|Home | About | Journals | Submit | Contact Us | Français|
Computed Tomography (CT) has become an effective diagnosis and evaluating tool in clinical; however, its radiation exposure has drawn great attention as more and more CT scans are performed every year. How to reduce the radiation dose and meanwhile keep the resultant CT images diagnosable becomes an important research topic. In this paper, we propose a dose reduction approach along with the adaptive bolus chasing CT Angiography (CTA) techniques, which are capable of tracking the contrast bolus peak over all the blood vessel segments during the CTA scan. By modulating the tube current (and collimator width) online, we can reduce the total radiation dose and maintain the contrast to noise ratio (CNR) of the blood vessel. Numerical experiments on reference DSA data sets show that by using the proposed dose reduction method, the effective radiation dose can be saved about 39%.
With the advent of multi-slice CT scanner and the sophisticated CT reconstruction algorithm, CTA is increasingly used to evaluate patient with vascular diseases [1–3]. It has many advantages over the conventional catheter angiograms: 1) non-invasiveness, 2) short acquisition time, 3) multiple viewing methods, and 4) low cost. However, the radiation exposure of CT scan has become of a wide concern due to its induction to genetic, cancerous and other diseases [4, 5]. According to , CT delivers 2/3 of the total radiation dose every year in US, and this number is climbing. On June 19th, 2007, the New York Times reported that “the per-capita dose of ionizing radiation from clinical imaging exams in the U.S. increased almost 600% from 1980 to 2006”. Berrington  reported in 2004 that diagnostic X-rays, indicating that radiation from medical and dental scans, causes about 700 cases of cancer per year in Britain and more than 5,600 cases in US. To that end, every effort to reduce the radiation dose is appreciated and the well-known ALARA (As Low As Reasonably Achievable) principle is widely accepted in the medical community.
There are many studies have been reported on dose reduction for CT scan. In , Wintersperger and his colleagues used lower kilo-voltage and claimed the resultant image had similar sensitivity and accuracy. In , lower tube currents were used and the resultant image quality was claimed to be good enough for diagnosis. Generally, dose reduction is contrary to the image quality. In other word, reducing the radiation dose will sacrifice the image quality. However, with the advanced CT reconstruction algorithm [9–12] and post-processing techniques [13–16], the resultant images could be held to certain diagnosis level while the radiation dose is reduced lot. Recently, control technique is found very useful in CT. It can be applied to improve the CT image quality [17, 18], and to reduce the radiation dose reduction, for example, Automatic Exposure Control (AEC) program, which modulates the tube current to maintain the same x-ray attenuation level according to the scan angle/position of the human body.
In this paper, we are going to present a radiation dose reduction approach using adaptive control techniques. It is specifically for CTA scan and is based on the adaptive bolus chasing techniques. The adaptive bolus chasing techniques track the contrast bolus peak during the CTA scan, and ensure every segment of the blood vessel be scanned with the possible highest bolus density inside it. To that end, it is very likely that the blood vessel has a much higher CNR, which is better than enough for diagnosing. With the ALARA principle, it is straightforward to modulate the tube current and/or vary the table increment to reduce the radiation dose while keeping the resultant CT images at the same diagnosis level.
The proposed dose reduction approach is made possible by the adaptive bolus chasing CTA techniques, which will be briefly introduced in this section. More details can be found in .
The contrast bolus is used to enhance the vasculature so that it can be distinguished from the surrounding soft tissue, which has similar physical property of the vasculature [20, 21]. During the CTA san, it is desirable to synchronize the contrast peak and CT imaging aperture, in such a way, the CNR is maximized and the vascular diseases on resultant CT images are better shown. However, the current scan method uses a pre-set constant speed to move the CT table, which is problematic due to the complicated bolus dynamics influenced by many factors such as, patient characteristics, injection patterns, and vascular diseases . The contrast bolus rarely flows constantly along the blood vessels. To that end, the current constant-speed method either needs a large amount of contrast volume or results in lower CNR vascular images due to the unmatched the longitudinal position of the bolus peak and CT imaging aperture.
The adaptive bolus chasing techniques are illustrated in Fig. 1. The patient is feeding into the CT gantry of the speed of the CT table. The real-time CT images are reconstructed and meanwhile the intravascular bolus density information is fed into the adaptive controller, which is used to predict the bolus peak time for the next position and sends the signal to the table driver to vary the speed. Thus, the synchronization of the bolus peak and CT imaging aperture will be realized.
Generally, the effective radiation dose is affected by the tube voltage, collimator width, tube current, scanning time, and scan volume . The following procedure is used to estimate the effective radiation dose for a CT scan.
where EDLP is the region-specific conversion factor, and it equals to 0.015 for abdomen and pelvis .
