Extremities CBCT. Advances in Cone-Beam CT and Emerging Applications: Clinical Motivation 7/14/2015

Transcription

1 JHU Radiology JHU Radiology Advances in Cone-Beam CT and Emerging Applications: Extremities CBCT W Zbijewski Department of Biomedical Engineering Johns Hopkins University The I-STAR Laboratory istar.jhu.edu Imaging for Surgery, Therapy, and Radiology JH Siewerdsen, JW Stayman, A Sisniega, G Gang, Q Cao, A Muhit, P DeJean, P Prakash Carestream Health J Yorkston, N Packard, R Senn, D Yang JHU Radiology JA Carrino, G Thawait, S Demehri, M Mahesh, A. Machado Funding Support Carestream Health NIH 2R01-CA NIH R21-AR NIH R01-EB Clinical Motivation Fracture / Trauma Monitoring of healing Insufficiency fractures Osteoporosis Dietary deficiencies Radiation therapy Need for biomarkers Arthritis Osteoarthritis (OA) ~30 million US adults Joint replacement Need for early biomarkers Rheumatoid arthritis ~1.5 million US adults Deformity and chronic pain Novel drugs Monitoring of treatment response Gout ~8 million US adults DE imaging Imaging capabilities: Contrast resolution Joint space, tendons, ligaments Spatial resolution Bone micoarchitecure Bone composition (quantitative) Weight-bearing (functional) Longitudinal studies 1

2 MSK Radiology Contrast resolution Spatial resolution Quantitative Functional (load-bearing) Functional (motion) Rad First line imaging Fluoro Motion CT Trauma Fracture MRI Joint internal derangement (+) CT ( US MRI ( Soft-tissue lesions Guidance NM Infection MR exquisite soft tissue MR limited spatial resolution CT high-resolution bone limited + soft tissue + (Dual Energy-CT) A Dedicated Extremity CBCT Scanner System Configuration Flat-panel detector (FPD) Compact gantry Sitting / standing examination Capabilities Weight-bearing scans Natural stance High isotropic spatial resolution Multi-mode Rad / Fluoro / CBCT Simplified logistics Modest imaging dose Longitudinal studies Carestream + JHU * Planmed Verity ** * Zbijewski et al. Med Phys 2011 * Carrino et al. Radiology 2014 ** Tuominen et al. AJR 2013 ** Huang et al. Skeletal Radiol 2015 planmed.com General considerations Side entry sliding door Natural stance FOV: 20 x 20 x 20 cm 3 ~20 sec/scan X-ray source and FPD Fixed anode, 0.5 focal spot kvp, 0.8 kw max mm detector pixel Scanner Configuration Standing Configuration Clinical Prototype Patient studies: Weight-bearing lower extremities Arthritis and trauma Application Development Joint space morphology Bone health Bone microarchitecture Bone Mineral Density (BMD) Sitting Configuration FPD Side Entry X-ray Tube 2

3 SDNR MTF Nominal Protocol (2x2 Binning) mm pixels (0.26 mm voxels) High-Res Protocol (1x1 Binning) mm pixels (0.13 mm voxels) x1 (Hann) Spatial Resolution 1x1 (Ramp) 2x2 2x2 (Ramp) (Hann) Line-Pair Pattern Nominal High-Res lp/cm Cadaver Hand lp/cm Frequency [lp/mm] Contrast Resolution Constant dose ~10 mgy Max Power ~0.875 kw 60 kvp mgy 3 mgy Adipose (-100 HU) BR-1 (-45 HU) 80 kvp kvp 0.5 Muscle (+10 HU) kvp Dynamic Gain, 2X2 Binning (0.388 mm pixels) 0.52 mm isotropic voxels, Hann Filter Cadaver knee Diagnostic Imaging Performance CBCT (10 mgy) MDCT (25 mgy) CBCT MDCT comparison Bone and soft-tissue visualization tasks CBCT: 1 st generation prototype MDCT: Siemens Definition Bone Recon Ramp Filter 0.26 mm voxels isotropic Bone Recon Ultra-High-Res 0.4 mm voxels isotropic Observer study Fresh cadavers 10x knee, 10x hand 4 expert radiologists 8 diagnostic tasks Preference and satisfaction Soft Tissues Soft Tissues Hann Fitler UHR Soft Protocol 0.52 mm voxels isotropic 0.6 mm voxels isotropic Demehri et al, Eur Radiol

