Cone-beam computed tomography (CBCT) is routinely used for many clinical applications in the fields of neurology, vascular, radiology, oncology, and cardiovascular interventions [
For contrast enhanced CBCT as performed during TACE procedures the long acquisition time also results in a long injection time as the contrast bolus has to be maintained during the whole time of the acquisition to provide a consistent filling of the imaged vessels and tumors. This means that the longer the acquisition time is the longer the contrast injection has to last and the more the contrast medium is required which can become a serious problem for patients with poor kidney function. The long acquisition time of image data is due to the limited maximum detector readout speed (30–60 f/s), limited maximum mechanical speed of CBCT system (40–60°/s), and number of projection images required for certain image quality. The present study intended to evaluate ultrafast CBCT (UF-CBCT) imaging protocols on a multiaxis robotic CBCT system regarding the capability to create image datasets suitable for guiding TACE procedures while at the same time save contrast media and radiation dose compared to values in the literature.
This is a prospective study conducted from October 2011 to June 2013 using 80 patients and the study protocol was accepted by institutional review board. The patient selection was randomly performed for each protocol without bias. The selection of the patients for UF-CBCT imaging was completely dependent on TACE inclusion and exclusion criteria, based on previous publication [
We used a multiaxis robotic CBCT system (Artis zeego, Siemens Healthcare, Forchheim Germany) to conduct patient examinations during TACE therapy. This UF-CBCT system offers the possibility to rotate the tube-detector system around the patient with a maximum speed of up to 100°/sec. The system is equipped with a latest generation 30 × 40 cm flat panel detector made of amorphous silicon with cesium iodide scintillator (aSi:CsI). During UF-CBCT acquisition the system acquires projection images on a 200° circular trajectory with a constant angular frame increment (AI); that is, each AI degree the system acquired an image. This means that the image acquisition frame rate differs between the acceleration phase, the phase with constant speed, and the deceleration phase of the CBCT.
The correlation between UF-CBCT rotation speed (
With a given angular increment for a UF-CBCT acquisition the maximum readout speed of the detector may limit the maximum CBCT rotation speed.
In this study we evaluated two UF-CBCT imaging protocols, each consisting of a CBCT acquisition protocol and a contrast injection protocol (Table
Described specifications of the clinical imaging parameters and contrast material injection protocol used for the ultrafast cone-beam CT imaging of patients during transarterial chemoembolization treatment.
Imaging parameters | Acquisition protocol 1 | Acquisition protocol 2 |
---|---|---|
Set kilo-voltage | 90 kV | 90 kV |
Number of images | 133 | 166 |
Angular increment (°/ima) | 1.5 | 1.2 |
Maximum CBCT speed (°/s) | 100 | 88.8 |
Maximum readout speed (ima/s) | 67 | 74 |
Total acquisition time (s) | 2.54 | 2.72 |
Detector entrance dose (uGy/frame) | 0.36 | 0.36 |
|
||
Contrast injection parameters | Injection protocol 1 | Injection protocol 2 |
|
||
Contrast volume (mL) | 6 | 7 |
Iodine/mL | 320 | 320 |
Saline (mL) | 12 | 20 |
Flow rate (mL/s) | 3 | 3 |
X-ray delay (s) | 3 | 6 |
Injection duration (s) | 6 | 9 |
For the UF-CBCT protocols the CBCT system had to be positioned in a head side position so that the acquisition is performed in a “propeller-like” mode where only one axis of the multiaxis robotic UF-CBCT system is moved. The 200° acquisition trajectory reaches from right-anterior-oblique (RAO) 170° to left-anterior-oblique (LAO) 30°. During TACE, the patients were positioned head-first in supine position on table top and their arms were positioned above the head during data acquisition. Furthermore, UF-CBCT image data was acquired for all patients on expiration condition. The UF-CBCT acquisition was performed with a 0.36
From the projection images a 3D dataset with isotropic voxels of 0.5 mm is automatically reconstructed on the connected workstation (syngo XWP, Siemens Healthcare, Forchheim, Germany) using a filtered back-projection (Feldkamp) algorithm. This 3D dataset is then loaded into the software application syngo InSpace, which allows the user to visualize the dataset in different rendering modes like multiplanar reformatted (MPR), maximum intensity projection (MIP), or volume rendered (VRT) images.
