tubes ļ¬lled with FDG and 2 tubes ļ¬lled with tritylspin probe solution. (b) A PET/EPRI registered image of the imaging phantom. (c) A proļ¬le of the PETimage on the line through the centers of the 3 capillary tubes ļ¬lled with FDG. (d) A tumor-borne mouse legis immobilized in the animal bed using polyvinyl siloxane mold cast; the bed is placed in the resonator forimaging.
Small animal imaging followed the protocol approved by the University of Chicagoās InstitutionalAnimal Care and Use Committee. Squamous cell carcinoma tumor cells were grown in the leg ofa C3H mouse. The animal was positioned into an animal bed which was equipped with a set of 4non-parallel capillary tubes for registration purpose. The tumor area was secured using polyvinylsiloxane (PVS) dental mold material, as shown in Figures 2(d), and the mold was placed in theresonator hole. F-MISO was used for PET hypoxic imaging.A bolus of F-MISO was injected intravenously through a tail vein cannula. EPRI acquisitionbegan 1.5 hours post F-MISO injection. During EPRI data acquisition, the trityl spin probe wasinfused continuously through cannulation at a rate of 240 š l/h. Two or three 7-minute EPRIimages were taken in sequence during which list mode PET data were continuously acquired.This simultaneous imaging session lasted 20-30 minutes. For evaluation, an additional PET-onlyimaging was conducted beginning 2 hours post F-MISO injection to collect PET data free of theinterference of the EPRI pulse sequences. ā 4 ān addition to PET/EPRI imaging, MR T2-weighted image of the animal was also obtained foranatomical reference. MRI was done separately, after PET/EPRI imaging, by using a preclinicalBruker 9.4 T scanner (Billerica, MA) employing the Rapid Imaging with Refocused Echo (RARE)sequence. The four non-parallel capillary tubes placed around the animal bed were used as ļ¬ducialmarkers for registration of EPR and MR images. Each capillary tube contained 1 mM of trityl spinprobe solution that is detectable by both EPRI and MRI. Eļ¬ects on EPRI due to PET was found to be negligible from the comparison of the capillarytube phantom images obtained before and after installation of the PET subsystem. However, thePET subsystem, as the detector modules had no RF shielding, picked-up EPRI RF pulses duringsimultaneous PET/EPRI imaging. To investigate the eļ¬ect, we observed the singles count rate andthe pulse shape as determined by the MVT samples. Figure 3(a) shows that during RF pulsing thesingles count rate increases signiļ¬cantly, due to the detection of spurious events associated with theEPRI RF pulses. However, with MVT data acquisition, these spurious RF events can be eļ¬ectivelyrejected by oļ¬-line data processing.
Figure 3 . (a) PET singles event rate as the EPRI RF pulsing is turned on and oļ¬. Horizontal axis representssingles event time-stamp recorded in the MVT board. (b) A typical scintillation event consisting of 8 MVTsamples (blue circles) and the signal waveforms estimated from them by using a simple piecewise linearinterpolation-based method (dashed-line curve) and a bi-exponential ļ¬tting function (solid-line curve). (c)The waveform of an exemplary spurious RF event picked up by PET data acquisition. (d) Histogram of thefalling time measured from the data acquired during EPRI RF pulsing. (e) Singles event rate after applyingthe RF rejection. (f) Pulse-height histograms for a detector pixel before and after applying the RF rejection. ā 5 ās stated above, MVT sampling produces a leading and trailing sample at each voltage thresholdto yield 8 samples for each signal. Figures 3(b) and 3(c) show the signal waveforms estimated fromthese MVT samples (circles) for a real scintillation pulse and a spurious RF event, respectively.In this work, a simple piecewise linear interpolation-based method is used for estimating signalwaveform. The leading portion of the waveform between the 1st and the 4th samples is obtainedby linearly interpolating these samples, that before the 1st sample by extrapolating the interpolationline of the 1st and 2nd samples until it reaches zero, and that after the 4th samples by extrapolatingthe interpolation line between the 3rd and 4th samples until it meets the trailing portion of thewaveform, which itself is obtained from the 5th to 8th samples in a similar way. Evidently, thepoint where the leading and trailing portions meet deļ¬nes the peak of the waveform. Also shown inFigure 3(b) is the pulse waveform obtained by a bi-exponential ļ¬tting for