Low-dose 2.5 MV cone-beam computed tomography with thick CsI flat-panel imager.

Most of the treatment units, both new and old models, are equipped with a megavoltage portal imager but its use for volumetric imaging is limited. This is mainly due to the poor image quality produced by the high-energy treatment beam (> 6MV). A linac at our center is equipped with a prototype 2.5 MV imaging beam. This study evaluates the feasibility of low-dose megavoltage cone-beam imaging with the 2.5MV beam and a thick cesium iodide detector, which is a high-efficiency imager. Basic imaging properties such as spatial resolution and modulation transfer function were assessed for the 2.5 MV prototype imaging system. For image quality and imaging dose, a series of megavoltage cone-beam scans were acquired for the head, thorax, and pelvis of an anthropomorphic phantom and were compared to kilovoltage cone-beam and 6X megavoltage cone-beam images. To demonstrate the advantage of MV imaging, a phantom with metallic inserts was scanned and the image quality was compared to CT and kilovoltage cone-beam scans. With a lower energy beam and higher detector efficiency, the 2.5 MV imaging system generally yields better image quality than does the 6 MV imaging system with the conventional MV imager. In particular, with the anthropomorphic phantom studies, the contrast to noise of bone to tissue is generally improved in the 2.5 MV images compared to 6 MV. With an image quality sufficient for bony alignment, the imaging dose for 2.5 MV cone-beam images is 2.4-3.4 MU compared to 26 MU in 6 MV cone-beam scans for the head, thorax, and pelvis regions of the phantom. Unlike kilovoltage cone-beam, the 2.5 MV imaging system does not suffer from high-Z image artifacts. This can be very useful for treatment planning in cases where high-Z prostheses are present.

Observation of interfractional variations in lung tumor position using respiratory gated and ungated megavoltage cone-beam computed tomography.

PURPOSE: To evaluate the use of megavoltage cone-beam computed tomography (MV CBCT) to measure interfractional variation in lung tumor position. METHODS AND MATERIALS: Eight non-small-cell lung cancer patients participated in the study, 4 with respiratory gating and 4 without. All patients underwent MV CBCT scanning at weekly intervals. Contoured planning CT and MV CBCT images were spatially registered based on vertebral anatomy, and displacements of the tumor centroid determined. Setup error was assessed by comparing weekly portal orthogonal radiographs with digitally reconstructed radiographs generated from planning CT images. Hypothesis testing was performed to test the statistical significance of the volume difference, centroid displacement, and setup uncertainty. RESULTS: The vertebral bodies and soft tissue portions of tumor within lung were visible on the MV CBCT scans. Statistically significant systematic volume decrease over the course of treatment was observed for 1 patient. The average centroid displacement between simulation CT and MV CBCT scans were 2.5 mm, -2.0 mm, and -1.5 mm with standard deviations of 2.7 mm, 2.7 mm, and 2.6 mm in the right-left, anterior-posterior and superior-inferior directions. The mean setup errors were smaller than the centroid shifts, while the standard deviations were comparable. In most cases, the gross tumor volume (GTV) defined on the MV CBCT was located on average at least 5 mm inside a 10 mm expansion of the GTV defined on the planning CT scan. CONCLUSIONS: The MV CBCT technique can be used to image lung tumors and may prove valuable for image-guided radiotherapy. Our conclusions must be verified in view of the small patient number.

Low-dose megavoltage cone-beam computed tomography for lung tumors using a high-efficiency image receptor.

We report on the capabilities of a low-dose megavoltage cone-beam computed tomography (MV CBCT) system. The high-efficiency image receptor consists of a photodiode array coupled to a scintillator composed of individual CsI crystals. The CBCT system uses the 6 MV beam from a linear accelerator. A synchronization circuit allows us to limit the exposure to one beam pulse [0.028 monitor units (MU)] per projection image. 150-500 images (4.2-13.9 MU total) are collected during a one-minute scan and reconstructed using a filtered backprojection algorithm. Anthropomorphic and contrast phantoms are imaged and the contrast-to-noise ratio of the reconstruction is studied as a function of the number of projections and the error in the projection angles. The detector dose response is linear (R2 value 0.9989). A 2% electron density difference is discernible using 460 projection images and a total exposure of 13 MU (corresponding to a maximum absorbed dose of about 12 cGy in a patient). We present first patient images acquired with this system. Tumors in lung are clearly visible and skeletal anatomy is observed in sufficient detail to allow reproducible registration with the planning kV CT images. The MV CBCT system is shown to be capable of obtaining good quality three-dimensional reconstructions at relatively low dose and to be clinically usable for improving the accuracy of radiotherapy patient positioning.