The above calculation is based on the constant-pitch (speed), which is not the case for the adaptive bolus chasing CTA, which varies speed during the scan. To that end, we need to modify the formula accordingly. In step 2), CTDIvol is computed at every step (gantry rotation time) under assumption that gantry rotation time is kept constant
Accordingly, the DLP is changed to
Where Ti is the table increment at each step and N is total number of gantry rotation.
It seems that Ti is not related to DLP in Eq. (6); however, Ti affects the total number of gantry rotation. Higher average Ti results in smaller N. With all other parameters kept constant, total DLP is smaller.
Dose reduction can be achieved under the condition that the resultant CT images are good enough for diagnosis, which is subjective. Different physicians may have different diagnosis results for the same set of CT images. On the other hand, CT image quality is affected by many factors, such as, image noise, signal to noise ratio, antifacts, and resolution. It is hard to find a standard that quantifies the image quality. In this paper, we focus on the CNR/SNR, which is important for a CTA scan. Meanwhile, we restrict the spatial resolution (slice thickness) within some range.
The idea of the dose reduction approach is as follows. Since the proposed bolus chasing CTA techniques always track the contrast bolus peak, and the scanned blood vessels are well enhanced, according to the literature, most of the segments have CT number greater than 250 HU. On the other hand, “the minimum adequate goal for contrast opcification of arterial system is 150–200 HU” . Therefore, we can modulate the tube current (and the collimator width) to achieve the desired CNR and reduce the radiation dose meanwhile.
In this subsection, we will reduce the radiation dose using adaptive-chasing method only through modulating the tube current to maintain the minimum CNR at each step. Varying the tube current and collimator width simultaneously will be presented in the next subsection.
Since we only modulate the tube current and the collimator width is kept constant, the image noise and CT number of blood vessel at step i is denoted by and Ci, respectively, then the CNR at step i is under assumption that the CT number of the surrounding tissue is zero (or a constant value). In the adaptive bolus chasing techniques, we use a two-variate 2nd order polynomial to approximate the bolus density function,
where t and z represent the time and distant starting at the current time and position, respectively. Substitute t = Tr= Δt and replace z with Ti (table increment at the step i), then we have
where ā0 =a0 (ti, zi) + a1 (ti, zi)Δt + a3 (ti, zi)Δt2, ā1 = a2 (ti, zi) + a4 (ti, zi)Δt, and ā2 =a5 (ti, zi).
Now, we can formulate the problem:
Minimize the radiation dose (DLP) normalized over the table increment at every step, , under condition that CNR is kept constant or above a designated value, that is , and the longitudinal resolution is not worsen very much, which is quantified by the slice thickness. The optimal solution will be given when .
Lagrange multiplier is used to solve this problem
The solution of the above problem is
And the corresponding estimated bolus density at Ti is
Generally, the bolus density profile is bell-shaped, to that end, ā2 is negative and ā1 and ā0 are positive. Without the dose reduction optimization, , which maximizes the bolus density at the next step. However, the table increment in this paper is . It is reasonable because table increment is expected to be maximized under the condition the next scan affords enough CNR.
Adaptive bolus chasing with dose optimization procedure through modulating tube current only
This algorithm alone does not guarantee that the whole vasculature length be scanned. However, by setting the minimum moving distance at each step, it is no longer a problem. This is reasonable because scanning the vasculature forever contradicts minimizing the radiation dose.
Adaptive-chasing method varies the table increment Ti at each step, and changes the pitch p accordingly, since , where W is the collimator width. Under condition that the collimator width is constant, the slice thickness, which depends on the collimator width and table increment (see (12)), also varies. This is not good to the image quality in the sense that the longitudinal spatial resolution is not uniform. To that end, we propose to modulate the tube current and collimator width simultaneously to reduce the radiation dose and maintain the slice thickness constant.
Assume uniform slice thickness, S0, is expected for the whole scan. For a single slice CT and half-scan reconstruction method, the slice thickness is given by
To keep slice thickness constant, we need to vary the collimator width at each step
Adaptive bolus chasing with dose optimization procedure through modulating tube current and collimator width
We apply the proposed dose reduction approach on the referenced digital subtraction angiogram (DSA) data sets of four patients, among which, two had occlusive diseases; one had abdominal aneurysm and one was normal. To show the advantage of the proposed method, we also apply the conventional constant-speed method on the same referenced data sets. The high frame DSA data sets were collected in University of Iowa Hospital and Clinic (UIHC). They recorded the bolus information in the abdominal-aorta to the femoral artery from the time of bolus injection to the end. We extracted the bolus information along the blood vessel from every frame and formed the bolus 3D profile (position-time-density), which would be used as the referenced bolus information. It can be seen that the CTA data sets do not have this ability due to its narrow FOV in the longitudinal direction short acquisition time.
