Advantages and challenges of whole-body MR-PET Gaspar Delso Axel Martinez-Möller Stephan Nekolla Sibylle Ziegler
Why MR+PET? MR High spatial resolution (~1mm) Good imaging of anatomy. Excellent soft-tissue contrast. Versatility to target biological processes (flow, perfusion, diffusion, ) Wide range of contrast agents (indirect measurement) No ionizing radiation. PET Very high sensitivity. Versatility to target biological processes. Wide range of radiopharmaceuticals (direct measurement) Long acquisition time.
Why whole-body MR-PET? Good soft tissue contrast (abdomen, prostate, ) Multiple molecular probes (dual tracer tumor delineation, ) No ionizing radiation due to MR (long/multiple/dynamic studies, follow-up, pediatric, ) Motion correction (respiratory/cardiac)
Example: Prostate imaging T2 DWI [ 11 C]Choline PET/CT Pat. M., R.-D. 72 Years Prostate Cancer ct1, cnx, cmx Gleason Score 3 + 4 = 7 VIBE (Perfusion)
Example: Prostate imaging Choline Creatinine Citrate Polyamine Prostate cancer - High phospholipid membrane turnover - Switch from citrateproduction to acetateoxidizing metabolism Krause et al., unpublished data
Example: Prostate imaging
Example: Cardiac imaging Myocardial perfusion imaging PET MRI Hybrid PET/MRI Allows quantification of MBF, accurate method to detect obstructive CAD and early changes in vasoreactivity. A potential alternative to nuclear imaging for diagnosis of CAD in experienced centers. MRI-based correction for attenuation, motion, and partial volume correction for improved quantification of MBF. Non-invasive coronary angiography Assessment of LV function - Presently inferior performance, but no radiation or contrast agents as in coronary CTA. Limited by low spatial resolution. Most accurate technique to determine LV function and structure. Possibility to combine anatomy (plaque burden, luminal obstruction) with hemodynamic consequences (ischemia) of CAD. Combination of LV function with perfusion, metabolic or molecular imaging for improved stratification of heart failure. Assessment of infarction and viability 18 F-FDG uptake is a gold standard of viability, prognostic value shown in several studies. High resolution delineation of infarction by Gd-DTPA delayed-enhancement, more studies on prognostic value expected. Potential of MRS? More detailed risk assessment by combining glucose uptake (jeopardized myocardium) and delayed-enhancement (irreversible scar) Molecular imaging Excellent sensitivity, tracers for many targets (inflammation, angiogenesis, sympathetic nervous function, therapeutic genes or cells...) Lower sensitivity than that of PET, evolving tracers and imaging techniques for molecular targets. Exact anatomical localization and volume correction by MRI for detection and quantification of molecular targets by PET. Nekolla et al., in press.
Architectures Integrated Sequential Technical realization Interference Simultaneous acquisition Room size Standalone MR performance? Cost?? Delso et al., in press.
Integrated architecture PET detector ring inserted between the gradient and RF coils. Squeeze in the available 5 cm. 70 cm - 2x5 cm = 60 cm bore.
Insert architecture Primary coil Gradient coils RF coil
Compatibility Challenges MR static field Typical 1.5 Tesla. Required homogeneity 1 ppm @ 50 cm sphere. Required stability 0.1 ppm/h. Required periodic effects in TR range <0.01 ppm. PET detectors Tolerant to magnetization (APD / SiPM detector) Low susceptibility ( <10-5 ) No generation of time varying fields. MR gradient fields Typical 40 mt/m @ slew rate 200 T/m/s. Required residual eddy currents < 0.1-0.04 %. Required integral stability phase error @TE 60 ms < 4. Required step linearity < 0.03% amplitude. Tolerant to vibration in the range of 1-10 m in the khz range. Tolerant to temperature changes from 20 to 70 C. Tolerant to induction up to 100 V in a loop of 1 mm². No eddy currents, no conductors thicker than 20 m.
