Development and validation of a 3D-printed bolus cap for total scalp irradiation.

PURPOSE: The goal of total scalp irradiation (TSI) is to deliver a uniform dose to the scalp, which requires the use of a bolus cap. Most current methods for fabricating bolus caps are laborious, yet still result in nonconformity and low reproducibility, which can lead to nonuniform irradiation of the scalp. We developed and validated patient-specific bolus caps for TSI using three-dimensional (3D) printing. METHODS AND MATERIALS: 3D-printing materials were radiologically analyzed to identify a material with properties suitable for use as a bolus cap. A Python script was developed within a commercial treatment planning system to automate the creation of a ready-to-print, patient-specific 3D bolus cap model. A bolus cap was printed for an anthropomorphic head phantom using a commercial vendor and a computed tomography simulation of the anthropomorphic head phantom and bolus cap was used to create a volumetric-modulated arc therapy TSI treatment plan. The planned treatment was delivered to the head phantom and dosimetric validation was performed using thermoluminescent dosimeters (TLD). The developed procedure was used to create a bolus cap for a clinical TSI patient, and in vivo TLD measurements were acquired for several fractions. RESULTS: Agilus-60 was validated as a new 3D-printing material suitable for use as bolus. A 3D-printed Agilus-60 bolus cap had excellent conformality to the phantom scalp, with a maximum air gap of 4 mm. TLD measurements showed that the bolus cap generated a uniform dose to the scalp within a 2.7% standard deviation, and the delivered doses agreed with calculated doses to within 2.4% on average. The patient bolus was conformal and the average difference between TLD measured and planned doses was 5.3%. CONCLUSIONS: We have developed a workflow to 3D-print highly conformal bolus caps for TSI and demonstrated these caps can reproducibly generate a uniform dose to the scalp.

Comprehensive Investigation on Controlling for CT Imaging Variabilities in Radiomics Studies.

Radiomics has shown promise in improving models for predicting patient outcomes. However, to maximize the information gain of the radiomics features, especially in larger patient cohorts, the variability in radiomics features owing to differences between scanners and scanning protocols must be accounted for. To this aim, the imaging variability of radiomics feature values was evaluated on 100 computed tomography scanners at 35 clinics by imaging a radiomics phantom using a controlled protocol and the commonly used chest and head protocols of the local clinic. We used a linear mixed-effects model to determine the degree to which the manufacturer and individual scanners contribute to the overall variability. Using a controlled protocol reduced the overall variability by 57% and 52% compared to the local chest and head protocols respectively. The controlled protocol also reduced the relative contribution of the manufacturer to the total variability. For almost all variabilities (manufacturer, scanner, and residual with different preprocesssing), the controlled protocol scans had a significantly smaller variability than the local protocol scans did. For most radiomics features, the imaging variability was small relative to the inter-patient feature variability in non-small cell lung cancer and head and neck squamous cell carcinoma patient cohorts. From this study, we conclude that using controlled scans can reduce the variability in radiomics features, and our results demonstrate the importance of using controlled protocols in prospective radiomics studies.

Effect of tube current on computed tomography radiomic features.

Variability in the x-ray tube current used in computed tomography may affect quantitative features extracted from the images. To investigate these effects, we scanned the Credence Cartridge Radiomics phantom 12 times, varying the tube current from 25 to 300 mA∙s while keeping the other acquisition parameters constant. For each of the scans, we extracted 48 radiomic features from the categories of intensity histogram (n = 10), gray-level run length matrix (n = 11), gray-level co-occurrence matrix (n = 22), and neighborhood gray tone difference matrix (n = 5). To gauge the size of the tube current effects, we scaled the features by the coefficient of variation of the corresponding features extracted from images of non-small cell lung cancer tumors. Variations in the tube current had more effect on features extracted from homogeneous materials (acrylic, sycamore wood) than from materials with more tissue-like textures (cork, rubber particles). Thirty-eight of the 48 features extracted from acrylic were affected by current reductions compared with only 2 of the 48 features extracted from rubber particles. These results indicate that variable x-ray tube current is unlikely to have a large effect on radiomic features extracted from computed tomography images of textured objects such as tumors.

