Materials and Methods
This retrospective study was approved by the local research ethics committee of the Sultan Qaboos University Hospital.
The study included 83 adult oncology patients (33 males,
50 females; age range 28- 78 years,
mean 54.8 ± 11.9 years),
who were scanned twice in the calendar year 2018 before and after the optimization of CT protocols.
Both CT examinations were performed on the same Siemens CT scanner (Somatom,
Sensation 64).
The first image set consists of a scan performed before optimizing the scanning protocols.
The second set corresponds to image sets obtained after the optimization; i.e.,
from July 2018 onwards.
The optimization technique adopted in this study is to refine the exposure parameters (kV and reference mAs) according to the patient’s size in order to reduce radiation dose while maintaining the clinically acceptable level of image quality.
Two different CT protocols were developed depending up on the variation in the patients’ size (above 80 kg and below 80 kg) as shown in table 1.
For the sake of ensuring comprehensive assessment,
a wide range of patients’ weights (40 kg – 124 kg) was considered in this study.
The patients were grouped according to their BMI into three categories (normal,
overweight and obese) to assess the significance of the dose reduction and its impact on the image quality for each group.
The two sets of images (pre and post optimization) were randomly distributed into two separated folders,
each folder containing 83 chest CT examinations.
Each folder had a mix of pre and post-optimization scans. The images were evaluated by following the European Guidelines on Quality Criteria for CT imaging [17-18].
Before the actual evaluation,
a pilot of 20 cases was evaluated by both the radiologists together to form a consensus and also to maintain a consistent interpretation of the defined criteria.
The individual scans were independently and blindly assessed by two experienced radiologists using the visual grading analysis (VGA) method.
Visual grading technique is used to evaluate image quality by grading the visibility of reproduction of anatomical or pathological structures [19].
Hence,
each CT scan was graded into a five-point scale of absolute visual grading,
taking into account the overall noise impact on the image,
as shown in table 2.
Each set of images were graded by two radiologists and were analyzed by visual grading characteristic (VGC) analyzer [19-22].
The VGC analyzer offers the possibility to handle multiple-readers’ analysis and also to analyze paired data that are collected from two compared conditions (pre and post-optimization) using the same group of patients.
Radiation dose to the patient was monitored for each study by means of the two standard dose indicators CTDIvol and DLP.
They were extracted from the patient protocols associated with CT examinations.
The data were statistically analyzed to determine the significance of the dose reduction over the 3 groups.
Results and Discussion
This study has investigated the potential of reducing chest CT doses while preserving the image quality.
The adopted technique was to create two chest CT protocols; one protocol for patients who weigh less than 80 kg and the second protocol for patients above 80 kg.
These protocols were designed to scan with (100 kV,
140 ref mAs) and (120 kV,
120 ref mAs) respectively.
The other exposure parameters were kept unchanged. The CT technologists selected the CT chest protocol based on the patient's weight as entered in the hospital information system (HIS).
For few patients,
this data was not available in the HIS.
In those patients,
the CT technologists subjectively choose the protocol based on the perception of the patient’s size,
which remains one of the limitations of the study.
All data analysis was performed using SPSS version 23.0 for Windows (SPSS,
Inc,
Chicago,
IL),
MedCalc version 11.0 (MedCalc Software,
Mariakerke,
Belgium),
and VGC analyzer [21].
Radiation Doses
The results obtained from the comparison of the dosimetric data obtained from both sets of images (pre and post-optimization) have shown a significant reduction in radiation doses among the different BMI groups.
Table 3 summarizes the mean of the radiation doses (CTDIvol, DLP) calculated for pre and post optimization image sets with the resultant reduction within all groups
The reduction in CTDIvol,
as illustrated in figure 1,
ranges from 23.8% to 69.3% (mean of 54.0 ± 9.1%) in the normal group,
whereas for the overweight group the reduction varies from 34.4% to 62.9% (mean of 50.6 ± 7.2%).
The reduction among the obese group is slightly lower than the other groups ranging from 19.5% to 64.2% (mean of 43.6 ± 11.6%).
Overall the results have shown substantial dose reduction among all groups.
Image Quality
Interobserver agreeability was calculated using MedCalc software for the data analysis.
There was no agreement between observer 1 and observer 2 in pre-optimization (weighted kappa= 0.043,
95% confidence interval = -0.0565 - 0.143).
There was only a fair agreement between observer 1 and observer 2 in post-optimization (weighted kappa= 0.240,
95% confidence interval= 0.0385 - 0.442).
However,
the agreeability increased significantly (0.66) if score 4 and 5 Excellent and good (Table 2) are combined.
The VGC curve analysis is based on the trapezoid VGC curve,
and the Area under the VGC curve is 0.43835 (95% confidence interval = 0.3613- 0.5225 and p-value: 0.034).
The results suggest a significant difference in the image quality before and after the optimization of the CT protocols.
There is degradation of the image quality after the dose optimization,
which is statistically significant. The deterioration of the visual score in most of the patients is 4 (good) from 5 (excellent).
Both the visual grade scores four and five should be considered as clinically-diagnostically adequate images without loss of any relevant information.
Figure 2 illustrates two CT images corresponding to pre and post optimization acquired for the same patient (BMI 23).
The two observers had consensually agreed that both images of pre and post-optimization were rated 5.
Figure 3 shows the CT examination images for a patient (BMI 35),
where pre-optimization images had better quality (5 against 4) than post- optimization images.
Interestingly,
if we combined the scores 4 and 5,
the difference in the image quality pre and post-optimization was not significant (Area under the VGC curve: 0.49081,
95% confidence interval = 0.4759 - 0.5000).