Machine learning for automated quality assurance in radiotherapy: A proof of principle using EPID data description.

PURPOSE: Developing automated methods to identify task-driven quality assurance (QA) procedures is key toward increasing safety, efficacy, and efficiency. We investigate the use of machine learning (ML) methods for possible visualization, automation, and targeting of QA, and assess its performance using multi-institutional data. METHODS: To enable automated analysis of QA data given its higher dimensional nature, we used nonlinear kernel mapping with support vector data description (SVDD) driven approaches. Instead of using labeled data as in typical support vector machine (SVM) applications, which requires exhaustive annotation, we applied a clustering extension of SVDD, which identifies the minimal enclosing hypersphere in the feature space defined by a kernel function separating normal operations from possible failures (i.e., outliers). In our case, QA test data are mapped by a Gaussian kernel to a higher dimensional feature space and then the minimal enclosing sphere was identified. This sphere, when mapped back to the input data space along the principal components, can separate the data into several components, each enclosing a separate cluster of QA points that could be used to evaluate tolerance boundaries and test reliability. We evaluated this approach for gantry sag, radiation field shift, and [multileaf collimator (MLC)] offset data acquired using electronic portal imaging devices (EPID), as representative examples. RESULTS: Data from eight LINACS and seven institutions (n = 119) were collected. A standardized EPID image of a phantom with fiducials provided deviation estimates between the radiation field and phantom center at four cardinal gantry angles. Deviation measurements in the horizontal direction (0°, 180°) were used to determine the gantry sag and deviations in the vertical direction (90°, 270°) were used to determine the field shift. These measurements were fed into the SVDD clustering algorithm with varying hypersphere radii (Gaussian widths). For gantry sag analysis, two clusters were identified one of which contained 2.5% of the outliers and also exceeded the 1 mm tolerance set by TG-142. In the case of field shifts, SVM clustering identified two distinct classes of measurements primarily driven by variations in the second principal component at 270°. Results from MLC analysis identified one outlier cluster (0.34%) along Leaf offset Constancy (LoC) axis that coincided with TG-142 limits. CONCLUSION: Machine learning methods based on SVDD clustering are promising for developing automated QA tools and providing insights into their reliability and reproducibility.

Automating linear accelerator quality assurance.

PURPOSE: The purpose of this study was 2-fold. One purpose was to develop an automated, streamlined quality assurance (QA) program for use by multiple centers. The second purpose was to evaluate machine performance over time for multiple centers using linear accelerator (Linac) log files and electronic portal images. The authors sought to evaluate variations in Linac performance to establish as a reference for other centers. METHODS: The authors developed analytical software tools for a QA program using both log files and electronic portal imaging device (EPID) measurements. The first tool is a general analysis tool which can read and visually represent data in the log file. This tool, which can be used to automatically analyze patient treatment or QA log files, examines the files for Linac deviations which exceed thresholds. The second set of tools consists of a test suite of QA fields, a standard phantom, and software to collect information from the log files on deviations from the expected values. The test suite was designed to focus on the mechanical tests of the Linac to include jaw, MLC, and collimator positions during static, IMRT, and volumetric modulated arc therapy delivery. A consortium of eight institutions delivered the test suite at monthly or weekly intervals on each Linac using a standard phantom. The behavior of various components was analyzed for eight TrueBeam Linacs. RESULTS: For the EPID and trajectory log file analysis, all observed deviations which exceeded established thresholds for Linac behavior resulted in a beam hold off. In the absence of an interlock-triggering event, the maximum observed log file deviations between the expected and actual component positions (such as MLC leaves) varied from less than 1% to 26% of published tolerance thresholds. The maximum and standard deviations of the variations due to gantry sag, collimator angle, jaw position, and MLC positions are presented. Gantry sag among Linacs was 0.336 ± 0.072 mm. The standard deviation in MLC position, as determined by EPID measurements, across the consortium was 0.33 mm for IMRT fields. With respect to the log files, the deviations between expected and actual positions for parameters were small (<0.12 mm) for all Linacs. Considering both log files and EPID measurements, all parameters were well within published tolerance values. Variations in collimator angle, MLC position, and gantry sag were also evaluated for all Linacs. CONCLUSIONS: The performance of the TrueBeam Linac model was shown to be consistent based on automated analysis of trajectory log files and EPID images acquired during delivery of a standardized test suite. The results can be compared directly to tolerance thresholds. In addition, sharing of results from standard tests across institutions can facilitate the identification of QA process and Linac changes. These reference values are presented along with the standard deviation for common tests so that the test suite can be used by other centers to evaluate their Linac performance against those in this consortium.