Deep-Inspiration Breath-Hold Intensity Modulated Radiation Therapy to the Mediastinum for Lymphoma Patients: Setup Uncertainties and Margins.

PURPOSE: Patient setup for treating large target volumes can be challenging. In the present study, we measured the local uncertainties in the treatment of mediastinal lymphoma and investigated the need for region-specific planning target volume (PTV) margins. METHODS AND MATERIALS: The data from 30 patients who had undergone radiation therapy for mediastinal lymphoma were retrospectively analyzed. A computed tomography (CT)-on-rails (CTOR) system in the treatment room was used for daily image guidance. The total PTV was split into 6 regions: neck, supraclavicular fossa, axilla, mediastinum, upper heart, and lower heart. The total PTV and the 6 local regions were separately aligned to the planning CT scan using automatic rigid registration. The residual local errors using 3 setup strategies were investigated: no image guidance, CTOR setup to total PTV, and simulated cone beam CT setup to the mediastinum. Errors were recorded in the anteroposterior, superoinferior, and right-left directions separately. Using the residual error calculations, the margins required to cover 95% of the clinical target volume for 90% of the patients was estimated. RESULTS: For each patient, 12 to 21 days of daily CTOR data were available for analysis. The residual errors for the total PTV and mediastinum setups were both smaller than those with no image guidance. The lower heart region had more uncertainty with all 3 setup strategies. Margin analysis revealed that the magnitude of the margin is dependent on the imaging strategy, direction, and local region inside the PTV. Margins >7 mm are necessary to account for uncertainty in the neck, lower heart, and axilla regions even under daily CT guidance. CONCLUSIONS: Setup uncertainties in the mediastinum are not uniform and are dependent on target location and imaging strategy. However, with the appropriate margin, we can target regions that might not be visualized with the available on-board imager system.

Average CT in PET studies of colorectal cancer patients with metastasis in the liver and esophageal cancer patients.

Average CT (ACT) and PET have a similar temporal resolution and it has been shown to improve registration of the CT and PET data for PET/CT imaging of the thorax. The purpose of this study was to quantify the effect of ACT attenuation correction on PET for gross tumor volume (GTV) delineation with standardized uptake value (SUV) for liver and esophageal lesions. Our study included 48 colorectal cancer patients with metastasis in the liver and 52 esophageal cancer patients. These patients underwent a routine PET/CT scan followed by a cine CT scan of the thoracic region for ACT. Differences between the two PET data sets (PET HCT and PET ACT ) corrected with the helical CT (HCT) and ACT were quantified by analyzing image alignment, maximum SUV (SUV max ), and GTV. The 67% of the colorectal and 73% of the esophageal studies demonstrated misregistration between the PET HCT and HCT data. ACT was effective in removing misregistration artifacts in 65% of the misregisted colorectal and in 76% of the misregisted esophageal cancer patients. Misregistration between the CT and PET data affected GTVs due to the change in SUV max with ACT. A change of SUV max greater than 20% between PET HCT and PET ACT was found in 15% of the colorectal and 17% of the esophageal cases. Our results demonstrated a more pronounced effect of misregistration for the smaller lesions (< 5 cm 3 ) near the diaphragm (< 5 cm). ACT was effective in improving registration between the CT and PET data in PET/CT for the colorectal and esophageal cancer patients.

Cine computed tomography without respiratory surrogate in planning stereotactic radiotherapy for non-small-cell lung cancer.

