![Page 1: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/1.jpg)
Advanced Reconstruction Methods on Philips CT Systems
Sandra Simon Halliburton, PhDDirector of Clinical Architecture
CT R&D
![Page 2: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/2.jpg)
Optimization through system & statistic models
iDose4
Philips Solution for Advance Reconstruction
PARADIGM
SHIFT
![Page 3: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/3.jpg)
128 mAs
20 mAs
IMRFBP iDose
![Page 4: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/4.jpg)
Iterative Model Reconstruction
Formulates image reconstruction as optimization of cost function, F 𝑥 , where 𝑥 is the image that minimizes F
Function is solved iteratively because no explicit solution exists
What is IMR?
function minimum
![Page 5: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/5.jpg)
System Model
Detailed description of system geometry.
Statistics Model
Representation of ideal measurement of object
based on Poisson model of x-ray transport.
Cost Function [F(𝒙)]
![Page 6: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/6.jpg)
Minimized when
estimated data most closely matches actual data
noise is low
Cost Function
F 𝑥 = 𝐷 𝑥 + 𝛽 ∙ 𝑅(𝑥)
Data fit:Difference between
estimated and acquired data
Regularization:Noise penalty
Optimization is balance between data fit and noise
![Page 7: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/7.jpg)
F 𝑥 = 𝐷 𝑥 + 𝛽 ∙ 𝑅(𝑥)
Starting image is xxxxxxxxxxxxxxxxxxxxxxxxxxx after xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx while preserving xxxxxxxxxxxxxxxxxxxxxx
Desired image is controlled by xxxxxxxxxxxxxxxxxxxxxxxtaking into account xxxxxxxxxxxxxxxxxxxxxxxxxxxxx and xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Defined by the desired image and the starting image
Cost Function
![Page 8: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/8.jpg)
= too noisy𝐷 𝑥
𝜷𝟏 ∙ 𝑅(𝑥)
𝐷 𝑥𝜷𝟐 ∙ 𝑅(𝑥)
𝐷 𝑥 𝜷𝟑 ∙ 𝑅(𝑥)
= too smooth
= “just right”
F 𝑥 = 𝐷 𝑥 + 𝜷 ∙ 𝑅(𝑥)
Noise Constraint
![Page 9: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/9.jpg)
True Noise (Monte Carlo)
Noise Map
Noise Modeling
Attenuation Image
![Page 10: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/10.jpg)
Attenuation Image
Noise Map
Noise Modeling
![Page 11: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/11.jpg)
Model-based IR attempts to create a final (optimized) image which is
closest match to projections through image voxels
t1t2
1 0 1
0 2 0
1 0 1
Image matches projections
Optimization achieved
High-resolution, low-noise image
Image
Projections
STATIC OBJECT
xy
Static vs. Moving Object
![Page 12: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/12.jpg)
Model-based IR attempts to create a final (optimized) image which is
closest match to projections through image voxels
t1 t2
1 0 0
0 1 0
0 0 1
Image DOES NOT match projections
Optimization NOT achieved
Image “forced” to match, artifacts
ImageProjections
MOVING OBJECT
Projections
xy
Static vs. Moving Object
![Page 13: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/13.jpg)
Manifestation of motion on model/knowledge based IR is unpredictable
May compromise image quality since optimization “forces” data match
IMR w/o motion compensation
Motion Sensitivity
![Page 14: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/14.jpg)
Yuki et al., JCCT, 2014
Standard iDose4
Cardiac CT
![Page 15: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/15.jpg)
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Brain16 cm
0.9 x 0.45 mm…
Brain CTA15 cm
0.9 x 0.45 mm…
Carotid CTA25 cm
0.9 x 0.45 mm…
Aorta CTA70 cm
0.9 x 0.45 mm…
Chest35 cm
0.9 x 0.45 mm…
Coronary CTA14 cm
0.9 x 0.45 mm…
Abdomen40 cm
0.9 x 0.45 mm…
Re
con
stru
ctio
n t
ime
(m
inu
tes)
Measured on iCT Elite
Number of Images =
1201
IMR Reconstruction Times
![Page 16: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/16.jpg)
↑ Low Contrast Detectability
↓ Dose
↓ Noise
Benefits of IMR
![Page 17: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/17.jpg)
Standard reconstruction
iDose4: Level 6
0
10
20
30
40
50
60
70
80
0 200 400 600
No
ise
(s
tan
da
rd d
evia
tio
n in
HU
)*
Tube Current-Time Product (mAs)
Standard Recon
iDose4 Level6
IMR Level 3
* Image noise as defined by IEC standard 61223-3-5. Assessed
using Reference Body Protocol on a CATPHAN phantom.