Stereotactic body radiation therapy for curative treatment of adrenal metastases.

The detection of oligometastatic adrenal metastases is increasing and there are limited data supporting the use of curative intent stereotactic body radiation therapy (SBRT) to treat patients with limited metastatic disease with adrenal involvement. Therefore, we utilized a prospectively maintained database of consecutive patients treated with SBRT for limited metastatic disease (≤5 sites) to identify patients with adrenal metastases. Patients were either treated on a three-fraction dose escalation protocol or a ten fraction off-protocol regimen. Outcomes including treated-metastasis control (TMC), distant control (DC), and overall survival (OS) were calculated by the Kaplan-Meier method. Ten patients with 13 adrenal metastases were identified for this case series. The median follow-up was 14.9 months. No patient experienced grade 3 toxicity. The most common grade 1-2 acute toxicities were fatigue (80%) and GI toxicity (40%). One patient experienced late grade 2 adrenal insufficiency. Overall, the 1-year TMC rate was 73%, DC was 30%, and OS was 90%. Three treated adrenal metastases progressed, all receiving the lowest BED10 (43.2 Gy), corresponding to 24 Gy in 3 fractions. After treatment of adrenal metastases with SBRT, the median time to salvage chemotherapy was 5.3 months (range 1.0-38.8 months) and 1-year freedom from salvage chemotherapy was 44%. These results suggest that SBRT to adrenal metastases was tolerated with low toxicity in limited metastatic patients and control rates are promising. This study supports the growing body of literature treating patients with adrenal metastases with SBRT.

The utility of FDG-PET for assessing outcomes in oligometastatic cancer patients treated with stereotactic body radiotherapy: a cohort study.

BACKGROUND: Studies suggest that patients with metastases limited in number and destination organ benefit from metastasis-directed therapy. Stereotactic body radiotherapy (SBRT) is commonly used for metastasis directed therapy in this group. However, the characterization of PET response following SBRT is unknown in this population. We analyzed our cohort of patients to describe the PET response following SBRT. METHODS: Patients enrolled on a prospective dose escalation trial of SBRT to all known sites of metastatic disease were reviewed to select patients with pre- and post-therapy PET scans. Response to SBRT was characterized on PET imaging based on standard PET response criteria and compared to CT based RECIST criteria for each treated lesion. RESULTS: 31 patients had PET and CT data available before and after treatment for analysis in this study. In total, 58 lesions were treated (19 lung, 11 osseous, 11 nodal, 9 liver, 6 adrenal and 2 soft tissue metastases). Median follow-up was 14 months (range: 3-41). Median time to first post-therapy PET was 1.2 months (range; 0.5-4.1). On initial post-therapy PET evaluation, 96% (56/58) of treated metastases responded to therapy. 60% (35/58) had a complete response (CR) on PET and 36% (21/58) had a partial response (PR). Of 22 patients with stable disease (SD) on initial CT scan, 13 had CR on PET, 8 had PR, and one had SD. Of 21 metastases with PET PR, 38% became CR, 52% remained PR, and 10% had progressive disease on follow-up PET. 10/35 lesions (29%) with an initial PET CR progressed on follow-up PET scan with median time to progression of 4.11 months (range: 2.75-9.56). Higher radiation dose correlated with long-term PET response. CONCLUSIONS: PET response to SBRT enables characterization of metastatic response in tumors non-measurable by CT. Increasing radiation dose is associated with prolonged complete response on PET.

Improvements in dose accuracy delivered with static-MLC IMRT on an integrated linear accelerator control system.

