There are differences in noise in the four protocols studied. The lowest noise is obtained in variation IV, while the highest noise is obtained in variation II. Variations I and III with the same measured CTDIvol value produce the same noise value (table 4). Difference in measured CTDIvol value up to 2.74 mGy does not produce a significant noise difference. Noise value decreases if the measured CTDIvol value increases, and results in the same noise value at the same measured CTDIvol value. This is consistent with the results of previous studies which stated that noise is influenced by tube current and rotation time (effective mAs) and pitch, where the three parameters will also affect the radiation dose [20]. Higher doses of radiation will produce lower noise.

There are differences in SNR in the four protocols studied. The highest SNR was obtained in variation IV, while the lowest SNR was obtained in variation II. Variations I and III have the same measured CTDIvol and noise values, but have different SNR values. SNR in variation III is higher compared to variation I (table 4). SNR is calculated by dividing the HU value of the object with noise. Variation III uses a faster rotation time of 0.75 seconds compared to variation I which is 1 second. Fast rotation time will reduce motion artifacts [16]. In CT, artifacts are defined as the systematic difference between the HU value and the attenuation coefficient of the actual object. Motion artifacts can affect the radio density measurement or HU on CT images. The signal measurements in artifacts area results lower radio density, this causes the object's HU value at variation I to be lower than variation III. The post hoc test results showed that a significant SNR difference was seen between variations I and IV also between variations II and IV. Variations I and II used the same effective mAs from the same tube current and rotation time variations of 150 mA and 1 second rotation time, while the variation IV uses greater effective mAs that is 187.5 mAs from the variation of the 250 mA and the rotation time of 0.75 seconds.

There were differences in anatomical structures scores in the four protocols studied. The highest anatomical structures score was obtained in variation IV, while the lowest was obtained in variation II (table 4). Of the seven questions in the anatomical assessment questionnaire, which showed a difference (p value <0.05) was the visualization of the basal ganglia and corpus callosum, while the visualization of the thalamus, cerebellum, cysterna, and ventricular system also the differentiation of white matter and gray matter did not differ (p value> 0.05). Post hoc test results state that significant differences in anatomical structures scores are seen between variations II and IV and between variations III and IV. Variation II uses a tube current of 150 mA which in theory will produce higher noise and lower SNR compared to variation IV which uses a tube current of 250 mA [16]. Meanwhile, although variation I and III use the same effective mAs which is 150 mAs, and variation I has a slightly lower SNR value, but variation I has a higher anatomical structures score. This happened because the difference in SNR values ​​between variations I and III were not significant, and because the anatomical structures score in this study was assessed subjectively by radiology specialists, so the assessment results were strongly influenced by perception and visual observer ability.

There was no difference in pathological scores in terms of visualization and location infarction in the images of the four protocols (p value> 0.05). But based on the mean value, the highest pathological score was obtained in variation IV, while the lowest value was obtained in variation II (table 4). In this study, images from all four variations were able to visualize infarction on head CT images. All head CT images can be received diagnostically and can be interpreted by radiology specialists. Measured CTDIvol reduction of 33.08 - 37.04 mGy or 46-51% (from the initial protocol 78.7 mGy) did not result in differences in the pathological score of infarction. Previous studies suggest that a reduction in dosage of 30-40% still produces images that can be received diagnostically, whereas other studies say a reduction in dosages of 60% is still possible without loss of diagnostic quality, but images produced with surface doses of less than 30 mGy cannot be interpreted [14].

There was no difference in low contrast resolution either based on visual assessments by medical physicists and based on CNR measurements between a 25 mm cylinder and the background (figure 2).

