DQE should be measured according to IEC 62220-1. The measurement is involved and complex and should be performed by a certified lab. Here are the basic steps:
- Establishment of a Standard X-ray Spectrum
- Measurement Geometry
- Measurement of Air Kerma at Detector Surface
- Conversion of Air Kerma to Quanta per Unit Area (f)
- Measurement of Conversion Function
- Measurement of MTF
- Measurement of NPS
The formula per the IEC standard is : DQE(f) = MTF2(f) * f / NPS(f)
MTF and NPS must be obtained from images that have been linearized to the quanta per unit area at the detector input.
In terms of end use, MTF is specified as the highest number of line pairs per millimeter that are visible to the eye. This turns out to be an MTF of about 1%. One way to quantify this is to use an MTF estimate based on the slanted-edge method described in the ISO-12233 standard. The derivatives of the pixel values along each row/column perpendicular to the edge, are computed to obtain the line-spread-functions (LSF) perpendicular to the edge. The centroids of each LSF in the ROI are computed and a linear fit to the centroid locations is generated. The linear fit is then used to project the LSF data along the edge direction to the top line of the ROI; the resulting shifted data is “binned” by sampling at ¼ the original image sampling. A Hamming window is applied to filter out the noise, then the windowed, binned LSF data is subjected to a discrete Fourier transform (DFT). Finally, the normalized modulus of the DFT is computed to give the MTF result in percent as a function of the spatial frequency (in line pairs/mm). The final MTF is the average of the values at multiple slanted edges.
Take exposures starting at 20uGy, making a step increase in dosage with each exposure. Plot a line based on a least square fit of the average pixel values at each step. The saturation dose will be the lowest dosage at which the line becomes non-linear. Overall linearity can be specified as the maximum deviation from a straight line before saturation.
All sensor arrays have defective pixels that must be identified and hidden from view in the final product. The process of “blemish correction” is performed by the sensor’s firmware and usually requires some type of blemish map to tell the firmware where it will find the bad pixels. In design verification, clusters of bad pixels can be only so big, the total number must be limited, and the number that appear in a given row or column must also be limited. Since the blemish map is generated during manufacture of the sensor array, verification that each unit complies with the specification is also checked in production.
The interesting part is when new blemishes appear after production. This can easily happen during temperature and vibration testing or in life test. Every test that may have an impact on the blemish map needs to perform a test where a new blemish map is created and then mathematically subtracted from the map (correction file) created in production.
Take exposures in large steps from 100-2000 uGy. Determine the dosage at which blooming occurs.
Take exposures with a copper screen or a printed circuit board with fine copper traces in a grid pattern. Verify that the lines are straight on the image to a given limit in um.
Plot the average returned dark current pixel values (that is with no x-rays) over a temperature range of 0-50º C. Maximum dark current values should be specified in the 25-40º C range. Special firmware will be needed to do this test because raw pixel values must be returned on command from the host computer.
This is a test of the x-ray system's ability to maintain a consistent density across the image. Take an image with no phantom, at both low and high dosages that are within the sensor's linear range. Calculate the standard deviation of the pixel values across the entire image in each case. Specify uniformity as the maximum allowed sigma when operating the sensor inside it’s linear dosage range.
Compare two test phantom images, first with the sensor surface aligned normally with the x-ray source and then repeated with the sensor rotated about 30 degrees off axis. There should be no degradation in the side illuminated image.
Drill holes in an aluminum plate at different depths. Use a mechanical drawing to identify the hole depth at each location. Examine the test phantom images and record the holes that are visible in the image. Specify low contrast capability as the shallowest hole that is still visible. This method has an advantage over an aluminum step phantom, because the human brain tends to fill in missing information – in other words, we see steps in the image that are not really visible.
Take exposures at various dosages with no phantom, then calculate SNR by dividing the median (not the mean) by the standard deviation. SNR should be checked globally and also locally in various regions on the sensor surface. Minimum SNR must be specified at each tested dosage.
Most intraoral digital sensors trigger automatically when they detect x-rays. They do this by scanning pixels continuously until a certain number detect a reading. To verify triggering, the following functionality should be checked:
- Minimum dosage needed to detect x-rays
- False triggers that can occur when a cold sensor warms up in the patient’s mouth
- False Triggers caused by visible light
- Location and minimum surface area required to trigger
For test #1, sensitivity is specified as total dosage (typically in uGy) that will trigger a reading. In practice, this is tricky to test because the sensor requires a minimum dose rate to operate, but the x-ray source also has a minimum exposure time. For example, if the trigger specification is 5 uGy at a minimum dose rate of 700 uGy/s, the x-ray source would need to generate a 7ms exposure and the triggering algorithm in the sensor would need to react within 7ms. In practice, the sensor will usually trigger just fine a little below it’s minimum dose rate, so just use a DC x-ray source that can generate exposures under 15ms, then reduce the dose rate by moving the tube head further from the sensor or use an aluminum filter. The only other way to get a trigger sensitivity value is to do the following:
1. Measure the average pixel value at two different dosages
2. Calculate the conversion gain by taking the difference in pixel values and dividing by the difference in dosage.
3. For the next measurement, the digital sensor needs to be triggered by the controlling computer and not by the presence of x-rays. The x-ray source should be controlled by the computer so that the exposure can begin a short time prior to the start of the image transfer. This will insure that 100% of the radiation contributes to the average pixel reading. Save this reading for the next step.
4. Repeat step 3, this time letting the sensor trigger on the start of exposure. Take the difference in pixel values and divide by the conversion gain to get a trigger value in uGy
Since there are various ways to do this in firmware, a good triggering test will expose very small sections on the sensor surface to see if it has any dead areas. The last thing a dentist wants is a sensor that won't trigger because the patient has too many fillings.
For test #2, place the sensor in an ice water bath for about 20 minutes, then move it to a warm water (37 deg C) bath. Connect the sensor to the computer and verify that there are no false triggers. Take exposures at various dosages and verify proper triggering as the sensor warms up.
For test #3, place the sensor under a strong visible or UV light source. Connect the sensor to the computer and verify that there are no false triggers. Take exposures at various dosages and verify proper triggering.