4 Joint Space Width (JSW) (mm) Diagnostic Imaging Performance CBCT (10 mgy) Bone Recon Ramp Filter 0.26 mm voxels isotropic Soft Tissues Hann Fitler 0.52 mm voxels isotropic MDCT (25 mgy) Bone Recon Ultra-High-Res 0.4 mm voxels isotropic Soft Tissues UHR Soft Protocol 0.6 mm voxels isotropic Knee Bone Tasks Hand Bone Tasks Excellent 5 Tissue * * Assessment 5 * * Criteria 4 Cortical Visibility 4 and integrity Bone Fair 3 3 Medullary Architecture of trabeculae 2 Bone 2 Poor 1 Fracture Detection 1 and (if present) characterization p = p = CBCT MDCT CBCT MDCT Knee Soft Tissue Tasks Tasks Hand Tissue Assessment Criteria Excellent 5 Tendon * Uniform 5 density * 4 Fair 3 * Integrity at attachment Distinguish 4 superficial * from deep 3 Muscle Architecture of muscle 2 Ligaments Integrity 2 of ligaments Poor 1 Fat Observer study Uniformity CBCT 1 and vs. MDCT visibility and p < Articular Joint space p < width radiography for fractures: Huang at el, Cartilage Skel CBCT Rad MDCT 2015 (Planmed CBCT Verity) MDCT Bone Imaging: Fracture Healing Week 9 Week 15 Joint Space Analysis in OA Joint space width (JSW) Bone image 3.3 Meniscal p < protrusion Soft-tissue image (MP) 4.65 mm 3.21 Medial femoral condyle to tibial plateau OA: 2.05 OA: 1.45 Study setup 17 patients with OA of the knee 18 patients without OA 3 observers Weight-bearing (WB) vs. Non-weight bearing (NWB): Significant difference in JSW for OA Normal OA Normal OA Non-Weight Bearing (NWB) Weight Bearing (WB) JSW change in OA Thawait et al. RSNA

5 Meniscal Extrusion (ME) (mm) Joint Space Analysis in OA Meniscal Extrusion (ME) Soft-tissue image p = mm Study setup 17 patients with OA of the knee 18 patients without OA 3 observers Weight-bearing (WB) vs. Non-weight bearing (NWB): Significant difference in JSW for OA Significant difference in ME for OA No significant difference in JSW or ME for non-oa Normal OA Normal OA Non-Weight Bearing (NWB) WB vs. NWB hindfoot alignement: Hirschmann at el, Eur Rad 2014 (Planmed Verity) Weight Bearing (WB) The Electrostatic Model Φ 1 Ω = V 1 E = Φ E = ρ ε 0 2 Φ x = 0 x Ω Ω R 2, R 3 Φ Electric potential [V] ρ Charge density [C m 3 ] Ω Joint space of interest Φ 0 Ω = V 0 For an electric field Echaracterized by field lines l (dl E = 0) s 1 JSW x 0, y 0, z 0 = dl s ds for all x 0, y 0, z 0 0 Ω s 0 Cao et al Phys Med Biol 2015 Joint Space Maps A015 A041 Normal A066 A067 A072 A073 Classification Accuracy Electrostatic Model JSM Osteoarthritic A021 A030 M A L P A038 A023 A031 A043 Minimum JSW Radiologist 5

6 Trabecular Volume Fraction Structure Model Index Degree of Anisotropy Trabecular Thickness (mm) Quantitative Imaging of Bone Health Bone health Structure: Micro-architecture (micro-ct) Composition: Bone Mineral Density BMD (DEXA / qct) Bone Marrow Edema BME (T2-weighted MRI) (increased fluid content) Risk of fracture Mechanical competency (e.g. response to load) Osteoporosis, arthritis, insufficiency fractures CBCT: integrated platform for bone health assessment Superior bone visualization to conventional CT Micro-architecture: High resolution required (~100 mm) Model-based reconstruction with deblurring Multi-resolution model-based reconstruction (speed) CMOS detectors Micro-CT T2 MRI qct BME * Sisniega et al, Phys. Med. Biol. (2015) Quantitative Imaging: Microarchitecture % Micro CT (mct) (Johns Hopkins Univ) Ex Vivo a vox = 0.1 mm % 1 % % MDCT (Siemens Somatom) In Situ a vox = 0.36 mm CBCT (Dedicated Prototype) In Situ a vox = 0.1 mm Muhit et al. SPIE (In Situ) MCCT (Ex Vivo) MDCT mct (In Situ) CBCT (In Situ) MCCT (Ex Vivo) MDCT mct (In Situ) CBCT % 3 % 42 % % 0 MDCT mct CBCT 0.0 MDCT mct CBCT MDCT (In Situ) MCCT (Ex Vivo) CBCT (In Situ) MDCT (In Situ) MCCT (Ex Vivo) CBCT (In Situ) Ultra-High-Resolution CBCT Imaging : Deblurring and Noise Correlations in PWLS Source Blur B s Extended X-ray Source B s B d Detector Blur B d X-ray Photons Object Light Photons Scintillator Photodiodes Uncorrelated Bare Beam Uncorrelated, Attenuated Correlated Light Photons Correlated Electrons X-rays with Focal Spot Blur with Light Spread with Readout Noise Penalized Weighted Least Squares with Deblurring and Noise Covariance: Image Estimate System Matrix Roughness Penalty μ = arg μ min[ Aμ l T K L 1 Aμ l + βr μ ] Covariance Matrix: Accounts for blur and deblurring Deblurred line integrals Tilley et al, Fully 3D Mtg