LYSO/MPPC MVT datathat was previously proposed and validated in [25]. Although the piecewise linear interpolation-based method was not as accurate, it was much faster to compute and was found to yield adequateenergy resolution for the proposed system.It was observed that the spurious RF events have a shorter duration than the scintillation pulses.The histogram in Figure 3(d) shows that the falling time (calculated as the time diļ¬erence betweenthe 5th and 8th MVT samples) of the spurious RF events and actual scintillation events have wellseparated distributions, with the former being approximately 67 ns and the latter approximately130 ns. Consequently, events having falling time smaller than 95 ns were identiļ¬ed as spuriousRF events and rejected before energy qualiļ¬cation and coincidence ļ¬ltering. Figure 3(e) shows thesingles count rate after implementation of this rejection method. Figure 3(f) shows the pulse-heighthistograms of a detector pixel before and after applying the rejection method on the EPRI-on data.(Note that the y-axis uses logarithmic scale in Figure 3(f).) The pulse-height of single event wascalculated by integrating the area under the interpolated waveform.To check the eļ¬ectiveness of the proposed RF-event rejection method, a tumor in the leg ofa female mouse was imaged. 4.0 MBq FDG was injected to the female mouse (21.0 gram), andimaging started 35 minutes post-injection. Two PET data sets were obtained for comparison: onedata set was collected while EPRI was oļ¬ (EPRI-oļ¬ data), and the other set was acquired whileEPRI was on but processed by employing the RF-event rejection method (EPRI-on, RF-rejecteddata). Table 3.1 summarizes the acquisition time and the resulting numbers of coincidence events.The comparable numbers of coincidence events suggests the eļ¬ective removal of spurious RFevents by the proposed method. Figure 4 shows the PET images obtained from the two data sets,co-registered to the MRI T2-weighted image for visualization. The PET images obtained fromthe EPRI-on, RF-rejected data set do not show noticeable artifacts, and are similar to the imagesfrom the EPRI-oļ¬ data. Note that the data sets were acquired at approximately 15 minutes apart;the slight but observable diļ¬erences between the images possibly reļ¬ect temporal physiologicalchanges that occur between the imaging sessions. : animal imaging We have conducted 10 mice F-MISO imaging using the PET/EPRI combined system. A typicalimage is shown in Figure 5 showing a mouse (23.9 gram) with a leg-born SCCVII squamous cellcarcinoma immobilized in way similar to that shown in Figure 2(d). An MR T image was acquiredfor delineating the tumor region as an anatomical reference after PET/EPRI imaging. Figure 5(top)ā 6 āata set EPR On EPR On, RF-rejected EPR Oļ¬DAQ time (minutes) 15.5 15.5 14.5Coincidence events 17.4 M 0.71 M 0.68 M Table 1 . Comparison of the data acquisition time and the number of coincidence event from the two datasets.
Figure 4 . (top) A MR T2-weighted image of the tumor in a mouse are acquired for anatomical reference.(middle) PET image obtained from the EPRI-oļ¬ data are registered to the MR image. (bottom) PET imagefor the same slice obtained from the EPRI-on and RF-rejected data. shows the tumor boundary marked on the MR images at three diļ¬erent slices. Figure 5(middle)shows the corresponding PET images. High F-MISO uptake are shown in red and they are observedto appear within the tumor boundary. Figure 5(bottom) shows a pO image by the EPRI for thesame image slices. The hypoxic regions within the tumor, deļ¬ned as pO ā¤
10 torr, is indicatedby dark blue color in the image. Some hypoxic regions are overlapped in both F-MISO image andEPRI oxygen images. However, there is also discrepancy between images.ā 7 ā igure 5 . (top) A MR T image of the tumor in a mouse. The tumor boundary is determined based on theMR image and marked by pink color line. (middle) A PET image for the same slice shows high F-MISOuptake in the tumor. (bottom) pO image acquired by the EPRI. The hypoxic region within the tumor isindicated by dark blue color. We have integrated a PET system with an EPRI system and produced successful imaging resultsfor animals. Recently, a combined PET/PERI system was reported by Tseytlin et al. [32], showingsuccessful imaging of a multi-modality phantom containing trityl and FDG solution, but no animalimaging results. This system is diļ¬erent from ours reported here in terms of the PET detectortechnology and the integration of