The CTA scan is assumed to be performed on the Siemens SOMATON Volume Zoom four-row detector CT scanner, and spiral scan mode is used. The peak Kilo-Voltage is set to 120 kV for both adaptive-chasing and constant-speed method. As for constant-speed method, we set the effective tube-current and collimator width to be 240 mAs and 10 mm, respectively. The resultant CNR of the vasculature is expected to be greater than or equal to 200 HU over the noise level generated by the above voltage, current and collimator settings.
During the simulation, the constant speed is set as 30mm/sec for all four cases. For adaptive-chasing method, the collimator width is fixed at 10mm and the pitch varies between 0.5 and 2, which also limited the CT table speed. As for the case of varying collimator width, the scanned HU results will be same. The difference is on the reconstructed CT image, and it is not discussed in this paper. Both methods use auto-trigged technique to start the CT table, i.e., monitor the HU in the designated region of interest (ROI), and starts the CT table when the HU reaches the pre set threshold.
The effective radiation doses for four patients are estimated in the way aforementioned in Section 2.2. To show the performance of the adaptive-chasing method and the constant-speed method, we use the performance index (PI), which is defined as
where B(zk, tk) is the actual maximum density at position zk, which happens at time tk, and (zk, k) is the observed density at position zk, scanned at time k. The higher the PI, the better the performance is. However, due to the limitation on the CT table velocity and acceleration, PI is less than 1.
Fig. 2 to Fig. 5 show the patient vascular diseases and corresponding simulated scan results (CT number of the vasculature) of the adaptive-chasing and constant-speed method results. In most of the time, the scanned CT number of adaptive-chasing method (dash-dot curve) is greater than that of the constant-speed method (dashed curve).
Table 1 shows the estimated effective dose and PI for adaptive-chasing and constant-speed method, respectively. We can see that adaptive-chasing method reduces the effective dose, on average, 39% of the constant-speed method, and PI increases about 17%.
The adaptive bolus chasing CTA with dose reduction is a new concept in dose reduction. It saves the radiation dose in two ways: 1) generally lowering the tube current, which is made possible through adaptive-chasing method. Though the adaptive-chasing method does not guarantee the scanned HU is greater than the expected, it generally has greater performance than the constant-speed method, which is validated in our simulations. 2) Tracking the bolus peak trajectory. The adaptive-chasing method allows the CT table move as fast as it can and still have a good resultant scan results for all segments of the vasculature. As we know, the faster scan the lower radiation dose if all other parameters are kept the same.
Although the simulation is performed on a four-row detector CT scanner, this dose reduction approach could be easily extended to the 16- or 64-row detector CT scanners. Generally, 16 or 64 slice CT scanner has a wider field of view (in z direction) and moves faster. It provides much more information in a shorter time. On one hand, it is good to the control algorithm because more bolus information will make the chasing results better; on the other hand, it requires more computation capacity and challenges the real time control. This problem could be solved in following ways: 1) fast reconstruction algorithm with the advanced CPU techniques, 2) part reconstruction, which only reconstruct the images that around the region of interest. Those images are only used for real time control and the high quality CT images will be reconstructed after the scan.
The constant-speed method is not so bad for this short injection. This is because we choose a relatively good pre-set threshold to start the CT table. It is hard to set the right threshold, but a lot of contrast dose would be used to compensate for the deficit of the constant-speed method. On the other hand, adaptive-chasing method is not sensitive to the pre-set threshold. It can adjust optimal starting time by itself.
Patient 3 has an aneurysm in the abdominal aorta. When the contrast bolus flows there, it stays inside for a while until it fills the aneurysm. It is like the prolonged maximum geometry. Hence, the constant-speed method has almost the same PI as adaptive-chasing method does. However, the adaptive-chasing method saves the radiation dose by lowering the tube current while the constant-speed method does not.
We did not apply uniform slice thickness method in the simulation because it would not change the table increment, product of the tube current and collimator width, and dose savage. The only difference is the varying collimator width. We will try this method for the experiment on the CT scanner and evaluate the resultant CT image quality.
In a summary, we have proposed an adaptive-chasing method with dose reduction. The proposed approach ensures the vasculature to be scanned with higher bolus density and maintains a minimum CNR. The effective radiation dose is reduced through modulating the tube current and using the possible minimum time to track the contrast bolus. Simulation results show that the adaptive-chasing method has greater PI than the constant-speed method (0.95 vs. 0.81) and the effective radiation dose is reduced 39% on average. The proposed techniques also have great potential in optimization of contrast agent, which may save the exam cost and benefit patient who has kidney diseases. To that end, the future work will be focused on two areas: 1) the real experiment and evaluation on the resultant CT images, and 2) optimization of contrast agent.