[PDF] - Inside a CT Scanner Web Stayman, Advanced Imaging Algorithms and PDF Document

SLIDE 1

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 1

J. Webster Stayman

Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) Johns Hopkins University October 16, 2019

Inside a CT Scanner

SLIDE 2

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 2

CT in Operation

https://www.youtube.com/watch?v=2CWpZKuy-NE

Mean measurements as a function of parameters

Forward Model

SLIDE 3

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 3

100 200 300 400 500 600 700 100 200 300 400 500 600 700

0.08
0.06
0.04
0.02

0.02 0.04 0.06 0.08 100 200 300 400 500 600 700 100 200 300 400 500 600 700 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000

Noise in Projection Data

Detector Patient X-ray Source

Beam Shaping – Bowtie Filters

SLIDE 4

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 4

Moiré patterns

Mathijs Delbaere, behance.net

Multiple Aperture Devices (MADs)

(Stayman et al., SPIE 2016)

Detector Patient MAD Filter X-ray Source

X-rays 135 mm 15 mm Thickness: 2mm

MAD0 MAD1

Spacing: 10 mm

Manufacturing:

Tungsten powder, Laser Sintering, EDM Wire cutting

Design Target:

Flatten fluence behind 1) a QRM phantom (30 x 20 cm), 2) cylinders 20~50 cm in diameter

Stayman, SPIE 2016 ; Mathews, CT Meeting 2016, SPIE 2017 ; Gang, PMB 2019

MAD0 MAD1

1 cm X-ray Source MADs Motion Stage Flat Panel Detector 0 cm SMD = 34 cm SAD = 80 cm SDD = 108 cm

Actuation Stages

Experimental CBCT Bench

Motion system on CT gantry

Diagnostic CT Scanner

Beam shape: Relative displacement between MADs Amplitude: View-dependent mA and/or ms modulation Miscentered object* and VOI imaging**: Both MADs moving to change beam center

*Mao, SPIE 2018, JMI 2018; ** Wang, CT Meeting 2018, SPIE 2019, JMI 2019

Linear motors

SLIDE 5

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 5

Uniform Elliptical acrylic phantom (25.8 x 14.1 cm) MAD gain in air Phantom acquisition “One period” of MAD profiles Actuation for elliptical phantom

MAD Acquisition - Ellipse Phantom

Prior Information about Patient Initial Ultra-Low Dose Scan Custom Acquisition

?

Patient-Driven Diagnostic CT 3D Scout Volume Customized Tube Current Modulation Customized Model-based Reconstruction Custom Regularization

r1 r2 r3 r4

Customized CT Workflow

𝜈 = argmax Φ 𝑧; 𝜈 Φ = 𝑀 𝑧, 𝜈 − 𝑆(𝜈)

SLIDE 6

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 6

Image Quality

True Signal Added noise realizations with different correlation while maintaining same variance “noise masquerading as signal” Contrary to CNR, task-based metrics: Accounts for spatial frequency characteristics of task, noise and system response Accounts for observer detection strategy and detection threshold

Task-Based Performance

SLIDE 7

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 7

Spatial resolution

Detectability Index (non-prewhitening observer)

Noise Imaging task 𝑒

Ω, Ω =

∫ ∫ ∫ 𝑁𝑈𝐺

2 Ω, Ω ⋅ 𝑋2𝑒𝑔 𝑒𝑔 𝑒𝑔

∫ ∫ ∫ 𝑂𝑄𝑇

Ω, Ω 𝑁𝑈𝐺 2 Ω, Ω ⋅ 𝑋2𝑒𝑔 𝑒𝑔 𝑒𝑔

Task Function

vs

𝑦 𝑧

𝑋

𝑔

𝑔
Task-Performance: Detectability Index

Hypothesis #1 Hypothesis #2

Quadratic Penalty Likelihood

𝜈 = argmax log 𝑀 𝜈; 𝑧 − 𝛾𝑆 𝜈

Penalized Likelihood Estimation (PLE)

𝑂𝑄𝑇

≈

ℱ AD 𝑧 A𝑓

ℱ AD 𝑧

A𝑓

+ 𝛾𝐒𝑓

Object dependence

(via the projection data)

𝑁𝑈𝐺

≈

ℱ AD 𝑧 A𝑓

ℱ AD 𝑧

A𝑓

+ 𝛾𝐒𝑓

𝑧

:

Fessler, IEEE-TIP 5(3), (1996); Stayman and Fessler, Trans. Med. Im. 23(12),2004; Zhang-O’Connor and Fessler, IEEE, 2007;

ej : Location Dependence

Forward Model:

𝛾𝐒: Regularization dependence

Gang et al., Med Phys 41(8) 2014

Local NPS and MTF

(1) (2) (3)

