Quality control test for electronic portal imaging device using QC-3 phantom with PIPSpro

Purpose :A Quality control (QC) test suitable for routinely daily use has been established for electronic portal imaging device (EPID) using PIPSpro software version 4.4 (Standard Imaging, Middleton, WI). It provides an objective and quantitative test for tolerable image quality on the basis of the high contrast spatial resolution, the contrast-to-noise ratio (CNR) and noise. Methods: The test uses a QC-3 phantom consisting of five sets of high contrast rectangular bar patterns with spatial frequeinces of 0.10, 0.20, 0.25, 0.43 and 0.75 lp/mm using 6MV and 15MV photon energy for accquiring high quality images. A “base line” value for the relative square wave modulation transfer function (RMTF), CNR and Noise data was obtained during a one week calibration period and one month test period. Results : Subsequent measurements shows significant deviations from baseline values, resulting in warning messages “potential problems in system performance”. The QC test uses high contrast spatial resolution and CNR for the system with acceptable performance. Conclusion: The method provides an automatic, objective, and sensitive measure of the system's imaging performance. This is a useful implementation during acceptance testing, commissioning, and routine quality control.


Introduction
Verification of the field placement is an indispensable part of a compressive quality assurance program for radiation oncology. Task group (TG) report 1 suggested by American Association of Physicists in Medicine (AAPM) recommends for acquiring portal images at least once in a week, which is frequently used in most of the clinical practices. With the recent expansion of electronic portal devices 2 (EPID), verification is now much simpler and could be carried out on a more frequent basis with the expectation of reducing gross field placement errors and increasing overall treatment accuracy.
However, the effectiveness of an EPID depends mostly on the image quality of the device to verify patient set-up and positioning prior to radiation therapy treatments. 3 Hence, it is equally important for device itself to maintain in its design and devise quality control (QC) tests to boost better image quality. These tests are essential by the manufacture at design, manufacture, operation, installation and by the user, all the way through the lifetime of the equipment. Nevertheless, as these systems are the part of regular clinical practice, it is very important to ensure the correct and reliable opera-tion of systems at all times. Lutz 4,5 and the Las Vegas 6,7 phantoms has been reported to check the accuracy and the image quality for the qualitative visual QA checks of patient imaging. 8 Here in, we report a less subjective approach for automatic daily quality control tests to get better image quality for patient. In a study presently underway at our institute, an integrated EPID-based QA system is being developed, which aims to replace the conventional device-dependent methods for daily and monthly QA tasks. In light of that, this investigation reports on the image quality, relative modulation transfer function, critical frequency and contrast-to-noise ratio obtained.

QC-3 PHANTOM
The QC-3(Standard Imaging, Middleton, WI) phantom is used to test the image quality from EPIDs. Megavoltage portal images are acquired with the phantom placed on the sur-face of an EPID at source-to-detector distance (SDD) 140 cm. The position of phantom can be at different distances such as 160 cm, 150 cm or at isocenter but need to maintain same distance always. A QC-3 phantom was designed for use in the test, which consists of five sets of high-contrast rectangular bars with spatial frequencies of 0.10, 0.20, 0.25, 0.43, and 0.75 lp/mm and dimensions of 13.5 × 11.3 × 3.6 cm 3 . A schematic diagram of the QC phantom is shown in Figure 1 (a). The diagram shows the numbered regions of phantom QC-3. The large numbers in the corners are used to for subjective quality control, as they are visible on the image with increasing density as (Number 1 is machined into a lead block to a depth of 1 mm, number 2 to a depth of 2 mm, etc.).
The small numbers indicate the region numbers. Regions 1 -5 are bars with different spatial separations, and are used for the analysis of the spatial resolution f50. Regions 6 -11 contain blocks of lead or plastic (PVC) with increasing thicknesses. With the EPID located under the patient (0 o in the Varian gantry coordinate systems), the phantom is placed on the top of the EPID detector housing in order to acquire test images. We prefer this location rather than the isocenter to minimize blurring due to the beam penumbra, since the test is intended to monitor the performance of the EPID and should be independent of the linac source size. The phantom is rotated to 45 0 relative to the EPID scan lines to prevent aliasing in the image of the bar patterns.