Integrating respiratory gating into a megavoltage cone-beam CT system.

We have previously described a low-dose megavoltage cone beam computed tomography (MV CBCT) system capable of producing projection image using one beam pulse. In this study, we report on its integration with respiratory gating for gated radiotherapy. The respiratory gating system tracks a reflective marker on the patient's abdomen midway between the xiphoid and umbilicus, and disables radiation delivery when the marker position is outside predefined thresholds. We investigate two strategies for acquiring gated scans. In the continuous rotation-gated acquisition, the linear accelerator (LINAC) is set to the fixed x-ray mode and the gantry makes a 5 min, 360 degree continuous rotation, during which the gating system turns the radiation beam on and off, resulting in projection images with an uneven distribution of projection angles (e.g., in 70 arcs each covering 2 degrees). In the gated rotation-continuous acquisition, the LINAC is set to the dynamic arc mode, which suspends the gantry rotation when the gating system inhibits the beam, leading to a slightly longer (6-7 min) scan time, but yielding projection images with more evenly distributed projection angles (e.g., approximately 0.8 degrees between two consecutive projection angles). We have tested both data acquisition schemes on stationary (a contrast detail and a thoracic) phantoms and protocol lung patients. For stationary phantoms, a separate motion phantom not visible in the images is used to trigger the RPM system. Frame rate is adjusted so that approximately 450 images (13 MU) are acquired for each scan and three-dimensional tomographic images reconstructed using a Feldkamp filtered backprojection algorithm. The gated rotation-continuous acquisition yield reconstructions free of breathing artifacts. The tumor in parenchymal lung and normal tissues are easily discernible and the boundary between the diaphragm and the lung sharply defined. Contrast-to-noise ratio (CNR) is not degraded relative to nongated scans of stationary phantoms. The continuous rotation-gated acquisition scan also yields tomographic images with discernible anatomic features; however, streak artifacts are observed and CNR is reduced by approximately a factor of 4. In conclusion, we have successfully developed a gated MV CBCT system to verify the patient positioning for gated radiotherapy.

Independent verification of ultrasound based image-guided radiation treatment, using electronic portal imaging and implanted gold markers.

The aim of this paper is to study the correction of prostate motion and position during external beam therapy. The correction was performed using a commercially available ultrasound-based repositioning tool. Electronic portal imaging with the use of fiducial markers was used to assess efficacy and accuracy. Patients undergoing radiation treatment for adenocarcinoma of the prostate were enrolled in a positioning study. Fifteen patients had five to six gold fiducial markers implanted in their prostate. These patients were positioned daily in a standard manner and then were repositioned every other day using an ultrasound-based correction system. Every fraction of a patients' treatment was imaged. This yielded 156 image pairs with and 119 pairs without repositioning available for analysis. This group of patients with markers had the following residual positions measured after the use of ultrasound repositioning. A mean error of -0.4 mm (LL), -2.6 mm (CC), and +2.5 mm (AP) with a standard deviation of 4.3, 5.4, and 5.7 mm. In two directions the improvements of treatment using the ultrasound correction were smaller than the precision of this experiment. They were no larger than 0.81 mm (LAT), and 0.95 mm (CC). In the AP direction a significant improvement was found of 1.6 mm. A highly significant correlation (p < 0.001) was found between the residual errors in the cranio-caudal direction and the shifts performed on the basis of the ultrasound measurements (Spearman ranking R = 0.53). We presented a method to objectively estimate improvements by a correction scheme. This method applied to ultrasound-based adjustment showed significant improvement in one direction and no measurable improvement in two other directions.