Compatibility Challenges MR RF signal Transmit Typical pulse length 1.5 ms. Typical RF field up to 100/200 T @ 63 MHz Required field homogeneity < 5%. Receive Typical signal in the nt-range, up to 23 dbm @ coils. Required pre-amplification 30 db. Required RF shielded cabin -100 db. SNR is crucial. Shielding+ filtering. PET detectors Low electromagnetic emission. Shielding requirements vs. no eddy currents. Induction in the range of V/cm 2 Avoid big conductive structures. Blanking necessary?. Be quiet!! Must not introduce noise. (At least in the MR frequency range) Outside of RF shield. Must not introduce excessive gamma attenuation (Reduce high Z and excess materials)
Simulations Monte Carlo simulations with GATE NEMA protocol 72 PET detector rings Ø 66 cm Axial field of view 19.2 cm Crystal dimensions 2.5x2.5x20 mm 3 Body coil (GRP plastic) MR bed and rails (glass fiber) Cables and housing (copper)
Simulations Monte Carlo simulations with GATE NEMA protocol 72 PET detector rings Ø 66 cm Axial field of view 19.2 cm Crystal dimensions 2.5x2.5x20 mm 3 Body coil (GRP plastic) MR bed and rails (glass fiber) Cables and housing (copper)
Compatibility Challenges Photomultiplier tubes (PMT) Avalanche photodiodes (APD) Silicon photomultipliers (SiPM) Sensitive to magnetic fields Yes No No Quantum efficiency 20% 70% 70% PDE 25% - 65% Rise time ~1 ns ~5 ns ~1 ns Gain Up to 10 6 Up to 10 3 Up to 10 6 Operation voltage >1000 V 300-1000V 30-80 V Temperature sensitivity <<1% per C ~ 3% per C ~ 2% per C Size Ø10-52mm 5x5mm2 1x1mm2
PET reconstruction challenges Attenuation map Field of view MR coils
Attenuation map Attenuation map in standalone PET and PET/CT Soft tissue Fat Soft tissue Lungs Lungs Bones Transmission image with rotating 68 Ge/ 68 Ga source CT-based Attenuation Map How to obtain the attenuation coefficients at 511 kev from an MR acquisition?
Attenuation map Hypothesis: A segmented attenuation map including four classes (background, lungs, fat, soft tissue) is suitable for AC Evaluation using PET/CT no misregistration bias Protocol and patients: 18 F-FDG PET/CT (Siemens Biograph 16) 35 oncology patients. 52 focal lesions (15 lung, 21 bone, 16 neck) Low-dose CT (no IV contrast media) Two patients underwent an additional MR.
Attenuation map Lesion SUV = -
Lesions Average SUV = -
Attenuation map Impact of segmentation (% of change in SUV mean ) Worst case (13% difference)
Attenuation map Impact of bone removal (% of change in SUV mean ) Agreement with Nakamoto et al. (JNM 2002), who reported 11% difference PET/CT vs. radionuclide attenuation for osseous lesions.
Field of view limitation Possible complementary information: Camera. Laser. Radiation source. Mechanical. PET pre-reconstruction.
Simulate MR field of view f.o.v. contours extracted from MR image CT used as gold standard for attenuation correction
Simulate MR field of view f.o.v. contours extracted from MR image CT with same field of view as MR
Error introduced by f.o.v. limitation ~50% ~30% Original Limited f.o.v. Normalized error (%)
Test robustness of possible correction Patient contour extracted from CT CT with rough correction of missing attenuation
Error after correction <10% ~15% Original Corrected f.o.v. Normalized error (%)
MR coils Redesign. Modeling. Positioning. Body matrix coils Peripheral angiography matrix coil Neck matrix coil Head matrix coil
Redesign Axial view Coronal view Sagittal view HR+ Emission HR+ Transmission Coronal view Sagittal view
Redesign Reconsider choice of materials. Avoid massive components. Avoid flat surfaces. Normalized transmission sinogram view
Modeling & Positioning HR+ transmission PET/CT transmission PET/CT transmission +Model
PET/CT transmission +Misregistered model 1cm
Modeling & Positioning
Conclusions Lots of advantages To-dos Sequential or integrated? Release vs. cost vs. application