the PET and EPRI systems. It is a 21 cm inner diameter ringmade of 12 detector modules, each of which consists of a 32x32 array 1.5x1.5x10 mm LYSOcrystals (1.57 mm pixel) coupled to a 10x10 array of 3x3 mm SiPMs (4.85 mm pitch). Readout ofthe 10x10 SiPMs is based on the conventional charge division multiplexing to reduce the signals to2x2, and these signals are acquired by using the conventional analog-to-digital converters (ADC)and TDCs. As its diameter is larger than the gap between the EPRI magnets, the PET ring cannotļ¬t with the EPRI system in the way shown in Figure 1(c). Instead, it "sits" on the bottom EPRImagnet with a 20 ā¦ tilt angle to allow some access to the imaging-sensitive volume. Anothercrucial important diļ¬erence lies in the design of the EPR oxygen imaging system. The EPR imagerof Tseytlin et al. utilizes Rapid Scan (RS) EPR acquisition [33] measuring pO with spin-spin(SS) relaxation based linewidth measurement while we use spin-lattice relaxation (SLR) measuredā 8 āith pulse mode inversion recovery acquisition. SS relaxation based pO measurements are moresusceptible to confounding broadening of the spin probe linewidth by the spin probe itself to whichSLR is far less susceptible [22]. Thus the proposed system is uniquely suitable to study the eļ¬ectsof hypoxia.While we reported successful imaging without PET detector shielding, for optimal imagingperformance we believe some shielding is still necessary. However, we desire to mitigate theshielding requirement as much as possible to maintain a simple and compact detector design.Recently, we considered a new SLB design to provide inherent shielding of the PET signal traces.The current SLB design shown in Figure 6(a) uses PCB with a single ground plane. In the new SLB,another ground plane is added so that the SiPM signal traces inside the PCB are laid between the twoground planes, as shown in Figure 6(b). A preliminary comparison test between the current and newdesign SLB was conducted by using single detector module that was coupled to the current or newSLB. The result shows that the EPR RF interference is signiļ¬cantly reduced with the newly designedSLB. Exemplary waveforms of RF noise events, recorded by the DRS4 waveform sampler [34], forthe two SLB boards are shown in Figures 6(c) and (d). The amplitude of RF signal appearing on thePET electronics was reduced from often exceeding 700 mV to 50 mV. The percentage of spuriousRF events in singles (without the proposed RF rejection method) also is reduced from 76% to 2.6%.We plan to use this new SLB in our next implementation of the PET system. We will also considerthe use of thin copper foils to wrap the PET insert for improved shielding. Figure 6 . (a) The PCB layout of strip-line board currently used in the PET (b) The PCB layout in a newlydesigned strip-line board. (c) An exemplary waveform of RF noise using the current PCB. (d) An exemplarywaveform of RF noise using the new PCB. The RF noise is riding on top of scintillation signal.
Precise delineation of hypoxic region in tumor is essential for improving the outcome of oxygenimage-guided radiation therapy treatment for hypoxic tumor. PET imaging with F-MISO has beenused for hypoxic tumor imaging in clinical trials. However, due to the complexity of F-MISOaccumulation and its possible dependence on factors other than hypoxia, F-MISO PET imaginghas failed to produce accurate assessment of hypoxic tumor regions in vivo . Toward quantitativeF-MISO PET imaging, we prototyped a PET/EPRI combined system for small animal imaging forā 9 āssessing, and potentially calibrating, F-MISO PET hypoxia imaging by using simultaneous EPRIoxygen images as reference. The initial F-MISO and EPR imaging on mouse model hypoxic tumorwas successfully conducted using the developed system.The PET scanner was observed to pick-up RF noises during simultaneous PET/EPRI operation.An eļ¬ective RF-event rejection method was developed by exploiting the diļ¬erence between thewaveform features of the scintillation and RF events. A newly designed SLB board that can bettershield the PET signals from EPRI pulsing is also tested and the result is encouraging. A morecomprehensive animal imaging with the developed PET/EPRI system is undergoing and we arein the process of analyzing the correlations between the resulting F-MISO PET and EPRI oxygenimages, and with other physiological parameters derived from MRI imaging, to develop a morecomplete understanding of F-MISO uptake by healthy and tumor tissues.
