1. Role of Adaptive Statistical Iterative Reconstruction (ASIR) in lowering radiation dose for pediatric head CT
Electronic Poster-eP 140
Azadeh Tabari, MD
Dhruv Singh,
Sarabjeet Singh, MD, MMST
Michael S. Gee, MD,
Sandra Rincon, MD
Paul Caruso, MD

2. Ethical Committee Approval
Financial Disclosure
None of the authors had any financial disclosures pertinent to this study
Ethical Committee Approval
The study was approved by the IRB and was compliant with HIPPA guidelines

3. Background
Iterative reconstruction has recently shown promising results for substantial CT dose reduction.
However, there are concerns about image quality at very low dose levels and careful selection of settings of the Adaptive Statistical Iterative Reconstruction (ASIR) is recommend to achieve optimal image quality.

4. Background
Filtered Back Projection {FBP} used since 1970’s with advantages
Speed
Closed loop solution
Ignores that fact that projection data is corrupted by quantum and electronic noise
Prorogates this noise and sometimes even amplify
ASIR
Takes into account these inconsistencies
Lower image noise and improves quality

5. Basics of Filtered “Back” Projection
Tube
Detectors
Patient
X-rays from one angle are recorded at detectors as measured projection data
As the tube rotates, x-rays from different source locations or “angles” results in a “set” of projection-data
Set of projection-data is called the sinogram

6. “filtered” back projection
Acquired sinogram data are back projected to obtain 2D image
Results in dense center and
blurred periphery
Hence the need to filter/preprocess the images
Projection 3

7. Image reconstruction Step 1: Data Acquisition Step 2: Reconstruction
One angle –one projection
Step 1:
Data Acquisition
Different angles-set of projections
(Sinogram)
Step 2:
Reconstruction
FBP
Iterative
Reconstruction
Better defined
mathematical estimates
Filtration
Back projection
Multiple iteration
Final image
Final image

8. FBP ignores actual dimensions
ASIR accounts for actual dimensions
FBP ignores actual dimensions
X-ray source/focal spot:
assumed to be infinitely small
Actual x-ray source/focal spot:
1 mm * 1 mm
Image voxel:
Depends on FOV and slice thickness
E.g. for 50 cm FOV and 1 mm slice thickness,
Reconstructed voxel size is 0.98*0.98*1 mm
Image voxel size ignored
Photon interaction assumed
To be at exact geometric center of detector
For 1mm detector cell spacing:
Active detector area is 80% of spacing 8

9. Purpose
The purpose of this study was to assess the potential benefit of ASIR in low dose pediatric head CT by comparing image quality with standard dose filtered back projection (FBP).

10. Materials & Methods
Institutional review board approval was obtained for this study and HIPPA guidelines were followed.
Study cohort was selected as all consecutive pediatric head CT performed on Discovery 750 HD (GE Healthcare) with ASIR 90%.

11. FBP and ASIR “Blending” interface
ASIR and FBP can be selected based on desired reconstruction
ASIR can be used as increments of 10%
ASIR review tool is available to select the desired levels of blending
Presently ASIR 30-50% is recommended by the vendor

12. Materials & Methods
For comparison, pediatric head CT examinations performed on scanners with standard Filtered Back Projection (FBP) was included.
Patient demographics, including
maximum skin-to-skin transverse head diameter
scanner information {mA, kVp} as well as
radiation dose information were recorded.

13. Materials & Methods
Effective dose was calculated as per the ICRP103 guidelines.
Image quality was assessed my measuring
image noise as standard deviation of HU values as well as
Contrast to Noise Ratio {CNR} for grey white matter.
Statistical analysis was performed with student t-test.

14. Results
Fifty-six children {average age 12.0 ± 4.0, M: F 32:24} underwent head CT examinations with ASIR 90 enabled protocol as compared to
82 head CT {average age 12.7 ± 4.7, M: F 46:36} with standard FBP reconstruction.
There was no significant difference in head diameter between ASIR90% {166.6 ± 10.5 mm} and FBP {170.5 ± 22.6 mm} (p = 0.07).

15. Results However, there was significant reduction of radiation dose of
52% with ASIR90% {CTDIvol 14.8 ± 10.4, DLP ± mGy.cm, 0.5 ± 0.4 mSv},
as opposed to FBP {CTDIvol 31.1 ± 17.4, DLP ± mGy.cm, 1.2 ± 0.7 mSv} (p <0.001).
Furthermore, objective image noise measure in white and grey matter with ASIR90% was found to be similar as FBP (p = ).
In addition, contrast to noise ratio {CNR} was not affected with ASIR90%, when compared to FBP (p = 0.67).

16. 14 yr girl with ischemic stroke and known infarct; hypodensity seen left caudate lobe, putamen, globus pallidus
Prior scan: (FBP)
Follow up: (ASIR)
Scan parameters
120 kVp
140 mA
Pitch: 0.531
ASIR:90
Estimated dose:
CTDIvol: 26.3 mGy
DLP: mGy.cm
0.9 mSv
Scan parameters
120 kVp
290 mA

17. 18yr old boy, s/p hemicraniotomy and right MCA infarct
Prior scan: (FBP)
Follow up: (ASIR)
Scan parameters
120 kVp
308 mA
NI: 35
Estimated dose:
CTDIvol: 52.1 mGy
DLP: mGy.cm
2.2 mSv
Scan parameters
80 kVp
100 mA
Pitch: 0.531
ASIR:100
Estimated dose:
CTDIvol: 5.4 mGy
DLP: mGy.cm
0.2 mSv

18. 18yr old boy, with subarachnoid blood along the b/l frontal convexities and small SDH along the left frontotemporal convexity
Prior scan: (FBP)
Follow up: (ASIR)
Scan parameters
120 kVp
159 mA, NI: 42
Pitch: 0.531
ASIR:90
Estimated dose:
CTDIvol: 25.4 mGy
DLP: mGy.cm
0.9 mSv
Scan parameters
120 kVp
231 mA
Estimated dose:
CTDIvol: 65.6 mGy
DLP: mGy.cm
2.4 mSv

19. Conclusion
Pediatric head CT could be acquired at 52% lower radiation dose with Adaptive Statistical Iterative Reconstruction {ASIR90%} without affecting image noise and contrast to noise ratio.

20. Thank you