aS1000 EPID
The Varian aS1000 (Portal Vision, Varian Medical Systems, Palo Alto, CA) is an amorphous silicon flat panel imaging device mounted on a robotic arm. It has an active imaging area of 40 × 30 cm 2 (at an SSD of 105 cm). The image matrix is created from an array of 1024 × 768 pixels. The maximum frame acquisition rate is 9.574 frames/second, the permitted energy range is 4 -25 MV, and the permitted dose rates are 50 -600 MU/min. The detector has four main components. Inside the exterior plastic housing there is a Copper build-up plate, 1 mm in thickness. This is useful in MV imaging to absorb x-ray photons and emit recoil electrons. It also helps to improve the efficiency of the entire imaging system, by partially shielding the downstream components (including the scintillation screen) from scattered radiation. Underneath this plate lies the phosphor screen. In this EPID it is a Kodak Lanex Fast B scintillating screen, made up of a 0.4 mm thick Gadolinium Oxysulfide (Gd2O2S: Tb) phosphor.
This component absorbs the recoil electrons coming from the Copper plate, and transforms them into visible light. Below the phosphor, there is a 1024 × 768 pixel matrix, deposited on a 1 mm glass substrate. This constitutes the sensitive image forming layer of the photodiode system, and it is 1.5 μm thick. Each pixel consists of a Si n-i-photodiode to integrate the incoming light in charge captures and a thin film transistor (TFT) to act as a three-terminal switch for readout. The final major component is the accessory electronics, which drive the TFT switches and read out the charge captures. The gate driver powers the gate lines during the time that the data lines are feeding the accumulated charge to the read-out electronics.
When a voltage is applied to a gate-line, all of the TFTs in that row become transparent and the charge is then transferred to the data lines. Each row is read out in succession, and as one row is read the TFTs in the next row become transparent. External charge sensitive amplifiers capture the charge data. To form one frame of an image, a sequential readout of all of the rows is necessary.
Before each set of EPID images is acquired is it advisable to first calibrate the detector. This can be accomplished by obtaining a dark field and delivering a flood field. The premise is that taking these images will allow for the elimination of background noise and provide a uniform response for imaging. Specifically, the dark field image provides information about background noise, and is obtained by reading out each pixel in the absence of radiation. The resulting image, seen in, is a series of narrow vertical stripes, which result from the photodiode leakage current and varying electrometer offsets. The flood field image, on the other hand, is taken with the entire matrix exposed to a uniform dose. This allows the Portal Vision software to internally correct for individual pixel sensitivities. There is much to say about the acquisition and use of these images inside the Portal Vision software package.

Acquiring QC Images
The phantom is setup on the EPID at 140 cm, oriented at 45°t o the sagittal plane, as shown in the Figure 1(b). The large number 1 points towards the gantry. Lines and marks on the surface of the phantom assist in lining it up to the central beam line. For routine daily or weekly quality control, it is advisable to mark the surface of the EPID with paint or tape so that placement of the phantom can be done quickly. A stand is available for lateral imaging. Two images of the phantom are acquired under identical conditions, preferable during the same irradiation sequence. The obtained images were transferred to the PIPSpro software for analysis.
The analysis program places a region of interest (ROI) over each set of bars as shown in Figure 2. The frequency dependent square wave modulation transfer function (SWMTF) is determined by the method proposed by Droege 9 and the frequency for 50% modulation (f50) is compared with the predetermined critical frequency fc as a test of system performance.

FIG. 2:
A portal image of the QC phantom at diagonal orientation obtained with the aS1000 portal imaging system. The ROIs used for the QC test are marked on the image.

Determination of the RMTF
The SWMTF of an imaging system is defined as Where ΔE0 and ΔE (f) are the modulations of input to and output from the system. Since we are not interested in the absolute measure of the SWMTF, but only in determine day to-day variations in the system resolution, we use a relative measure 10  where, ΔE (f1) is the output modulation for the lowest frequency.
Usually the output modulation ΔE (f) is difficult to obtain from a noisy image; therefore, Droge and Morin 11 suggest using the relationship between signal amplitude and its variance. For sinusoidal output, (ΔE) 2 is proportional to the Variance (M) 2 within the ROI containing the bar pattern, and the above relation can be written as In the presence of random image noise (f) can be obtained by where, σm 2 (f) and σ 2 (f) are the measured total variance and the variance due to random noise, respectively. The total variance σm 2 (f) is obtained by measuring the variance of the pixel in the ROI corresponding the frequency f. In order to measure the random noise in an image, a pair of similar images are subtracted, and the standard deviation is obtained from the difference, thus avoiding contributions from fixed pattern noise. In this case, the variance of the subtracted ROI (σsub 2 ) will be σsub 2 = σ1 2 + σ2 2 where, σ1 2 and σ2 2 are the random noise variance of the ROIs for each image. We assume that these two variance are equal and hence; The variance σ 2 (f) is calculated once using above equation for the set of bars with the highest frequency (0.75 lp/mm) on the assumption that random noise is same for all ROIs.

Critical Frequency (f50)
Critical frequency is defined as the spatial resolution corresponding to 50% RMTF. This value is obtained using a piecewise linear interpolation of the RMTF graph to locate the 50% relative frequency response. (f40) is defined as the spatial resolution corresponding to 40% maximum of the relative modulation transfer function (RMTF). f30 is defined as the spatial resolution corresponding to 30% maximum of the relative modulation transfer function RMTF.

