Purpose
To evaluate the potential radiation dose reduction through image space iterative reconstruction (ISIR) algorithms in computed tomography (CT).
Methods and Materials
Five adult human cadavers were scanned with a standardized examination protocol (64x0.6 mm,
120 kV) at incremental doses of 20-200 mAsrot in steps of 20 mAsrot.
For image reconstruction standard soft and sharp convolution kernels (B20f,
B30f and B50f) and the corresponding ISIR kernels (I30f,
I50f,
IRIS,
Siemens,
Germany) were used (5/4 mm slice thickness/increment).
CT images were quantitatively analyzed for image noise and randomly presented to two radiologists,
who were instructed to define the minimal dose at which image quality was sufficient (criteria according...
Results
Noise values were as following for the different reconstruction kernels (increasing order): B20f:22.3±14.1,
I30f:21.9±26.7,
B30f:25.7±15.6,
I50f:43.8±28.4,
B50f:61±35.4.
All pairs of image noise values exhibit statistically significant differences (ANOVA,
p<0.0001).
The radiologist’s estimations of average necessary doses for the different kernels were: I30f: 97±44.7mAsrot,
B30f: 98.9±43.4mAsrot,
B20f: 99.7±44.3mAsrot,
I50f: 117±44.3mAsrot,
B50f: 132.3±45.8mAsrot.
Statistically significant dose reductions (12%) could be found for the combination of B50f versus I50f (p<0.007).
Conclusion
Image space iterative reconstruction lead to a significant reduction of image noise compared to the corresponding standard convolution kernels.
While in smooth convolution kernels only a slight dose reduction could be realized,
significant estimated dose reductions were found for the sharp B50/I50 kernels.
Thus,
ISIR will be a useful technique for bone and lung applications,
while the impact on soft tissue applications is less distinct,
warranting further investigations.