A UF-CBCT acquisition with contrast material injection (Visipaque 320 from GE Healthcare Braunschweig, Germany) was performed for all examined patients. The injection was performed into the right or left hepatic artery using a coaxial microcatheter (2.7F/2.4F × 150 cm; Trevo Pro 18 microcatheter, Concentric medical, CA, USA). A road mapping of the target tumor is possible using contrast material injection as described by Wallace et al. [
After the examination of patients using UF-CBCT, the images were evaluated by three radiologists with 4, 6, and 20 years of experience in abdominal imaging. They used a scoring system to analyze image data qualitatively based on tumor delineation, vascular contrast material perfusion, and appearance of artifacts on the cross-sectional images (Table
Classification of grading score based on tumor delineation, vascular contrast material perfusion, and appearance of the artifacts.
Grading point score | Tumor delineation, vascular contrast material perfusion, and appearance of artifacts. |
---|---|
Description | |
1 | Not suitable for diagnosis (nondiagnostic image data) |
2 | Suboptimal contrast perfusion and tumor delineation, strong appearance of artifacts |
3 | Less than standard contrast perfusion and tumor delineation with hazy appearance of artifacts |
4 | Standard contrast perfusion and tumor delineation with hazy appearance of artifacts |
5 | Vascular contrast material perfusion and tumor delineation are higher than necessary with little or no artifacts |
Details of patient tumor characteristics determined using gold standard MR-image data. The determined tumor characteristics using MRI displayed separately for UF-CBCT imaging protocols 1 (P1) and 2 (P2) patient groups.
Type of hepatic tumor | Tumor involvement | Hepatic tumor groups | Number of patients | Number of tumors | Mean dimension in cm ( | |||
---|---|---|---|---|---|---|---|---|
(P1/P2) | P1 | P2 | P1 | P2 | P1 | P2 | ||
Hypoenhanced tumor | Right lobe: 21/21 | Metastasis from: | ||||||
Left lobe: 13/14 | Thyroid carcinoma | 5 | 6 | 16 | 20 | 4.4 × 3.9 | 4 × 3.8 | |
Caudate lobe: 2/5 | Colorectal carcinoma | 6 | 7 | 25 | 23 | 4.8 × 4.7 | 4.5 × 4.4 | |
Quadrate lobe: 5/3 | ||||||||
|
||||||||
Heterogeneously enhanced tumor | Right lobe: 21/24 | Cholangiocarcinoma | 5 | 4 | 17 | 15 | 4 × 3.8 | 3.8 × 3.6 |
Left lobe: 15/12 | Metastasis from: | |||||||
Caudate lobe: 7/3 | Colorectal carcinoma | 4 | 7 | 12 | 12 | 4.3 × 3.7 | 4.5 × 4 | |
Quadrate lobe: 1/2 | Breast carcinoma | 4 | 3 | 15 | 14 | 4.8 × 4.3 | 4.3 × 3.9 | |
|
||||||||
Homogenously enhanced tumor | Right lobe: 26/22 | Hepatocell. carcinoma | 7 | 5 | 28 | 20 | 4.1 × 3.6 | 4 × 3.6 |
Left lobe: 15/14 | Cholangiocarcinoma | 5 | 5 | 9 | 13 | 4.9 × 4.2 | 4.2 × 3.8 | |
Caudate lobe: 2/4 | Metastasis from: | |||||||
Quadrate lobe: 3/5 | Colorectal carcinoma | 4 | 3 | 9 | 12 | 4.7 × 4.4 | 4.5 × 3.8 |
Two radiation dose parameters, dose area product (DAP) and patient entrance dose (PED), were obtained during exposure from the patient examination protocol generated on the UF-CBCT system [
Results of the present study are displayed as means ± standard deviation and range for continuous variables. The statistical analyses were performed using computer based BiAS software (BiAS for Windows, Epsilon 2008, version 8.4.2). A
Details of the patient hepatic tumor characteristics obtained using gold standard MRI are displayed in Table
Details of the patient tumor characteristics obtained using ultrafast CBCT image data (P1: imaging protocol 1; P2: imaging protocol 2).