Figure 2 Image of Low Contrast Resolution Phantom (a) Variation I ,( b) Variat ion II, (c). Variation III, (d) Variation IV

Based on the CNR mean values, the highest low contrast resolution was obtained in variation IV, while the lowest was obtained in variation II (table 4). The CNRs of the four protocols were more than 1.0, meet the ACR standard. While based on a visual assessment by three medical physicists, the four protocols showed the same result that the cylinder group that is clearly visible is cylinder group with a diameter of 5 mm (table 4), which meaning that it still meets the ACR standard (at least a cylinder group with a diameter of 6 mm must be seen in the image) [3]. The ability of CT images to detect and characterize low contrast lesions cannot be compromised in an effort to reduce the level of radiation dose [21]. Low contrast resolution is strongly influenced by noise, so parameters that affect noise will also affect low contrast resolution, including mA, rotation time, and pitch. In this study, visualized low contrast resolution showed no difference in the four variations even though there was a statistical difference in noise. Because the difference in noise from the four protocols is small, proportional to the dose difference which is also small, that is 1.22 - 3.96 mGy (3.2% - 11.4%). This is in line with the results of other studies which state that a 25% - 50% lower dose difference results in lower contrast resolution [21].

The evaluation of three medical physicists on the low contrast resolution phantom image stated that the four protocols showed the same results, the line pairs that were able to visualize the lines and the space between them clearly was 8 lp / cm. This means that the four protocols still meets the ACR standard which states the pairs of lines must be clearly visible on images of at least 5 lp / cm for abdominal and head protocols, and 6 lp / cm for high resolution protocols on the thorax CT [3]. Spatial resolution is influenced by several factors, including focal spot size, detector size, pixel size, reconstruction algorithm and pitch [22]. The size of the focal spot is related to tube current. Large focal spot size is used on large tube current, and vice versa. To compensate for the potential reduction in spatial resolution due to this, a longer rotation time and lower pitch are used, but this will affect the length of examination and radiation dose [20]. In this study, protocols I, II and III use "small" focal spots, whereas in variation IV use "large" focal spots because of the use of large tube currents (250 mA). Variation IV also uses a greater pitch compared to variations I, II and III. But the results of the spatial resolution assessment on variation IV do not differ from variations I, II and III (table 4). Because the spatial resolution assessment is a subjective assessment. So the assessment results are strongly influenced by the observer's perception and visual ability. Previous research stated that tube currents from 30 to 190 mA with increments of 20 mA and kV remained, did not produce different spatial resolutions.

The measured CTDIvol in this study is lower than the monitor CTDIvol with a deviation range between 8.98% - 9.02% (table 2). The measured CTDIvol in this study is the average of three measurement results.

Table 2 CTDIvol Measurement Results

The monitor CTDIvol value for each CT scanner is a value that has been set by the manufacturer, which is different for each brand and type of CT scanner [23]. The deviation between the monitor CTDIvol and the measured CTDIvol depend on the CT scanner performance, so it is important to periodically evaluate the accuracy of the CTDIvol value. The deviation between the measured CTDIvol and monitor CTDIvol value in this study can be used as a reference in optimizing the image quality in the other indication of head CT examination, because the results of three measurements for four different parameter variations show a consistent deviation value of 9 ± 0.2 %.

There is a correlation between measured CTDIvol and noise with a moderate level of relationship, an increase in measured CTDIvol will cause a decrease in noise (table 3). This is consistent with the statement that the dose affects the noise, where the decrease in the dose that occurs due to the setting of the CT scan parameters, will increase the noise value in the image [20]. In this study, if the measured CTDIvol value increases by 1 then the noise will decrease by 0.078. The results of previous studies stated that a dose reduction of 10 mGy caused an increase in image noise of 0.5 HU to 1.8 HU with a correlation coefficient between 0.62 and 0.97.

There is a correlation between measured CTDIvol and SNR, with a moderate level of relationship, an increase in measured CTDIvol will cause an increase in SNR (table 3). This is consistent with the statement that the use of higher doses will result in a decrease in image noise, which will also affect the value of the image SNR because the SNR is calculated by dividing the signal object with noise. In this study, if the measured CTDIvol increases by 1, the SNR value will increase by 0.691.

There is a correlation between measured CTDIvol and anatomical structure scores, with a weak relationship level, an increase in measured CTDIvol will cause an increase in anatomical structure scores (table 3). In this study, if the measured CTDIvol value increases by 1, the anatomical structure score will increase by 0.047. The results of this study are in accordance with the results of previous studies which state that the increase in measured CTDIvol by 25% only slightly increases the value of the anatomical structures scores of the image, with a tendency to increase weakly [25].