Contrast-to-Noise Ratio
A high quality image typically has a large CNR. This can be manipulated by increasing the Contrast, decreasing the noise, or a combination of both. CNR is defined by the following equation: where, Pbright is the average pixel value in the areas receiving the least radiation, Pdark is the average pixel value in the areas receiving the most radiation dose, and Noise represents the average noise value calculated from the uniformly irradiated regions.

Phantom Alignment
Electronic portal Imaging systems utilize non-square pixels giving rise to different MTFs in horizontal and vertical directions. By using the diagonal phantom orientation, a combine measure of resolution in both directions could be obtained, as well as reduction in error due to possible changes in ROI size and positioning. 9 RMTF (f) were exhaustively tested against changes in the ROI size and position. Images are acquired with a 6MV and 15MV photon beams when the phantom was physically displaced relative to the beam center by each of the four directions in a plane orthogonal to the beam. Changes in f50 and CNR were 0.86% and 0.72% respectively for 6MV and 0.27% and 0.8% for 15MV. Rotating the phantom by ± 5 0 from the correct 45 0 angle introduced changes in f50 and CNR of 0.22% and 0.82% respectively for 6MV and 0.27% and 1.12% for 15MV.
This displacement and rotations are much larger than any anticipated in normal use of the phantom for routine quality control measurements as mentioned in Table-1. The complete QC procedure was tested by repeating daily for 7 consecutive days. Deliberately moving the ROIs by two pixels (about 1.2mm at isocenter) to the left, right, up and down gave change in f50 and CNR of less than 0.3% and1.1%, respectively. Changing the field size and increasing the image acquisition times did not have a significant influence on the measured results.

Results
Measurements were made daily on the EPID imaging system with dual-energy linear accelerator by using 6MV and 15MV photons to 3-5 MU at images for high quality images distance with SDD 140 cm. System performances was monitored during a test period of 1 month and these data were used to determine the mean and standard deviation of f50, f40 and f30.
Before the QC needs to established Base line values for resolution, CNR and noise, after calibration of imager a series of one QC tests per day over the first one week. "base line" value is shown (Figure 3) by the blue line in PIPSpro software. The green area is within the acceptable parameters, the yellow area represents the caution levels, and the red areas represent the reject levels. Suggested values for accept and caution ranges are 5 and 10 percent respectively for all values except noise where 10 and 20 percent is suggested due to the fact that noise tends to fluctuate more than the other parameters.
During the one week calibration the mean values of f50, f40 and f30 (± standard deviation) were 0.463± 0.003, 0.602 ± 0.003 and 0.741 ± 0.002 respectively for 6MV. The mean values of f50, f40 and f30 (± standard deviation) were 0.365 ± 0.001, 0.457 ± 0.002 and 0.603 ± 0.002 respectively for 15MV. The mean value of CNR and Noise were 192.636 ± 1.348, 9.82 ± 0.094 for 6MV and 170.634 ± 1.464, 9.916 ± 0.135 for 15MV respectively. Figure 4 and Figure 5 shows f50 and CNR a plot of recorded on a daily basis for the as1000 system which included one month QC data for both high and low energy.  It is seen that system resolution at 6MV is superior to that 15 MV, an effect which has been observed previously and is a result of the larger physical beam penumbra at higher energys 12,13,14 and increased transmission through the bar patterns by the higher energy photons. It is seen that CNR for 6MV is higher than that for 15MV. Gantry angle at 90 o also checked during the one week calibration for both 6MV and 15MV.

Discussion
According to Task Group 142 report 15 : Quality assurance of medical accelerator, spatial resolution, Contrast, uniformity and noise should match to base line. In our case deviation of f50 CNR and noise were 0.22%, 0.70%and 0.47 for 6MV and 0.27%, 0.36%and 0.15 % respectively for 15MV from Base line which is meeting excellently. Quantitatively, specification value for Varian aS1000 f50 was 0.45 for 6MV and 0.379 for 10-25 MV. In our case spatial Resolution f50 for 6MV is 0.463 ± 0.003 and 0.365 ± 0.001 for 15MV which is very close to given data. Deviation for 6MV and 15MV were 2.2% and 3.4%, respectively. Spatial resolution of f40 and f30 following the f50 in proper sequence. The mean values of f50, f40, f30 (± standard deviation), CNR and noise excellently matching for both during the calibration period and test period for both of 6MV and 15MV.

Conclusion
We have established a QC test for portal imaging devices suitable for routine daily use for testing the system for acceptable performance in high contrast spatial resolution and CNR. It is shown that this method provides an automatic, objective, and sensitive measure of the system`s imaging performance and it is useful tool during acceptance testing, commissioning, and routine quality control.