Acknowledgments
The authors acknowledge that this work was supported in part by the National Institutes of Healthunder Grant numbers R01 CA236385, R01 CA098575, R03 EB027343, R01 EB029948, R50CA211408, P30 CA014599, P41 EB002034, S10 OD025265, and T32 EB002103. We are alsograteful for excellent support from Integrated Small Animal Imaging Research Resource and theCyclotron facility at the University of Chicago.
References [1] M. Hockel and P. Vaupel,
Tumor hypoxia: deļ¬nitions and current clinical, biologic, and molecularaspects , J. Natl. Cancer Inst. (2001) 266.[2] J. C. Walsh, A. Lebedev, E. Aten, K. Madsen, L. Marciano and H. C. Kolb, The clinical importance ofassessing tumor hypoxia: relationship of tumor hypoxia to trognosis and therapeutic opportunities , Antioxid redox signal. (2014) 1516.[3] R. Sullivan and C. H. Graham, Hypoxia-driven selection of the metastatic phenotype , CancerMetastasis Rev. (2007) 319.[4] X. Lu and Y. Kang, Hypoxia and hypoxia-inducible factors: master regulators of metastasis , Clin.Cancer. Res. (2010) 5928.[5] L. H. Gray, A. D. Conger, M. Ebert, S. Hornsey and O. C. Scott, The concentration of oxygendissolved in tissues at the time of irradiation as a factor in radiotherapy , Br. J. Radiol. (1953) 638.[6] D. M. Brizel, R. K. Dodge, R. W. Clough and M. W. Dewhirst, Oxygenation of head and neck cancer:changes during radiotherapy and impact on treatment outcome , Radiother. Oncol. (1999) 113.[7] B. J. Moeller, R. A. Richardson and M. W. Dewhirst, Hypoxia and radiotherapy: opportunities forimproved outcomes in cancer treatment , Cancer Metastasis Rev. (2007) 241.[8] J. G. Rajendran, K. R. G. Hendrickson, A. M. Spence, M. Muzi, K. A. Krohn and D. A. Mankoļ¬, Hypoxia imaging-directed radiation treatment planning , Eur. J. Nucl. Med. Mol Imaging (2006)44.[9] D. Thorwarth, S. M. Eschmann, F. Paulsen and M. Albert, Hypoxia imaging-directed radiationtreatment planning , Int. J. Radat. Oncol. Biol. Phys. (2007) 291. ā 10 ā
10] C. T. Lee, M. K. Boss and M. W. Dewhirst,
Imaging tumor hypoxia to advance radiation oncology , Antioxid redox signal. (2014) 313.[11] B. Epel, M. C. Maggio, E. D. Barth, R. C. Miller, C. A. Pelizzari, M. Krzykawska-Serda et al., Oxygen-guided radiation therapy , Int. J. Radiation Oncol. Biol. Phys. (2019) 977.[12] J. S. Rasey, Z. Grunbaum, S. Magee, N. J. Nelson, P. L. Olive, R. E. Durand et al.,
Characterization ofradiolabeled ļ¬uoromisonidazole as a probe for hypoxic cells , Radiat. Res. (1987) 292.[13] J. G. Rajendran and K. A. Krohn,
F18 ļ¬uoromisonidazole for imaging tumor hypoxia: imaging themicroenvironment for personalized cancer therapy , Semin. Nucl. Med. (2015) 151.[14] P. Vera, S. Thureau, P. Chaumet-Riļ¬aud, R. Modzelewski, P. Bohn, M. Vermandel et al., Phase iistudy of a radiotherapy total dose increase in hypoxic lesions identiļ¬ed by f-misonidazole pet/ct inpatients with non-small cell lung carcinoma (rtep5 study) , J. Nucl. Med. (2017) 1045.[15] Y. Masaki, Y. Shimizu, T. Yoshioka, Y. Tanaka, K. Nishijima, S. Zhao et al., The accumulationmechanism of the hypoxia imaging probe āfmisoā by imaging mass spectrometry: possibleinvolvement of low-molecular metabolites , Scientiļ¬c Reports (2015) 1.[16] D. Monnich, E. Troost, J. Kaanders, W. Oyen, M. Alber and D. Thorwarth, Modelling and simulationof [18f]ļ¬uoromisonidazole dynamics based on histology-derived microvessel maps , Phys. Med. Biol. (2011) 2045.