There is a correlation between measured CTDIvol and pathological visualization score, with a weak relationship level, an increase in measured CTDIvol will result an increase in pathological visualization scores (table 3). If the measured CTDIvol increases by 1, the pathological visualization scores will increase by 0.049. Previous study of radiation dose reduction states that decreasing the dose by 30-60% still produces images that can be received diagnostically, but in these studies no further correlation analysis was performed.

Correlation test between measured CTDIvol and low contrast and spatial resolution cannot be done because the value of both low contrast resolution and spatial resolution is constant. Measured CTDIvol values with a range of 34.56 mGy - 38.52 mGy or a difference of 3.96 mGy does not produce different low contrast resolution and spatial resolutions.

Physical image quality (noise & SNR) and clinical image quality (anatomical structure & pathological visualization scores) is shown in table 3.

Table 3 Measured CTDIvol and Image Quality Correlation and Phys ical and Clinical Image Quality Correlation

Table 3 shows that there is no correlation between physical image quality (noise and SNR) and clinical image quality (anatomic scores structure and pathological visualization). Because in this study the value of noise and SNR have small difference. The small difference in noise and SNR due to the selection of CTDI values in this study has a slight difference, because it considers the range of examination parameter values recommended by the international radiology and medical physics institute / commission. The small difference in HU in the image will not be seen visually difference. And in this study, anatomical structure scores and pathological visualization scores are assessed subjectively which is influenced by the ability of the observer's eye and observer's competence in assessing the visibility of anatomic structures and pathological abnormalities in the image.

Although in this study there is no correlation between physical quality (noise, SNR, low contrast resolution, and spatial resolution) and clinical quality (anatomical structure scores and pathological visualization scores), it is still important to measure physical quality and clinical quality. Physical quality measurements can be used as consideration and added value in analyzing contrast and detail quality in CT images, while clinical quality measurements must be performed to fully represent the actual clinical image quality performance [24], then used as consideration in determining best protocol as an effort to optimize the radiation dose in a CT examination procedure. The tabulated values for measurement and assessment of noise, SNR, low contrast resolution, spatial resolution, anatomical structure scores, pathological visualization scores, and measured CTDIvol are as in table 4.

Table 4 Measurement Result of Noise, SNR, Anatomical Structure Score, Pathological Visualization Score,

Low Contrast Resolution, Spatial Resolution, and Measured CTDIvol

To determine the best protocol of tube current, rotation time, and pitch variation, which is capable of producing optimal image quality and measured CTDIvol values not exceeding BAPETEN recommendations, normalization of noise data, SNR, and anatomical structure scores is performed. Pathological visualization scores, low contrast resolution and spatial resolution data were not normalized because there were no differences in the four treatment groups, as well as the measured CTDIvol values because the values of the four treatment groups did not exceed BAPETEN recommendations of 50 mGy. The results of normalization of noise data, SNR, and anatomical structures score are shown in table 5.

Table 5 Normalization of Noise, SNR, and Anatomical Structure Score

CT is a radio diagnostic examination that uses exposure to low radiation doses, with a dose range of 5-50 mSv. But the potential effects of radiation due to CT radiation exposure should not be ignored because no matter how low the radiation dose is received, there is always the opportunity for changes in biological systems at both the molecular and cellular levels. In addition, other studies reveal that exposure to low-dose radiation triggers induction which causes instability of genetic material after exposure which may play a role in the process of cancer formation. So that optimization of radiation doses is absolutely necessary, even at low doses of radiation exposure.

In this study, the use of faster rotation time on variation IV (0.75 seconds) compared to variations I and II (1 second) still resulted in higher noise, SNR, anatomical structure scores and pathological visualization scores compared to variation I and II. The helical pitch used in this study is still in the range recommended by IEC and AAPM. Although the helical pitch in variation IV is greater than variation I, II, and III which theoretically have the potential to increase the emergence of streak artifacts, this does not make noise, SNR, an atomic structure scores, and pathological visualization scores on CT images variation IV become lower. Because effective mAs variation IV is higher (187.5 mAs) compared to variations I, II and III (150 mAs). The effective mAs value is linear with the radiation dose, if the effective mAs is lowered then the radiation dose will also decrease, and the reduction in the radiation dose will increase the noise and subsequently will affect the image quality.