This experiment shows how to use the scanning beam digital x-ray system to save dose by locally adapting the exposure to the opacity of the object rather than exposing it uniformly as done conventionally. The first step is to acquire a first pass reference image and define the parameters of the adaptive exposure algorithm system. The computer then outputs a new scan map with lower dose requirements, which is loaded in the system, so an equalized image can be acquired.
Dose area product measurements show that adaptive exposure leads to dose savings in the range of 30%We are developing an imaging system called the scanning beam Digital x-ray or SPDX system that enables inverse geometry x-ray fluoroscopy. The main difference between a conventional and an inverse geometry imaging system is that we are not using a single focus spot x-ray tube that projects onto large field of view detector. Our x-ray source consists of thousands of focus spots with each coated onto a small photon counting detector to acquire a single imaging frame.
We acquire thousands of detector images with each detector image only viewing a small portion of the object inverse. Geometry imaging leads to a significant reduction of scatter in the images. This allows us to acquire these images at much lower dose.
This is in particular of advantage for larger patients as scatter scales with patient size. In pediatric patients, dose savings due scatter reduction are much less. However, inverse geometry allows for implementation of dose saving strategies that also benefit pediatric patients.
Inverse geometry enables a technique that we are calling adaptive exposure. As we are acquiring thousand of detector images, we can adapt exposure in real time, depending on the upper city of the region exposed, so we can reduce exposure in radiolucent area and maintain exposure in some more OPAC region. Therefore, we can achieve uniform image quality in the overall image and reduce dose in area that are overexposed in conventional systems.
To monitor the radiation dose, a dose area product meter is placed in front of the collimator. The patient or phantom is typically positioned relatively close to the collimator with the center of the imaging volume at about 40 centimeters from the collimator. Next, the system operating mode is selected.
Typical settings are field of view of seven inches and frame rate of 15 frames per second. The X-ray source peak voltage is set to 80 peak kilovolts and the source power to 17 kilowatts. Then turn on the SBDX system and start image acquisition.
During image acquisition, an electron beam is sequentially scanned to every focal spot position in a raster fashion. At every focal spot position, the electron beam hits the transmission target and generates x-rays. The resulting x-ray photons pass through a focusing collimator and illuminate the detector.
Thus, only a small portion of the imaging volume is projected. During illumination, the detector acquires a detector image, which is directly stored in the system memory. For each focal spot, an illumination of up to eight microseconds is desired.
However, due to thermal limitations of the target, the maximum continuous exposure is one. Microsecond and acquisition is broken up into eight one microsecond increments. In total, over about 60 milliseconds, over 40, 000 detector images are recorded at the end of image acquisition.
Be sure to record the measurement on the dose area product meter. The SBDX is intrinsically a tomosynthesis system, as the object is illuminated under different angles from each source focal spot, any plane within the imaging volume located between collator and detector can be reconstructed. Begin by setting the imagery.
Construction parameters. Select how many planes to reconstruct and their location. Next, run the image reconstruction algorithm.
The algorithm processes each detector image by first scaling the detector image for the respective reconstruction plane, and second, shifting the image according to its focal spot position before adding to the reconstruction. Next, a post-process filtering removes the pattern created by the shift and add operation. Typically, 32 planes with the spacing of 0.5 millimeters are reconstructed in a stack of reconstructed images.
Structures go in and out of focus depending on their distance from the detector. The plane selection algorithm selects the structures that are in focus in every plane, and with that information, a composite image can be formed that only shows structures and focus. This 2D image is equivalent to images obtained on a conventional system.
However, the plane selection algorithm has determined for every structure where it is in respect to a collimator. This allows us to play it as a 3D animation, Load the saved images into the adaptive exposure simulator and run the algorithm for each focal spot. The number of photons in the image is determined.
The images are then accumulated until the target number of photons is reached, or all eight rescans are added. The output of the adaptive exposure simulator is a re-scan map that details how many times each focal spot position is illuminated. This map gets merged to the operation mode file used to run the SBDX system.
After loading the updated operation mode file into the SBDX system, taken images before only now the x-ray beam will be turned on or off at the focal spot positions according to the re-scan map, reducing the total number of illuminations, run the image reconstruction algorithm on the new data to reconstruct the equalized image. Again, be sure to record the dose area product reading dose measurements with the dose area product meter demonstrated a dose savings of 30%in the equalized image using the re-scan mask. Preliminary results showed that the image quality is equivalent and that the equalized image has more visual appeal than the non equalized image.
We are exploring adaptive exposure as a technique to reduce dose and fluoroscopic procedures. Initial results show that dose savings of at least 30%are feasible at equal image quality. We are in the process of implementing adaptive exposure as a real time feature.
During image acquisition, our algorithm will automatically perform thresholding. Therefore, many of the steps shown here will be invisible to the user. Our current results are encouraging and we are looking forward to investigating our technique Further.
In particular, we will have to show in clinical trials that the diagnostic value of our images is equivalent or better than conventional images.