PURPOSE: To determine whether cine computed tomography (CT) can serve as an alternative to four-dimensional (4D)-CT by providing tumor motion information and producing equivalent target volumes when used to contour in radiotherapy planning without a respiratory surrogate. METHODS AND MATERIALS: Cine CT images from a commercial CT scanner were used to form maximum intensity projection and respiratory-averaged CT image sets. These image sets then were used together to define the targets for radiotherapy. Phantoms oscillating under irregular motion were used to assess the differences between contouring using cine CT and 4D-CT. We also retrospectively reviewed the image sets for 26 patients (27 lesions) at our institution who had undergone stereotactic radiotherapy for Stage I non-small-cell lung cancer. The patients were included if the tumor motion was >1 cm. The lesions were first contoured using maximum intensity projection and respiratory-averaged CT image sets processed from cine CT and then with 4D-CT maximum intensity projection and 10-phase image sets. The mean ratios of the volume magnitude were compared with intraobserver variation, the mean centroid shifts were calculated, and the volume overlap was assessed with the normalized Dice similarity coefficient index. RESULTS: The phantom studies demonstrated that cine CT captured a greater extent of irregular tumor motion than did 4D-CT, producing a larger tumor volume. The patient studies demonstrated that the gross tumor defined using cine CT imaging was similar to, or slightly larger than, that defined using 4D-CT. CONCLUSION: The results of our study have shown that cine CT is a promising alternative to 4D-CT for stereotactic radiotherapy planning.

Effects of respiration-averaged computed tomography on positron emission tomography/computed tomography quantification and its potential impact on gross tumor volume delineation.

PURPOSE: Patient respiratory motion can cause image artifacts in positron emission tomography (PET) from PET/computed tomography (CT) and change the quantification of PET for thoracic patients. In this study, respiration-averaged CT (ACT) was used to remove the artifacts, and the changes in standardized uptake value (SUV) and gross tumor volume (GTV) were quantified. METHODS AND MATERIALS: We incorporated the ACT acquisition in a PET/CT session for 216 lung patients, generating two PET/CT data sets for each patient. The first data set (PET(HCT)/HCT) contained the clinical PET/CT in which PET was attenuation corrected with a helical CT (HCT). The second data set (PET(ACT)/ACT) contained the PET/CT in which PET was corrected with ACT. We quantified the differences between the two datasets in image alignment, maximum SUV (SUV(max)), and GTV contours. RESULTS: Of the patients, 68% demonstrated respiratory artifacts in the PET(HCT), and for all patients the artifact was removed or reduced in the corresponding PET(ACT). The impact of respiration artifact was the worst for lesions less than 50 cm(3) and located below the dome of the diaphragm. For lesions in this group, the mean SUV(max) difference, GTV volume change, shift in GTV centroid location, and concordance index were 21%, 154%, 2.4 mm, and 0.61, respectively. CONCLUSION: This study benchmarked the differences between the PET data with and without artifacts. It is important to pay attention to the potential existence of these artifacts during GTV contouring, as such artifacts may increase the uncertainties in the lesion volume and the centroid location.

Four-dimensional cone beam CT with adaptive gantry rotation and adaptive data sampling.

We have developed a new four-dimensional cone beam CT (4D-CBCT) on a Varian image-guided radiation therapy system, which has radiation therapy treatment and cone beam CT imaging capabilities. We adapted the speed of gantry rotation time of the CBCT to the average breath cycle of the patient to maintain the same level of image quality and adjusted the data sampling frequency to keep a similar level of radiation exposure to the patient. Our design utilized the real-time positioning and monitoring system to record the respiratory signal of the patient during the acquisition of the CBCT data. We used the full-fan bowtie filter during data acquisition, acquired the projection data over 200 deg of gantry rotation, and reconstructed the images with a half-scan cone beam reconstruction. The scan time for a 200-deg gantry rotation per patient ranged from 3.3 to 6.6 min for the average breath cycle of 3-6 s. The radiation dose of the 4D-CBCT was about 1-2 times the radiation dose of the 4D-CT on a multislice CT scanner. We evaluated the 4D-CBCT in scanning, data processing and image quality with phantom studies. We demonstrated the clinical applicability of the 4D-CBCT and compared the 4D-CBCT and the 4D-CT scans in four patient studies. The contrast-to-noise ratio of the 4D-CT was 2.8-3.5 times of the contrast-to-noise ratio of the 4D-CBCT in the four patient studies.

Design of respiration averaged CT for attenuation correction of the PET data from PET/CT.

