Enhancement of image quality with a fast iterative scatter and beam hardening correction method for kV CBCT.

The problem of the enormous amount of scattered radiation in kV CBCT (kilo voltage cone beam computer tomography) is addressed. Scatter causes undesirable streak- and cup-artifacts and results in a quantitative inaccuracy of reconstructed CT numbers, so that an accurate dose calculation might be impossible. Image contrast is also significantly reduced. Therefore we checked whether an appropriate implementation of the fast iterative scatter correction algorithm we have developed for MV (mega voltage) CBCT reduces the scatter contribution in a kV CBCT as well. This scatter correction method is based on a superposition of pre-calculated Monte Carlo generated pencil beam scatter kernels. The algorithm requires only a system calibration by measuring homogeneous slab phantoms with known water-equivalent thicknesses. In this study we compare scatter corrected CBCT images of several phantoms to the fan beam CT images acquired with a reduced cone angle (a slice-thickness of 14 mm in the isocenter) at the same system. Additional measurements at a different CBCT system were made (different energy spectrum and phantom-to-detector distance) and a first order approach of a fast beam hardening correction will be introduced. The observed image quality of the scatter corrected CBCT images is comparable concerning resolution, noise and contrast-to-noise ratio to the images acquired in fan beam geometry. Compared to the CBCT without any corrections the contrast of the contrast-and-resolution phantom with scatter correction and additional beam hardening correction is improved by a factor of about 1.5. The reconstructed attenuation coefficients and the CT numbers of the scatter corrected CBCT images are close to the values of the images acquired in fan beam geometry for the most pronounced tissue types. Only for extreme dense tissue types like cortical bone we see a difference in CT numbers of 5.2%, which can be improved to 4.4% with the additional beam hardening correction. Cupping is reduced from 20% to 4% with scatter correction and 3% with an additional beam hardening correction. After 3 iterations (small phantoms) and 6 to 7 iterations (large phantoms) the algorithm converges. Therefore the algorithm is very fast, that means 1.3 seconds per projection for 3 iterations on a standard PC.

Correction of patient positioning errors based on in-line cone beam CTs: clinical implementation and first experiences.

BACKGROUND: The purpose of the study was the clinical implementation of a kV cone beam CT (CBCT) for setup correction in radiotherapy. PATIENTS AND METHODS: For evaluation of the setup correction workflow, six tumor patients (lung cancer, sacral chordoma, head-and-neck and paraspinal tumor, and two prostate cancer patients) were selected. All patients were treated with fractionated stereotactic radiotherapy, five of them with intensity modulated radiotherapy (IMRT). For patient fixation, a scotch cast body frame or a vacuum pillow, each in combination with a scotch cast head mask, were used. The imaging equipment, consisting of an x-ray tube and a flat panel imager (FPI), was attached to a Siemens linear accelerator according to the in-line approach, i.e. with the imaging beam mounted opposite to the treatment beam sharing the same isocenter. For dose delivery, the treatment beam has to traverse the FPI which is mounted in the accessory tray below the multi-leaf collimator. For each patient, a predefined number of imaging projections over a range of at least 200 degrees were acquired. The fast reconstruction of the 3D-CBCT dataset was done with an implementation of the Feldkamp-David-Kress (FDK) algorithm. For the registration of the treatment planning CT with the acquired CBCT, an automatic mutual information matcher and manual matching was used. RESULTS AND DISCUSSION: Bony landmarks were easily detected and the table shifts for correction of setup deviations could be automatically calculated in all cases. The image quality was sufficient for a visual comparison of the desired target point with the isocenter visible on the CBCT. Soft tissue contrast was problematic for the prostate of an obese patient, but good in the lung tumor case. The detected maximum setup deviation was 3 mm for patients fixated with the body frame, and 6 mm for patients positioned in the vacuum pillow. Using an action level of 2 mm translational error, a target point correction was carried out in 4 cases. The additional workload of the described workflow compared to a normal treatment fraction led to an extra time of about 10-12 minutes, which can be further reduced by streamlining the different steps. CONCLUSION: The cone beam CT attached to a LINAC allows the acquisition of a CT scan of the patient in treatment position directly before treatment. Its image quality is sufficient for determining target point correction vectors. With the presented workflow, a target point correction within a clinically reasonable time frame is possible. This increases the treatment precision, and potentially the complex patient fixation techniques will become dispensable.