Type of hepatic tumor | Tumor involvement | Hepatic tumor groups | Number of patients | Number of tumors | Mean dimension in cm ( | |||
---|---|---|---|---|---|---|---|---|
(P1/P2) | P1 | P2 | P1 | P2 | P1 | P2 | ||
Hypoenhanced tumor | Right lobe: 16/19 | Metastasis from: | ||||||
Left lobe: 10/13 | Thyroid carcinoma | 5 | 6 | 11 | 19 | 3.9 × 3.4 | 3.8 × 3.5 | |
Caudate lobe: 0/5 | Colorectal carcinoma | 6 | 7 | 17 | 21 | 4.5 × 4.2 | 4.3 × 4.1 | |
Quadrate lobe: 2/3 | ||||||||
|
||||||||
Heterogeneously enhanced tumor | Right lobe: 18/21 | Cholangiocarcinoma | 5 | 4 | 13 | 13 | 3.5 × 3.4 | 3.6 × 3.4 |
Left lobe: 11/11 | Metastasis from: | |||||||
Caudate lobe: 4/3 | Colorectal carcinoma | 4 | 7 | 9 | 11 | 3.8 × 3.2 | 4.2 × 3.7 | |
Quadrate lobe: 0/2 | Breast carcinoma | 4 | 3 | 11 | 13 | 4.2 × 3.7 | 4 × 3.7 | |
|
||||||||
Homogenously enhanced tumor | Right lobe: 19/21 | Hepatocell. carcinoma | 7 | 5 | 21 | 19 | 3.6 × 3.1 | 3.7 × 3.3 |
Left lobe: 11/13 | Cholangiocarcinoma | 5 | 5 | 6 | 12 | 4.6 × 3.6 | 3.9 × 3.5 | |
Caudate lobe: 2/4 | Metastasis from: | |||||||
Quadrate lobe: 2/5 | Colorectal carcinoma | 4 | 3 | 7 | 12 | 4 × 3.8 | 4.2 × 3.5 |
Displayed image quality qualitative analysis scores (mean ± standard deviation and range) obtained from the readers using both ultrafast CBCT patient imaging protocols.
Imaging protocol | Tumor classification: | ||
---|---|---|---|
Hypoenhanced tumor (A) | Heterogeneously enhanced tumor (B) | Homogenously enhanced tumor (C) | |
Protocol 1 | 2.9 ± 0.4 (2.6–3.1) | 3 ± 0.4 (2.6–3.25) | 3.1 ± 0.5 (2.7–3.4) |
Protocol 2 | 4 ± 0.6 (3.7–4.2) | 4.3 ± 0.4 (4–4.5) | 4.6 ± 0.3 (4.3–4.8) |
|
|||
|
0.0001 | 0.0001 | 0.0001 |
Frames (a) and (b) represented the images obtained from a 62-year-old patient during TACE therapy, generated using ultrafast cone-beam CT with imaging protocol 1. Pretreatment magnetic resonance cross-sectional images ((c) and (f)) show a clear view of embedded tumor in the hepatic parenchyma. Hepatic tumor detection was insufficient using imaging protocol 1 data (tumor indicated using black arrow) compared to imaging protocol 2 due to a reduction of contrast material in the tumors. Coronal reconstructed images ((d) and (e)) were acquired using imaging protocol 2 during a 60-year-old patient TACE examination. The images show excellent tumor(s), feeding vessels, and hepatic parenchymal visualization; furthermore, notice the strong contrast material enhancement of the tumors with little or no artifacts.
Quantitative image quality analysis also showed similar results to qualitative analysis (Table
Quantitative image quality parameters and tumor delineation obtained from ultrafast CBCT image data with contrast material injection for both examined protocols. Furthermore, TLC represents tumor-to-liver contrast.
Image quality parameter | Measurement locations | |||
---|---|---|---|---|
Normal hepatic parenchyma | Hypoenhanced tumor (A) | Heterogeneously enhanced tumor (B) | Homogenously enhanced tumor (C) | |
Protocol 1 | ||||
Hounsfield unit (HU) | 54 ± 13 (41–68) | 25 ± 8 (14–37) | 63 ± 15 (47–77) | 151 ± 17 (119–196) |
Image noise (HU) | 42 ± 11 (37–49) | 52 ± 18 (39–59) | 62 ± 17 (51–73) | 83 ± 22 (69–99) |
Signal-to-noise ratio (SNR) | 1.3 (1.15–1.4) | 0.48 (0.35–0.58) | 1 (0.98–1.1) | 1.8 (1.7–1.93) |
Tumor-to-liver contrast (TLC) | — | −29 (−18–−39) | 9 (4–14) | 97 (91–113) |
|
||||
Protocol 2 | ||||
Hounsfield unit (HU) | 83 ± 14 (69–95) | 10 ± 3 (6–15) | 115 ± 20 (74–147) | 217 ± 21 (169–244) |
Image noise (HU) | 29 ± 9 (25–35) | 17 ± 7 (13–23) | 44 ± 15 (32–52) | 60 ± 19 (49–66) |
Signal-to-noise ratio (SNR) | 2.85 (2.75–3) | 0.7 (0.4–0.8) | 2.6 (2.3–2.8) | 3.6 (3.5–3.7) |
Tumor-to-liver contrast (TLC) | — | −73 (−59–−89) | 32 (21–41) | 134 (119–162) |
|
||||
|
0.0001 | 0.0001 | 0.0001 | 0.0001 |
The volume of iodine injected into the patients during UF-CBCT was 1920 mg and 2240 mg, respectively, for injection protocols 1 and 2 during contrast material injection. In [
Displayed sensitivity and predictive values, calculated using UF-CBCT imaging protocols 1 and 2 and magnetic resonance image data.