iDose4Standard
Recon
iDose4
1 mm slice thickness10 mGy CTDIvol
SD: 15.4 HU SD: 1.9 HUSD: 8.7 HU
↓ Noise
![Page 18: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/18.jpg)
FDA CLEARANCE FOR CLAIMS OF DOSE
REDUCTION
Task-Based Image Quality Assessment
IMAGE GENERATION
TASK OBSERVERFIGURE OF
MERIT
↓ Noise + ↓ Dose + ↑ LCD
![Page 19: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/19.jpg)
IMAGE GENERATION
TASK OBSERVERFIGURE OF
MERIT3mm+14HU
• MITA Test Phantom
![Page 20: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/20.jpg)
Scanned 200x (100x @ each dose)
Reconstructed 200x
Mode Helical
Gantry Rotation Time 750 ms
Beam Collimation 64x 0.625 mm
Pitch 0.6
Tube Potential 120 kV
Tube Current-Time Product 153 mAs 31 mAs
CTDIvol 10 mGy 2 mGy
Reconstruction algorithm FBP IMR
Slice Thickness 0.8
Slice Increment 0.4
IMAGE GENERATION
TASK OBSERVERFIGURE OF
MERIT
• MITA Test Phantom
• 4AFC • Human
• N=36
![Page 21: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/21.jpg)
Created 200 sets of 4 images (100 @ each dose level)– Each image with same dose and
recon algorithm – 1 image 3 mm pin
3 images uniformity section– Random position of image
containing pin
Task = choose image w/ pin
IMAGE GENERATION
TASK OBSERVERFIGURE OF
MERIT
• MITA Test Phantom
• 4AFC • Human
• N=36
![Page 22: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/22.jpg)
36 human observers
Each observer executed 100 trials per dose level
TOTAL TRIALS PER DOSE LEVEL = 3,600
IMAGE GENERATION
TASK OBSERVERFIGURE OF
MERIT
• MITA Test Phantom
• 4AFC • Human
• N=36
![Page 23: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/23.jpg)
Receiver Operating Characteristic (ROC)
A
C
B
D
(1- Specificity) (FPF)
(Se
nsi
tivi
ty)
(TFP
)
Fre
qu
en
cy
True
Positive
Rate
False
Positive
Rate
A
B
C
Test Output
IMAGE GENERATION
TASK OBSERVERFIGURE OF
MERIT
• MITA Test Phantom
• 4AFC • Human
• N=36
![Page 24: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/24.jpg)
Detectability Index (d’)
𝑑′ = 𝑡𝑇𝑃 − 𝑡𝐹𝑃𝜎
-6 -4 -2 0 2 4 6 8 100
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
t
Pro
ba
bil
ity D
en
sit
y
True Positive Rate
False Positive Rate
d'
Burgess, Med Phys, 1995
𝑡𝑇𝑃
𝑡𝐹𝑃
𝜎
Mean of measurementof signal present
Mean of measurementof signal absent
Standard deviation, assumed same for both distributions
IMAGE GENERATION
TASK OBSERVERFIGURE OF
MERIT
• MITA Test Phantom
• 4AFC • Human
• N=36• Detectability
Index, d’
![Page 25: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/25.jpg)
0 0.2 0.4 0.6 0.8 1
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
FPRT
PR
ROC
d'=0
d'=1
d'=2
d'=3
PercentCorrect
d’ AUC
0.25 0 0.50
0.55 1 0.76
0.82 2 0.92
0.96 3 0.98
d’ = 0 Unable to distinguish images with low contrast object
present from image with low contrast object absent
d’ = ∞ Always able to distinguish image with low contrast object
present from image with low contrast object absent
(Se
nsi
tivi
ty)
(TFP
)(1- Specificity) (FPF)
0
IMAGE GENERATION
TASK OBSERVERFIGURE OF
MERIT
• MITA Test Phantom
• 4AFC • Human
• N=36• Detectability
Index, d’
ROC
1
2
3
0 0.2 0.4 0.6 0.8 1
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
FPR
TP
R
ROC
d'=0
d'=1
d'=2
d'=3