Our previous patient studies have shown that the use of respiration averaged computed tomography (ACT) for attenuation correction of the positron emission tomography (PET) data from PET/CT reduces the potential misalignment in the thorax region by matching the temporal resolution of the CT to that of the PET. In the present work, we investigated other approaches of acquiring ACT in order to reduce the CT dose and to improve the ease of clinical implementation. Four-dimensional CT (4DCT) data sets for ten patients (17 lung/esophageal tumors) were acquired in the thoracic region immediately after the routine PET/CT scan. For each patient, multiple sets of ACTs were generated based on both phase image averaging (phase approach) and fixed cine duration image averaging (cine approach). In the phase approach, the ACTs were calculated from CT images corresponding to the significant phases of the respiratory cycle: ACT(050phs) from end-inspiration (0%) and end-expiration (50%), ACT(2070phs) from mid-inspiration (20%) and mid-expiration (70%), ACT(4phs) from 0%, 20%, 50% and 70%, and ACT(10phs) from all ten phases, which was the original approach. In the cine approach, which does not require 4DCT, the ACTs were calculated based on the cine images from cine durations of 1 to 6 s at 1 s increments. PET emission data for each patient were attenuation corrected with each of the above mentioned ACTs and the tumor maximum standard uptake value (SUVmax), average SUV (SUVavg), and tumor volume measurements were compared. Percent differences were calculated between PET data corrected with various ACTs and that corrected with ACT(10phs). In the phase approach, the ACT(10phs) can be approximated by the ACT(4phs) to within a mean percent difference of 2% in SUV and tumor volume measurements. In cine approach, ACT(10phs) can be approximated to within a mean percent difference of 3% by ACTs computed from cine durations > or =3 s. Acquiring CT images only at the four significant phases for the ACT can reduce radiation dose to 1/3 of the current 4DCT dose; however, the implementation of this approach requires additional hardware that is not standard equipment on PET/CT scanners. In the cine approach, we recommend a duration of 6 +/- 1 s in order to include variations of respiratory patterns in a larger population. This approach can be easily implemented because cine acquisition mode is available on all GE PET/CT scanners. The CT dose in the cine approach can be reduced to approximately 5 mGy by using the lowest mA setting (10 mA), while still maintaining good quality CT data for PET attenuation correction. In our scanning protocol, the ACT is only acquired if respiration-induced misregistration is observed (determined before the PET scan is completed), and therefore patients do not receive unnecessary CT radiation dose.

Attenuation correction of PET cardiac data with low-dose average CT in PET/CT.

We proposed a low-dose average computer tomography (ACT) for attenuation correction (AC) of the PET cardiac data in PET/CT. The ACT was obtained from a cine CT scan of over one breath cycle per couch position while the patient was free breathing. We applied this technique on four patients who underwent tumor imaging with 18F-FDG in PET/CT, whose PET data showed high uptake of 18F-FDG in the heart and whose CT and PET data had misregistration. All four patients did not have known myocardiac infarction or ischemia. The patients were injected with 555-740 MBq of 18F-FDG and scanned 1 h after injection. The helical CT (HCT) data were acquired in 16 s for the coverage of 100 cm. The PET acquisition was 3 min per bed of 15 cm. The duration of cine CT acquisition per 2 cm was 5.9 s. We used a fast gantry rotation cycle time of 0.5 s to minimize motion induced reconstruction artifacts in the cine CT images, which were averaged to become the ACT images for AC of the PET data. The radiation dose was about 5 mGy for 5.9 s cine duration. The selection of 5.9 s was based on our analysis of the respiratory signals of 600 patients; 87% of the patients had average breath cycles of less than 6 s and 90% had standard deviations of less than 1 s in the period of breath cycle. In all four patient studies, registrations between the CT and the PET data were improved. An increase of average uptake in the anterior and the lateral walls up to 48% and a decrease of average uptake in the septal and the inferior walls up to 16% with ACT were observed. We also compared ACT and conventional slow scan CT (SSCT) of 4 s duration in one patient study and found ACT was better than SSCT in depicting average respiratory motion and the SSCT images showed motion-induced reconstruction artifacts. In conclusion, low-dose ACT improved registration of the CT and the PET data in the heart region in our study of four patients. ACT was superior than SSCT for depicting average respiration motion in a patient study.

Relation of external surface to internal tumor motion studied with cine CT.

