In this initial assessment,
we have demonstrated that a prototype automatic kV selection tool can be used to select the most dose-efficient kV for a particular patient,
based on their size,
diagnostic task,
and scanner limitations while insuring that diagnostic image quality is maintained. This kV prescription also selects the corresponding automatic exposure control settings (i.e.,
quality reference mAs) that is needed in order to achieve a desired image quality defined by the reference scanning technique and the contrast gain setting (noise constraint) in the Auto kV tool. For the large majority of patients,
tube potential will be changed to 100 kV,
resulting in a corresponding dose reduction of about 23%.
In a very recent study by Winklehner et al,
the dose reduction and image quality of thoracoabdominal CTA using an automated kV selection tool were evaluated [18].
They observed an overall dose reduction of 25.1% if including all kV patients and 39.3% if only including 80 and 100 kV patients.
The lower dose reduction observed in our study is expected because of the strong dependency of dose reduction at lower kV on diagnostic tasks.
In CTA examinations,
because the diagnostic task is to visualize high-contrast vessels and the improved iodine contrast enhancement at lower kV allows a greater noise increase to be tolerated,
so the strength setting of Auto kV can be high.
This corresponds to a high contrast gain and a greater radiation dose reduction.
In contrast-enhanced abdominopelvic examinations,
the detection and characterization of lesions may benefit from higher enhancement of iodine at lower kV,
but cannot tolerate as much noise increase.
Therefore,
the strength setting of Auto kV for abdominopelvic examinations is weaker than in CTA examinations.
The basic physics principle of the dependency of kV selection and dose reduction on diagnostic task was recently described in detail by Yu et al [19].
Because the appropriate strength setting of Auto kV and the potential dose reduction depend on both patient size and diagnostic task,
more extensive clinical validation across a range of diagnostic tasks is required.
It should also be noted that the selection of the optimal kV by the current Auto kV software is based on the traditional filtered backprojection reconstruction method without application of any modern noise reduction methods.
With the development of iterative reconstruction and various noise reduction methods,
image noise can be reduced substantially without sacrificing image spatial resolution,
which further improves the contrast-to-noise ratio for a given radiation output and may allow the use of lower kV on relatively larger patient sizes and more radiation dose reduction.
Therefore,
with the use of iterative reconstruction and noise reduction methods,
the parameter setting of Auto kV may need to be further adjusted.
There are several limitations to our study.
To calculate radiation dose reduction,
each patient served as his or her own control.
Whilst we know the actual CTDIvol of the scan that was performed with Auto kV,
we used the scanner-generated estimate of CTDIvol for the dose of the base protocol exam at 120 kV,
and this estimate may contain error,
as there is typically a 1-5% difference between prescribed CTDIvol and actual delivered CTDIvol.
However,
this error is considerably below the dose reduction that we were able to demonstrate,
and should not materially affect our results. In our secondary analysis,
we also estimated dose reduction by comparing to the CTDIvol of size-matched controls who were scanned with the same slice thickness and phase of enhancement. Size was considered equivalent if the patient width matched to within 5 cm,
measured skin-to-skin at the level of the superior border of the liver on the coronal topogram. However,
patients come in a variety of shapes,
and body fat distribution is not uniform. Even though the patients and controls were matched for width at one level,
they may have had non-matching widths at other scanned levels.
Furthermore,
measuring the lateral width of a patient does not account for the antero-posterior dimension,
so there is likely some inaccuracy associated with size-matching of patients. These measurement errors underscore the inadequacy of simple size-based technique charts and need for a semi-automated tool.
There were 2 patients who experienced a dose increase of about 20% compared to what was prescribed with the reference protocol at 120 kV,
as shown in Figure 1B.
These were both large patients in whom,
with the 120 kV reference protocol,
the scanner output was limited by the system tube current limit,
which was lower than that requested by the AEC software. After loading the scan,
the technologist was prompted to reduce the pitch to meet the requirement of the AEC software,
resulting in the dose increase.
Therefore,
the CTDIvol recorded before the pitch adjustment underestimated the reference CTDIvol.
In conclusion, we have demonstrated that with the use of an automated kV selection tool,
lower doses can be achieved whilst maintaining satisfactory image quality,
diagnostic confidence,
and unchanged iCNR.
The Auto kV tool was effective in the majority of patients,
and has the greatest benefit for dose reduction in small and medium sized patients,
in whom scanning with 100 kV can usually be achieved.
The automated kV tool can be introduced into routine clinical practice for abdominopelvic CT.