![Page 26: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/26.jpg)
IMAGE GENERATION
TASK OBSERVERFIGURE OF
MERIT
Median % Correct
Median d’ CTDIvol
FBP 50% 0.821 10 mGy
IMR 70% 1.475 2 mGy
80% 80%
• MITA Test Phantom
• 4AFC • Human
• N=36
↓ Noise + ↓ Dose + ↑ LCD
• Detectability Index, d’
![Page 27: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/27.jpg)
↑ Low Contrast Detectability 43-80%
↓ Dose 60-80%
↓ Noise 70-83%
In clinical practice, use of IMR may reduce CT patient dose depending on the clinical task, patient size, anatomical location, and clinical practice. A consultation with a radiologist and physicist should be made to determine appropriate dose to obtain diagnostic image quality for particular clinical task.
Benefits of IMR
![Page 28: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/28.jpg)
L¨ove A, et al. Six iterative reconstruction algorithms in brain CT: a phantom study on image quality at different radiation dose levels. Br J Radiol 2013; 86:20130388.
Image Quality Improvement with IMR
IMR provided
Lowest noise
Best low-contrast detectability, including at the lowest dose
Improved resolution and lowered noise simultaneously
![Page 29: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/29.jpg)
Oda S. et al, Iterative model reconstruction: Improved image quality of low-tube-voltage prospective ECG-gated coronary CT angiography images at 256-slice CT, Eur J Radiol (2014)
Low-kVp Made Routine with IMR
Standard iDose4
Standard iDose4 p-value
![Page 30: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/30.jpg)
Sub-mSv Imaging for Nodule Assessment with IMR
Khawaja R. et al, CT of Chest at <1 mSv: An Ongoing Prospective Clinical Trial of Chest CT at Sub-mSv Doses with IMR and iDose4 , JCAT 2014;38: 613–619
Conclusions: Sub-mSv IMR improves delineation of lesion margins compared to standard-dose FBP and sub-mSv iDose4.
![Page 31: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/31.jpg)
Improved Detection of PE with IMR
Kligerman. et al, Detection of PE on CT: Improvement using Model-Based Iterative Reconstruction compared with FBP and Iterative Reconstruction. J Thorac Imaging 2015;30:60–68
Standard iDose4
Kligerman. et al, Detection of PE on CT: Improvement using Model-Based Iterative Reconstruction compared with FBP and Iterative Reconstruction. J Thorac Imaging 2015;30:60–68
![Page 32: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/32.jpg)
Proven benefitsRigorous human observer studies evidence simultaneously lowering noise, dose, and improving low contrast detectability
Fast reconstruction timesMajority of reference protocols reconstructed in < 3 min
Model-based solution for cardiacIMR is currently only model-based iterative algorithm available for cardiovascular image reconstrcution
Installs250 sites with IMR by end of 2015
Scientific papers 30 peer-reviewed publications
IMR Highlights
3T MR
![Page 33: Advanced Reconstruction Methods on Philips CT Systemsamos3.aapm.org › abstracts › pdf › 99-27426-365478-111835-1013775324.pdfStandard Recon iDose4 Level6 IMR Level 3 * Image](https://reader030.vdocuments.mx/reader030/viewer/2022041112/5f187694209d69574d6fe9d1/html5/thumbnails/33.jpg)