The accuracy of delivering gated-radiation therapy to lung tumors using an external respiratory surrogate relies on not only interfractional and intrafractional reproducibility, but also a strong correlation between external motion and internal tumor motion. The purpose of this work was to use the cine images acquired by four-dimensional computed tomography acquisition protocol to study the relation between external surface motion and internal tumor motion. The respiratory phase information of tumor motion and chest wall motion was measured on the cine images using a proposed region-of-interest (ROI) method and compared to measurement of an external respiratory monitoring device. On eight lung patient data sets, the phase shifts were measured between (1) the signal of a real-time positioning-management (RPM) respiratory monitoring device placed in the abdominal region and four surface locations on the chest wall, (2) the RPM signal in the abdominal region and tumor motions, and (3) chest wall surface motions and tumor motions. Respiratory waveforms measured at different surface locations during the same respiratory cycle often varied and had significant phase shifts. Seven of the 8 patients showed the abdominal motion leading chest wall motion. The best correlation (smallest phase shift) was found between the abdominal motion and the superior-inferior (S-I) tumor motion. A wide range of phase shifts was observed between external surface motion and tumor anterior-posterior (A-P)/lateral motion. The result supported the placement of the RPM block in the abdominal region and suggested that during a gated therapy utilizing the RPM system, it is necessary to place the RPM block at the same location as it is during treatment simulation in order to reduce potential errors introduced by the position of the RPM block. Correlations between external motions and lateral/A-P tumor motions were inconclusive due to a combination of patient selection and the limitation of the ROI method.

Application of the electron pencil beam redefinition algorithm to electron arc therapy.

This project investigated the potential of summing fixed-beam dose distributions calculated using the pencil-beam redefinition algorithm (PBRA) at small angular steps (1 degree) to model an electron arc therapy beam. The PRBA, previously modified to model skin collimation, was modified further by incorporating two correction factors. One correction factor that is energy, SSD (source-to-surface distance), and field-width dependent constrained the calculated dose output to be the same as the measured dose output for fixed-beam geometries within the range of field widths and SSDs encountered in arc therapy. Another correction factor (single field-width correction factor for each energy) compensated for large-angle scattering not being modeled, allowing a more accurate calculation of dose output at mid arc. The PBRA was commissioned to accurately calculate dose in a water phantom for fixed-beam geometries typical of electron arc therapy. Calculated central-axis depth doses agreed with measured doses to within 2% in the low-dose gradient regions and within 1-mm in the high-dose gradient regions. Off-axis doses agreed to within 2 mm in the high-dose gradient regions and within 3% in the low-dose gradient regions. Arced-beam calculations of dose output and depth dose at mid arc were evaluated by comparing to data measured using two cylindrical water phantoms with radii of 12 and 15 cm at 10 and 15 MeV. Dose output was measured for all combinations of phantom radii of curvature, collimator widths (4, 5, and 6 cm), and arc angles (0 degrees, 20 degrees, 40 degrees, 60 degrees, 80 degrees, and 90 degrees) for both beam energies. Results showed the calculated mid-arc dose output to agree within 2% of measurement for all combinations. For a 90 degree arc angle and 5 x 20 cm2 field size, the calculated mid-arc depth dose in the low-dose gradient region agreed to within 2% of measurement for all depths at 10 MeV and for depths greater than depth of dose maximum R100 at 15 MeV. For depths in the buildup region at 15 MeV the calculations overestimated the measured dose by as much as 3.4%. Mid-arc depth dose in the high-dose gradient region agreed to within 2.2 mm of measured dose. Calculated two-dimensional relative dose distributions in the plane of rotation were compared to dose measurements using film in a cylindrical polystyrene phantom for a 90 degree arc angle and field widths of 4, 5, and 6 cm at 10 and 15 MeV. Results showed that off-axis dose at the ends of arc (without skin collimation) agreed to within 2% in the low-dose gradient region and to within 1.2 mm in the high-dose gradient region. This work showed that the accuracy of the PBRA arced-beam dose model met the criteria specified by Van Dyk et al. [Int. J. Radiat. Oncol. Biol. Phys. 26, 261-273 (1993)] with the exception of the buildup region of the 15 MeV beam. Based on the present results, results of a previous study showing acceptable accuracy in the presence of skin collimation, and results of a previous study showing acceptable accuracy in the presence of internal heterogeneities, it is concluded that the PBRA arced-beam dose model should be adequate for planning electron arc therapy.