𝑧 = 𝐽 𝑣, 𝜄 𝑓

𝑣: Detector location, q: Projection angle

NPS MTF

x10-4

0.5 1.0 1.2

(1) (2) (3)

0.4

0.4

fx

0.4

0.4

fy

0.4

0.4

fx

0.4

0.4

fx

0.4

0.4

fy

System Modeling

SLIDE 8

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 8

Task-Driven Optimization System Model Objective Optimizer

Low Dose 3D Scout

Location Contrast Spatial frequency

argmax

,

𝑒 Ω, Ω 𝑁𝑈𝐺 Ω, Ω Spatial resolution: Noise: 𝑂𝑄𝑇 Ω, Ω

Detectability index

Imaging Task

Anatomical Model

𝑒′ Ω, Ω

𝛁𝑩

∗

Ω Ω

mAs, kV, Orbit Fluence field Kernel (FBP) Regularization (MBIR) Conventional Task-driven

0o

90o 180o 270o

0.8 mAs 0.6 0.4 0.2

Task-Driven Imaging Framework

r1 r2 r3 r4

𝛁𝑺

∗

Task-Driven Imaging

0.65

−0.05 0.37 −1.07

1.12 0.50

0.91
0.71

𝑠

Unmodulated Uniform Signal Minimum Variance* (FBP) Task-driven TCM Task-driven Regularization Task-driven TCM + Reg.

90° 0° 180° 270° 0° 180° 0° 180° 0° 180° 0° 180° 0° 180° *Gies et al, MedPhys, 1999

Traditional and Task-Driven Acq/Recon

SLIDE 9

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 9

Unmodulated Uniform Signal Minimum Variance (FBP) Task-driven TCM Task-driven Anisotropic Regularization Task-driven TCM + Aniso Reg.

𝑒′ = 0.77 𝑒

= 1.0

𝑒′ = 0.90 𝑒′ = 1.08 𝑒′ = 1.0 𝑒′ = 1.10

All recons: are quadratic penalized likelihood have the 3 target stimulus have optimal b selection

Sample Reconstructions

G. Gang et al. SPIE Medical Imaging, March 2016, 9783, PMCID: PMC4841467
G. Gang et al. Physics in Medicine and Biology, December 2017, PMCID: PMC5738673

Traditional versus Task-Driven Fluence-Field Modulation

Phantom Task and Phantom Definition rij(x,y) Task Function Stimulus b (x,y) Flattened Fluence

Min. Mean Variance (FBP)

Task-Driven Design Objective

argmax

,

min 𝑒

Ω, Ω

𝑒

Ω, Ω

⋮ 𝑒

Ω, Ω

SLIDE 10

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 10

Reconstructions of Simulated FFM Data

Unmodulated a = 0.5 Task-Driven a = 1.0

Detectability Maps

G. J. Gang, J. H. Siewerdsen, and J. W. Stayman,

"Task-driven optimization of fluence field and regularization for model-based iterative reconstruction in computed tomography", IEEE Transactions on Medical Imaging (Special Issue on Low-Dose CT), 36(12), 2424-35 (December 2017)

Other Novel Acquisition: Orbits

SLIDE 11

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 11

Preoperative Image Planning Data Task Definition Conventional Intraoperative CT

X-ray Source Flat-Panel Detector

Traditional Circular Trajectory

Interventional Imaging Conventionally Ignored by Interventional Devices

Task-Driven Trajectory

Prior Information about Patient and Task

?

Patient- and Task-Driven Intraoperative CT Diagnostic Imaging

Task-Driven Interventional Imaging Trajectory Design

SLIDE 12

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 12

Anthropomorphic Head Phantom and Synthetic Vasculature CBCT Testbench with 6DOF Object Platform

Emulated Workflow on CBCT Bench

Circular Scan Task-Driven Trajectory Preoperative Scan CBCT Bench Results

J. W. Stayman and J. H. Siewerdsen, Int'l Mtg. Fully 3D Image Recon. June 16-21, 2013.
S. Capostagno et al. and Stayman et al.,”Task-driven source–detector trajectories in cone-beam computed tomography: I and II” J. Med. Im. 2019

SLIDE 13

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 13

Optimization:

30 stimulus locations on ellipsoid surrounding embolization coil 9 orbital bases

50° ≤ F ≤ 50°

0° ≤ q ≤ 360° CMA-ES (pop=40)

 

   

 

   

 

   

 

 

 