Type of tumor | Interpreted as tumor on |
|
|
|
|
Sensitivity (%) | Specificity (%) | Positive predictive value (%) | Negative predictive value (%) | |
---|---|---|---|---|---|---|---|---|---|---|
MRI | UF-CBCT protocol 1 | |||||||||
Hypoenhanced tumors | 41 | 28 | 23 | 18 | 5 | 0 | 56 | 0 | 82 | 0 |
Heterogeneously enhanced tumors | 44 | 33 | 27 | 17 | 6 | 0 | 61.3 | 0 | 81.8 | 0 |
Homogeneously enhanced tumors | 46 | 34 | 29 | 17 | 5 | 0 | 63 | 0 | 85.3 | 0 |
|
||||||||||
MRI | UF-CBCT protocol 2 | |||||||||
|
||||||||||
Hypoenhanced tumors | 43 | 40 | 36 | 7 | 4 | 0 | 83.7 | 0 | 90 | 0 |
Heterogeneously enhanced tumors | 41 | 37 | 34 | 7 | 3 | 0 | 82.9 | 0 | 91.9 | 0 |
Homogeneously enhanced tumors | 45 | 43 | 41 | 4 | 2 | 0 | 91 | 0 | 95.3 | 0 |
Mean PED estimated for UF-CBCT imaging protocols 1 and 2 were
Long patient breath hold is necessary for conventional CBCT image data acquisition to avoid motion artifacts and thus obtain reasonable image quality. A continuous contrast material injection should be maintained in entire duration during contrast enhanced CBCT data acquisition. These two conditions directly affect image quality as well as contrast material load in the patients. The high rotation speed of the tube-detector system and shorter acquisition time of the UF-CBCT make it easier for patients to comply with the breath-hold requirements and thus reduce the occurrence of motion artifacts, which produce good image quality. Ultrafast CBCT imaging bears the potential to reduce volume of contrast material (up to 6 and 7 mL) required for contrast enhanced CBCT during TACE therapy of patients due to a reduction of imaging time and appropriate contrast protocol used for imaging. The volume of iodine in the contrast material injected during CBCT imaging as regards TACE is significantly different in various published materials [
We obtained a reduced radiation dose on patients due to ultrafast imaging time compared to published data [
Injection protocols and delay time are highly influencing parameters on the enhancement of hepatic parenchyma and tumors. Imaging protocol 1 produced an insufficient visualization of hepatic tumors/metastasis. This is due to the short X-ray delay time of 3 seconds, which prevented proper tumor enhancement before the UF-CBCT data acquisition. In imaging protocol 2, the X-ray delay was extended to 6 seconds which allowed the further perfusion of the tumor feeding vessels with contrast material. To keep the contrast load for the patient at a reasonable level the contrast dilution was increased from a mixing level of 1 : 2 (contrast/saline) in imaging protocol 1 to 1 : 3 in imaging protocol 2. Despite the higher contrast/water dilution compared to imaging protocol 1, imaging protocol 2 produced excellent tumor enhancement with the possibility of good tumor delineation during UF-CBCT imaging. The reasons are the higher delay time and the higher number of frames acquired. To obtain reasonable image quality for CBCT image data a certain number of projection images (frames) are required [
In the present study we used an ultrafast robotic CBCT imaging system and protocols with less than 3 seconds image acquisition time acquiring image data during TACE. Imaging protocol 2 provided an excellent visualization of tumor(s) and feeding vasculature as well as hepatic parenchyma due to adequate bolus timing, mixing ratio X-ray delay time, and higher number of acquired frames compared to imaging protocol 1. Ultrafast image acquisitions reduce contrast material injection volume to patients during TACE examination using UF-CBCT. Furthermore, UF-CBCT imaging achieved a reduction of radiation dose due to reduction of total number of acquired frames during imaging. The reduction of imaging time helps to prevent the appearance of motion artifacts which was previously reported as a problem with longer CBCT acquisitions. Based on image quality results of imaging protocol 2 we recommend that this imaging protocol should be used for UF-CBCT image acquisitions during patient imaging.
The authors declare that there is no conflict of interests regarding the publication of this paper.