Modeling skin collimation using the electron pencil beam redefinition algorithm.

Skin collimation is an important tool for electron beam therapy that is used to minimize the penumbra when treating near critical structures, at extended treatment distances, with bolus, or using arc therapy. It is usually made of lead or lead alloy material that conforms to and is placed on patient surface. Presently, commercially available treatment-planning systems lack the ability to model skin collimation and to accurately calculate dose in its presence. The purpose of the present work was to evaluate the use of the pencil beam redefinition algorithm (PBRA) in calculating dose in the presence of skin collimation. Skin collimation was incorporated into the PBRA by terminating the transport of electrons once they enter the skin collimator. Both fixed- and arced-beam dose calculations for arced-beam geometries were evaluated by comparing them with measured dose distributions for 10- and 15-MeV beams. Fixed-beam dose distributions were measured in water at 88-cm source-to-surface distance with an air gap of 32 cm. The 6 x 20-cm2 field (dimensions projected to isocenter) had a 10-mm thick lead collimator placed on the surface of the water with its edge 5 cm inside the field's edge located at +10 cm. Arced-beam dose distributions were measured in a 13.5-cm radius polystyrene circular phantom. The beam was arced 90 degrees (-45 degrees to +45 degrees), and 10-mm thick lead collimation was placed at +/- 30 degrees. For the fixed beam at 10 MeV, the PBRA- calculated dose agreed with measured dose to within 2.0-mm distance to agreement (DTA) in the regions of high-dose gradient and 2.0% in regions of low dose gradient. At 15 MeV, the PBRA agreed to within a 2.0-mm DTA in the regions of high-dose gradient; however, the PBRA underestimated the dose by as much as 5.3% over small regions at depths less than 2 cm because it did not model electrons scattered from the edge of the skin collimation. For arced beams at 10 MeV, the agreement was 1-mm DTA in the high-dose gradient regions, and 2% in the low-dose gradient regions. For arced beams at 15 MeV, the agreement was 1 mm in the high-dose gradient regions, and in the low-dose gradient region at depth less than 2 cm, as much as 5% dose difference was observed. This study demonstrated the ease with which skin collimation can be incorporated into the PBRA. The good agreement of PBRA calculated with measured dose shows that the PBRA is likely sufficiently accurate for clinical use in the presence of skin collimation for electron arc therapy. To further improve the accuracy of the PBRA in regions having significant electrons scattered from the edge of the skin collimation would require transporting the electrons through the lead skin collimation near its edges.

Linearity and uniformity response as an indicator of performance for Agfa ADC-MD10 computed radiography plates.

Computed radiography (CR) plates are currently used in radiation therapy clinics to acquire digital radiographic images for the purpose of verifying the treatment field size, shape, and location. Each CR plate may be used numerous times, and the use of these digital images allows for easy storage and retrieval of patient data. Over prolonged repeat exposures of the CR plates, however, the image quality begins to degrade, making it increasingly more difficult for the therapists and physicians to determine where one anatomical structure begins, and the other ends. The purpose of this project was to analyze and compare the linearity and uniformity responses of new CR plates, versus CR plates that have been used clinically for a period of 2 years, and determine whether linearity or uniformity response may be used as an indicator of image quality degradation. To determine this, 44 old Agfa MD10 CR plates and 56 new Agfa MD10 CR plates were tested. When comparing the results of the uniformity test, we found both the old and the new plates varied from approximately 0.5% to 3.2%. When comparing the results of the linearity test, we found that the correlation coefficient, R(2), for both the old and the new plates varied from approximately 0.996 to 0.998, with the mean values being 0.9972 and 0.9979, respectively. We concluded that linearity and uniformity response cannot be used as an effective method for the evaluation of CR plate performance. Additional research is currently underway to evaluate various other methods of assessing CR plate performance.