7 Fine Grid RMSE Deblurring and Noise Correlations in PWLS Experimental setup dominated by source blur Deblurring the source blur introduces correlations Modelling the correlations reduces noise in deblurred PWLS Both with Deblur Ultra-High-Resolution CBCT Imaging : Multi-Resolution Model-Based Reconstruction Image Estimate System Matrix μ = arg μ min[ Aμ l T K L 1 Aμ l + βr μ ] Challenges for High Resolution PWLS System matrix A is large Multi-resolution PWLS System matrix separates into: -standard resolution component A STD -fine resolution component A FINE y = I 0 exp Aμ = I 0 exp A STD D m μ S exp(a FINE μ F ) Cao et al, Fully 3D Mtg 2015 Multi-Resolution Reconstruction No Downsampling Downsampling=4 Downsampling=10 Downsampling=10 No Downsampling Digital Phantom Study Fixed regularization in m F Varied regularization and downsampling in m S No visible artifacts in m F for downsampling 4 10x 100x acceleration while maintaining quality in m F 7

8 Multi-Resolution Reconstruction Knee phantom on a CBCT test-bench in extremities CBCT configuration Ultra-High-Resolution CBCT Imaging : CMOS detectors CMOS vs. asi Flat Panel Higher resolution (smaller pixels) Lower electronic noise ~500 electrons/pixel Faster read-out 30 fr/sec for 30x30 cm FOV Experimental evaluation 600 mm CsI for CMOS and Flat Panel Extremities CBCT configuration 0.4 focal spot x-ray tube 75 mm voxels, Ramp filter Results and Future Work ~10-15% improvement in PSF Flat Panel Varian mm pixels Flat Panel CMOS Dalsa Xineos mm pixels PSF of a 0.15 mm steel wire CMOS FWHM ~0.26 mm FWHM ~0.31 mm [mm] Ultra-High-Resolution CBCT Imaging : CMOS detectors CMOS vs. asi Flat Panel Higher resolution (smaller pixels) Lower electronic noise ~500 electrons/pixel Faster read-out 30 fr/sec for 30x30 cm FOV Experimental evaluation 600 mm CsI for CMOS and Flat Panel Extremities CBCT configuration 0.4 focal spot x-ray tube 75 mm voxels, Ramp filter CMOS Dalsa Xineos 0.1 mm pixels Flat Panel Varian mm pixels Results and Future Work ~10-20% improvement in PSF Better delineation of trabecular detail CMOS limited by thick scintillator Analytical optimization of scintillator thickness Task-based Detectability d 8

9 BMD (mg/ml) Quantitative Imaging of Bone Health Bone health Structure: Micro-architecture (micro-ct) Composition: Bone Mineral Density BMD (DEXA / qct) Bone Marrow Edema BME (T2-weighted MRI) (increased fluid content) Risk of fracture Mechanical competency (e.g. response to load) Osteoporosis, arthritis, insufficiency fractures Micro-CT qct T2 MRI CBCT: integrated platform for bone health assessment Bone Mineral Density / Bone Marrow Edema Accurate attenuation values Comprehensive artifact correction * BME challenging in conventional CT (partial volume from trabeculae) Material decomposition Dual Energy (DE) imaging BME * Sisniega et al, Phys. Med. Biol. (2015) Quantitative Imaging: BMD BMD calculated using CBCT and MDCT (Mindways QCT) BMD standard materials (CaHA rods) embedded in scanner door CBCT with scatter correction BMD calibration: 0, 75, and 150 mg/ml CaHA rods Osteoporotic (75 mg/ml) Healthy (150 mg/ml) % 0.3% % 3.1% 65 CBCT 1 2 MDCT CBCT MDCT BMD Test Phantom: 16 cm polyethylene cylinder CaHA inserts: 0, 75, 125, 150, 250 mg/ml Muhit et al. SPIE 2013 Dual Energy Imaging of Bone Health μ LE μ HE Image-based DE Three-material decomposition Volume preservation constraint Volume fractions of: Water + Fat/Marrow + Cortical Bone Requires good image uniformity DE Measurement of BMD Bone fraction BMD 1/mm Water Fraction Marrow Fraction Bone Fraction DE Imaging of BME DE: Increased water fraction=edema * Vol. Frac Vol. Frac Vol. Frac. * Pache et al, Radiology (2010) 9