2 2 2 1 2 (1) (2 ( ) , )

arg max min ' ; , ' , , , ; ,..., ' ; ˆ ˆ ˆ ,

Task Task Task L L

d d d W W W a a a a a a a

 F

   F  F  F  F

Multi-Location (or Multi-Task) Optimization Circular vs. Designed Orbit

SLIDE 14

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 14

Circular Scan Task-Driven Trajectory

0.014 0.016 0.018 0.02 0.022 0.024 0.026 0.028

Circular vs. Designed Orbit Reconstructions

J. W. Stayman et al., Int'l Mtg. Fully 3D Image Recon. June 1-4, 2015.
S. Capostagno et al. and Stayman et al.,”Task-driven source–detector trajectories in cone-beam computed tomography: I and II” J. Med. Im. 2019

Spectral CT involves measurements with varied spectral sensitivity and allows for enhanced material discrimination. Single-shot multi-phasic studies New contrast agent development

Gold Nanoparticles (angiography, lung cancer) Bismuth Sulphide (lymph nodes) Tantalum Oxide (cartilage, lymph nodes) Xenon (lung ventilation)

Symons R, Krauss B, Sahbaee P, Cork TE, Lakshmanan MN, Bluemke DA, Pourmorteza A. Photon-counting CT for simultaneous imaging of multiple contrast agents in the abdomen: An in vivo study. Med Phys. 2017 Oct 26 44(10):5120–5127.

Spectral CT with Photon Counting Detectors

Spectral CT

SLIDE 15

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 15

Energy-Dependent Attenuation

𝑧 = 𝑇 exp − 𝑟

𝑏𝜍
Forward Model

𝑧 = 𝐂exp −𝐍𝜍

Φ = 𝑀 𝑧; 𝜍 + 𝑆 𝑀 𝑧; 𝜍 = 𝑥 𝑧 𝜍 − 𝑧

Penalized-Likelihood Objective Function

Spatial-Spectral Filters

X-ray Focal Spot Spatial-Spectral Filter Multi-Contrast Phantom Energy- Integrating Detector

Overall System Geometry Varied Spectral Beam-lets

Stayman JW and Tilley II S (May 2018) Model-based multi-material decomposition using spatial-spectral filters CT Meeting 2018 102-105. Filtered Beamlets Pb Au Lu Er Polyenergetic Incident Beam

K-Edge Filter Tiles Translating Tiled Filter

Filter Details Data Collection

SLIDE 16

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 16

Spatial-Spectral Simulation

I Gd Au Tivnan M, Wang W, Tilley II S, Stayman JW (July 2019) Designing Spatial-Spectral Filters for Spectral CT AAPM Annual Meeting

Physical Experiments

Sb Er W PbSbErW Pb

SLIDE 17

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 17

Physical Experiments

Teflon ~1900mg/ mL (x7) ~1000mg/ mL Solid Water 2mg/mL I 5mg/mL I 7.5mg/mL I 10mg/mL I 15mg/mL I 20mg/mL I Teflon

Spatial-Spectral Filters

Filter Order: Pb, Au, Er, Lu
Filter Speed: 3 pixels/view
Filter Thickness: 0.25mm

Filter Tile Width: 2mm /10 pixels

Materials: Solid Water/Iodine
Teflon

Closing Thoughts

Customized data acquisitions allow task-based optimization Image quality assessments should take the imaging task into account Understanding and prediction of image properties can drive customization Challenges:

SLIDE 18

IMA Workshop - "Novel CT Data Acquisition and Processing" 10/16/2019 Web Stayman, Advanced Imaging Algorithms and Instrumentation Lab (aiai.jhu.edu) 18

Collaborators Grace Gang – Biomedical Engineering Kelvin Hong - (Interventional) Radiology Satomi Kawamoto - Radiology A Jay Khanna - Orthopaedic Surgery Eleni Liapi – Radiology Jeff Siewerdsen – Biomedical Engineering Marc Sussman - Thoracic Surgery Nick Theodore - Neurosurgery Clifford Weiss - (Interventional) Radiology Wojtek Zbijewski – Biomedical Engineering Students/Post-docs Jessica Flores Andrew Leong Junyuan Li Stephen Liu Yiqun (Quinn) Ma Hui (Amalie) Shi Matt Tivnan Wenying Wang

AIAI Laboratory Advanced Imaging Algorithms and Instrumentation Laboratory

Website: aiai.jhu.edu

Email: web.stayman@jhu.edu

Johns Hopkins University

Biomedical Engineering, School of Medicine, Medical Campus

NIH Funding R21CA219608 (Prior Images) R01EB025829 (Observers) R01EB025470 (DE Fractures) R21EB026849 (Spatial/Spectral) R43CA239777 (Sparse CT) R01EB027127 (Orbits) R21EB014964 (Metal Implants) U01EB018758 (Beam Modulation) Industry Collaborations Canon Medical Systems Elekta Fischer Imaging Philips Healthcare Siemens Healthineers Varex Imaging