Linac-integrated 4D cone beam CT: first experimental results.

A new online imaging approach, linac-integrated cone beam CT (CBCT), has been developed over the past few years. It has the advantage that a patient can be examined in their treatment position directly before or during a radiotherapy treatment. Unfortunately, respiratory organ motion, one of the largest intrafractional organ motions, often leads to artefacts in the reconstructed 3D images. One way to take this into account is to register the breathing phase during image acquisition for a phase-correlated image reconstruction. Therefore, the main focus of this work is to present a system which has the potential to investigate the correlation between internal (movement of the diaphragm) and external (data of a respiratory gating system) information about breathing phase and amplitude using an inline CBCT scanner. This also includes a feasibility study about using the acquired information for a respiratory-correlated 4D CBCT reconstruction. First, a moving lung phantom was used to develop and to specify the required methods which are based on an image reconstruction using only projections belonging to a certain moving phase. For that purpose, the corresponding phase has to be detected for each projection. In the case of the phantom, an electrical signal allows one to track the movement in real time. The number of projections available for the image reconstruction depends on the breathing phase and the size of the position range from which projections should be used for the reconstruction. The narrower this range is, the better the inner structures can be located, but also the noise of the images increases due to the limited number of projections. This correlation has also been analysed. In a second step, the methods were clinically applied using data sets of patients with lung tumours. In this case, the breathing phase was detected by an external gating system (AZ-733V, Anzai Medical Co.) based on a pressure sensor attached to the patient's abdominal region with a fixation belt. The comparison of the reconstructed 4D CBCT images and the corresponding 4D CT images used for the treatment planning provides the required information for the calculation of possible setup errors. So, a repositioning of the patient is feasible even though the patient moves due to respiration. In addition to the external signal, the position of the diaphragm in the cranial-caudal direction could be extracted from each projection. Both independent sources of information show a very good agreement of the phase and even the amplitude of the movement and the external signal respectively. This suggests the usability of such a system for a gated dose delivery approach. However, more studies involving patients with different incidences have to be carried out to confirm these first results.

Linac-integrated kV-cone beam CT: technical features and first applications.

One of the most prominent imaging techniques in image-guided radiotherapy (IGRT) is the acquisition of cone beam computed tomographies (CBCTs) at the linac with the patient in treatment position. CBCTs provide accurate 3-dimensional (3D) knowledge about the patient's anatomy for every treatment fraction and are therefore well suited for all adaptive corrections of errors related to interfractional uncertainties of the treatment process. In this paper, we first describe the technical development and implementation of this new imaging technique at our linac, i.e., the hardware components and their operating parameters are discussed in detail for a standard image acquisition of CBCTs. Then, an extension of this approach for the acquisition of complete images for extended field of views--the "shifted detector" technique--is presented followed by a first investigation of how CBCTs can be reliably used for adaptive dose calculations. Finally, a first clinical application, the process of automatic patient positioning based on CBCT images, is discussed. From our investigations, we conclude that the technical development of linac-integrated CBCTs bears an enormous potential for the correction of interfractional treatment errors. However, image quality and reconstruction speed of the images leave room for improvement. The development of clinical strategies for the optimal application of this new image modality in a clinical environment is one the major tasks for the future.

Compensation for respiratory motion by gated radiotherapy: an experimental study.