10 Marrow Vol. Fraction Bone100 Bone75 10% 25% 40% Marrow Cortical Bone Vol. Fraction Bone100 Bone75 10% 25% 40% Marrow Base Materials Edema Inserts DE CBCT: Experimental Study Volume Composition Tissue-Mimicking Materials * Cortical Bone=Dipotassium Phosphate K 2 HPO 4 Marrow/Fat=Ethanol Edema: Decreasing fraction of ethanol Water 100% H 2O Marrow 100% C 2H 6O Cortical Bone 100% K 2HPO4 Bone75 75 mg/ml K 2HPO 4 in H 2O Bone mg/ml K 2HPO 4 in H 20 Marrow Bone100 Imaging Experiments ~10 cm diameter phantom HE: 105 kvp, 0.1 mas/frame, ~7 mgy CTDI LE: 60 kvp, 0.8 mas/frame, ~7 mgy CTDI Artifact Correction Fast GPU Monte Carlo scatter correction ** ~5 min correction per scan 10% 10% 90% 25% 25% 75% 40% 40% 60% 25% 10% Bone100 Marrow 40% Bone75 1/mm * Goodsitt et al, Investig. Radiol (1987) ** Sisniega et al, Phys. Med. Biol. (2015) Narrow Beam in-air Three-Material DE Decomposition Water Marrow Bone Narrow Beam 10% 25% Water Marrow Bone 40% Bone75 Marrow Full Beam Scatter Correction /mm BMD Accuracy with Dual Energy CBCT DE estimates of bone and marrow volume fractions Measured Relative difference with in-air narrow beam DE in-air Measured in-air 25% Rel. diff.= % Rel. diff. = % Rel. diff.=0.01 Bone100 Rel. diff. =0.01 Bone75 Rel. diff.= Bone vol. frac. 10

11 Dual Energy with Three Source CBCT 1 source 3 source HE sup LE central HE inf Three-Source CBCT Configuration Custom fixed-anode unit Three x-ray tubes arranged axially Increased Field-of-View Reduced cone beam artifacts New method for Dual-Energy CBCT Obviates need for double scan Image Reconstruction in 3-Source DE CBCT Penalized Likelihood (PL) ˆ m arg max m L m; y R m Forward model L(µ; y) Edge-preserving Huber penalty R Low Energy reconstruction L(µ; y) involves LE central projections High Energy reconstruction L(µ; y) combines HE sup and HE inf LE central * central HE sup * * HE inf HE3 reconstruction HE1 reconstruction slice 1/mm Combined PL reconstruction Bone and Fat Fractions in Cadaveric Knee LE: 60 kvp, 7 mgy HE: 105 kvp, 7 mgy PL reconstruction Narrow Beam 105 kvp Narrow Beam Double Scan DE Double Scan DE Three-Source DE Marrow frac.=0.74±0.13 Bone frac.=0.059±0.008 Marrow frac.=0.88±0.08 Bone frac.=0.064±0.005 Marrow frac.=0.65±0.16 Bone frac.=0.034±0.004 Marrow frac.=0.48±0.18 Bone frac.=0.027± /mm Fat vol. frac Bone vol. frac. 11

12 Bone and Fat Fractions in Cadaveric Knee LE: 60 kvp, 7 mgy HE: 105 kvp, 7 mgy PL reconstruction Double Scan DE Three-Source DE Marrow frac.=0.95±0.03 Bone frac. =0.032±0.005 Marrow frac.=0.85±0.09 Bone frac.=0.029± /mm Fat vol. frac Bone vol. frac. Extremities CBCT Weight bearing Soft tissue contrast Isotropic spatial resolution (~0.5 mm) Low dose (~10 mgy) (~20-40 mgy for high-resolution MDCT) Novel applications 3D Joint space maps Quantitative imaging of bone health Conclusions Ultra-high resolution CBCT for bone morphometry Advanced blur and noise models in PWLS Multi-resolution PWLS Improved delineation of trabeculae with CMOS Scintillator optimization Motion correction Quantitative CBCT of bone composition Peripheral qct DE imaging BMD accuracy within 15% of in-air value Detection of edema in DE CBCT Marrow volume fraction changes 20% 12