[17] B. H. Kipp, C. Faraj, G. Li and D. Njus, Imidazole facilitates electron transfer from organicreductants , Bioelectrochemistry (2004) 7.[18] M. Elas, J. M. Magwood, B. Butler, C. Li, R. Wardak, R. DeVries et al., Epr oxygen images predicttumor control by a 50% tumor control radiation dose , Cancer Res. (2013) 5328.[19] B. Epel, M. Krzykawska-Serda, V. Tormyshev, M. C. Maggio, E. D. Barth, C. A. Pelizzari et al., Spinlattice relaxation epr po images may direct the location of radiation tumor boosts to enhance tumorcure , Cell Biochem. Biophys. (2017) 295.[20] J. L. Berliner and H. Fujii, Magnetic resonance imaging of biological specimens by electronparamagnetic resonance of nitroxide spin labels , Science (1985) 517.[21] H. J. Halpern, D. P. Spencer and J. van Polen,
Imaging radio frequency electron-spin-resonancespectrometer with high resolution and sensitivity for in vivo measurements , Rev. Sci. Instrum. (1989) 1040.[22] B. Epel, M. K. Bowman, C. Mailer and H. J. Halpern, Absolute oxygen r š imaging in vivo with pulseelectron paramagnetic resonance , Magn. Reson. Med. (2014) 362.[23] H. Kim, C.-T. Chen, N. Eclov, A. Ronzhin, P. Murat, E. Ramberg et al., A feasibility study of a pet/mriinsert detector using strip-line and waveform sampling data acquisition , Nucl. Instrum. MethodsPhys. Res. A (2015) 557.[24] H. Kim, C.-T. Chen, N. Eclov, A. Ronzhin, P. Murat, E. Ramberg et al.,
A silicon photo-multipliersignal readout using strip-line and waveform sampling for positron emission tomography , Nucl.Instrum. Methods Phys. Res. A (2016) 119.[25] D. Xi, C. M. Kao, W. Liu, C. Zeng, X. Liu and Q. Xie,
Fpga-only mvt digitizer for tof pet , IEEETrans. Nucl. Sci. (2013) 3253.[26] H. Kim, Y. Hua, H.-T. Chen, H.-M. Tsai, C.-T. Chen, G. Karczmar et al., Design, evaluation andinitial imaging results of a pet insert based on strip-line readout for simultaneous pet/mri , Nucl.Instrum. Methods Phys. Res. A (2020) 163575. ā 11 ā
27] B. Epel, S. V. Sundramoorthy, C. Mailer and H. J. Halpern,
A versatile high speed 250-mhz pulseimager for biomedical applications , Concepts Magn. Reson. B (2008) 163.[28] B. Epel, I. Gromov, S. Stoll, A. Schweiger and D. Goldfarb, Spectrometer manager: a versatilecontrol software for pulse epr spectrometers , Concepts Magn. Reson. B (2005) 36.[29] K. H. Anh and H. J. Halpern, Spatially uniform sampling in 4-d epr spectral-spatial imaging , J.Magn. Reson. (2007) 152.[30] L. Shepp and Y. Vardi,
Maximum likelihood reconstruction for emission tomography , IEEE Trans.Med. Imaging (1982) 113.[31] A. A. Kuzhelev, D. V. Trukhin, O. A. Krumkacheva, R. K. Strizhakov, O. Y. Rogozhnikova, T. I.Troitskaya et al., Room-temperature electron spin relaxation of triarylmethyl radicals at the x- andq-bands , J. Phys. Chem. B (2015) 13630.[32] M. Tseytlin, A. V. Stolin, P. Guggilapu, A. A. Bobko, V. V. Khramtsov, O. Tseytlin et al.,
A combinedpositron emission tomography (pet)-electron paramagnetic resonance imaging (epri) system: initialevalution of a prototype scanner , Phys. Med. Biol. (2018) 105010.[33] J. P. Joshi, J. R. Ballard, G. A. Rinard, R. W. Quine, S. S. Eaton and G. R. Eaton, Rapid-scan epr withtriangular scans and fourier deconvolution to recover the slow-scan spectrum , J. Magn. Reson. (2005) 44.[34] S. Ritt, R. Dinapoli and U. Hartmann,
Application of the drs chip for fast waveform digitizing , Nucl.Instrum. Methods Phys. Res. A (2010) 486.(2010) 486.