Respiratory organ motion is known to be one of the largest intrafractional organ motions. Therefore, it is important to investigate the potential benefit of gated dose delivery approaches which aim to account for the respective dose uncertainties. In this study respiration is simulated by a moving lung phantom; the movement is not restricted to a normal sinusoidal progression and simulates the one of the embedded lung tumour in the cranial-caudal direction. An IMRT plan with a total of 29 beam segments was designed for the treatment of this tumour. It was irradiated in its resting position-which is the position at exhalation-and during movement. Furthermore the irradiation was triggered using different amplitude thresholds, which means that the irradiation only proceeded if the deviation of the tumour's position from its resting position is smaller than the given threshold. We determined the gating-related increase of the treatment time for various gating procedures. We also measured the resulting dose distribution in specific slices of the phantom perpendicular to the direction of the movement using film dosimetry and compared it to the dose distribution of the static case. Since these film measurements cannot be done inside the whole tumour, additionally the movement and gating was simulated using the planning software to calculate the 3D dose distribution inside the tumour and to generate dose volume histograms for different treatment modalities. The total treatment time was observed to increase by 20%-100% depending on the individual gating threshold and can be calculated easily. The analysis of the films showed that irradiation without gating leads to significant underdosages up to 33%, especially at the edge of the tumour. With gating it is possible to considerably reduce this underdosage down to 9% depending on the trigger threshold. The calculation of the dose volume histograms makes it possible to find a reasonable compromise between the improvement of the dose distribution and the increase of the treatment time.

Intensity modulated radiation therapy with multileaf collimators of different leaf widths: a comparison of achievable dose distributions.

PURPOSE: A planning study to analyze the impact of different leaf widths on the achievable dose distributions with intensity modulated radiation therapy (IMRT). METHODS: Five patients (3 intra- and 2 extra-cranial) with projected planning target volume (PTV) sizes smaller than 10 cm by 10 cm were re-planned with four different multileaf collimators (MLC). Two internal collimators with an isocentric leaf width of 4 and 10 mm and two add-on collimators with an isocentric leaf width of 2.75 and were evaluated. The inverse treatment planning system KonRad (Siemens Medical Solutions) was used to create IMRT 'step & shoot' plans. For each patient the same arrangement of beams and the same parameters for the optimization were used for all MLCs. The beamlet size for all treatment plans was chosen to coincide with the leaf width of the respective MLC. To evaluate the treatment plans 3D dose distributions and dose volume histograms were analyzed. As indicators for the quality of the PTV dose distribution the minimum dose, maximum dose and the standard deviation were used. For the organs at risk (OAR) the equivalent uniform dose (EUD) was calculated. To measure the dose coverage of the PTV the volume (V(90)) that received doses higher than 90% of the prescribed dose was calculated where for the conformity the dose conformity index given by Baltas et al. was determined. RESULTS: The MLC with the smallest leaf width yields the best mean value of all five patients for the PTV coverage and for the conformity. For the MLCs with the same leaf width, the add-on MLC leads to superior treatment plans than the internal MLC. This is due to the sharper penumbra of the add-on MLC. The number of IMRT field segments to deliver increased by approximately a factor of two if 2. MLC leafs are used instead of the standard 10 mm leafs. In case of the para-spinal patients the EUD value for the spinal cord is only reduced slightly by using MLCs with leaf widths smaller than 5 mm. For the intra-cranial the EUD value for some organs improved with reduced leaf widths while for some organs the 10 mm MLC leafs give comparable values. CONCLUSION: As expected the MLC with the smallest leaf width always yields the best PTV coverage. Reducing the leaf width from 4 to 2.75 mm results in a slight enhancement of the PTV coverage. With the selected organ parameters no significant improvement for most OAR was found. The disadvantage of the reduction of the leaf width is the increasing number of segments due to the more complex fluence patterns and therefore an increased delivery time.

IMRT-application with an add-on MMLC.

In order to provide automatic IMRT dose delivery with an add-on MMLC a technical integration of a MMLC system with a linear accelerator was realized. The principle of this integration and the changes and enhancements of the existing hard- and software are briefly described. The system was tested by the automatic delivery of an IMRT plan designed for a head and neck phantom. A verification of dose delivery was performed with film dosimetry. The plan consisting of 78 "step and shoot" segments could be delivered within 17 minutes. A high spatial accuracy of the fluence pattern at the isocentre was reached by a resolution of 2.75x2.75 mm(2). The measured dose profiles were within 3% of the maximum dose of the calculated profiles.