PURPOSE: Dose accuracy has been shown to vary with dose per segment and dose rate when delivered with static multileaf collimator (SMLC) intensity modulated radiation therapy (IMRT) by Varian C-series MLC controllers. The authors investigated the impact of monitor units (MUs) per segment and dose rate on the dose delivery accuracy of SMLC-IMRT fields on a Varian TrueBeam linear accelerator (LINAC), which delivers dose and manages motion of all components using a single integrated controller. METHODS: An SMLC sequence was created consisting of ten identical 10 × 10 cm(2) segments with identical MUs. Beam holding between segments was achieved by moving one out-of-field MLC leaf pair. Measurements were repeated for various combinations of MU/segment ranging from 1 to 40 and dose rates of 100-600 MU/min for a 6 MV photon beam (6X) and dose rates of 800-2400 MU/min for a 10 MV flattening-filter free photon (10XFFF) beam. All measurements were made with a Farmer (0.6 cm(3)) ionization chamber placed at the isocenter in a solid-water phantom at 10 cm depth. The measurements were performed on two Varian LINACs: C-series Trilogy and TrueBeam. Each sequence was delivered three times and the dose readings for the corresponding segments were averaged. The effects of MU/segment, dose rate, and LINAC type on the relative dose variation (Δ(i)) were compared using F-tests (α = 0.05). RESULTS: On the Trilogy, large Δ(i) was observed in small MU segments: at 1 MU/segment, the maximum Δ(i) was 10.1% and 57.9% at 100 MU/min and 600 MU/min, respectively. Also, the first segment of each sequence consistently overshot (Δ(i) > 0), while the last segment consistently undershot (Δ(i) < 0). On the TrueBeam, at 1 MU/segment, Δ(i) ranged from 3.0% to 4.5% at 100 and 600 MU/min; no obvious overshoot/undershoot trend was observed. F-tests showed statistically significant difference [(1 - ß) =1.0000] between the Trilogy and the TrueBeam up to 10 MU/segment, at all dose rates greater than 100 MU/min. The linear trend of decreasing dose accuracy as a function of increasing dose rate on the Trilogy is no longer apparent on TrueBeam, even for dose rates as high as 2400 MU/min. Dose inaccuracy averaged over all ten segments in each beam delivery sequence was larger for Trilogy than TrueBeam, with the largest discrepancy (0.2% vs 3%) occurring for 1 MU/segment beams at both 300 and 600 MU/min. CONCLUSIONS: Earlier generations of Varian LINACs exhibited large dose variations for small MU segments in SMLC-IMRT delivery. Our results confirmed these findings. The dose delivery accuracy for SMLC-IMRT is significantly improved on TrueBeam compared to Trilogy for every combination of low MU/segment (1-10) and high dose rate (200-600 MU/min), in part due to the faster sampling rate (100 vs 20 Hz) and enhanced electronic integration of the MLC controller with the LINAC. SMLC-IMRT can be implemented on TrueBeam with higher dose accuracy per beam (±0.2% vs ±3%) than previous generations of Varian C-series LINACs for 1 MU/segment delivered at 600 MU/min).

Development of a frameless stereotactic radiosurgery system based on real-time 6D position monitoring and adaptive head motion compensation.

Stereotactic radiosurgery delivers radiation with great spatial accuracy. To achieve sub-millimeter accuracy for intracranial SRS, a head ring is rigidly fixated to the skull to create a fixed reference. For some patients, the invasiveness of the ring can be highly uncomfortable and not well tolerated. In addition, placing and removing the ring requires special expertise from a neurosurgeon, and patient setup time for SRS can often be long. To reduce the invasiveness, hardware limitations and setup time, we are developing a system for performing accurate head positioning without the use of a head ring. The proposed method uses real-time 6D optical position feedback for turning on and off the treatment beam (gating) and guiding a motor-controlled 3D head motion compensation stage. The setup consists of a central control computer, an optical patient motion tracking system and a 3D motion compensation stage attached to the front of the LINAC couch. A styrofoam head cast was custom-built for patient support and was mounted on the compensation stage. The motion feedback of the markers was processed by the control computer, and the resulting motion of the target was calculated using a rigid body model. If the target deviated beyond a preset position of 0.2 mm, an automatic position correction was performed with stepper motors to adjust the head position via the couch mount motion platform. In the event the target deviated more than 1 mm, a safety relay switch was activated and the treatment beam was turned off. The feasibility of the concept was tested using five healthy volunteers. Head motion data were acquired with and without the use of motion compensation over treatment times of 15 min. On average, test subjects exceeded the 0.5 mm tolerance 86% of the time and the 1.0 mm tolerance 45% of the time without motion correction. With correction, this percentage was reduced to 5% and 2% for the 0.5 mm and 1.0 mm tolerances, respectively.

Intensity-modulated radiotherapy for neoadjuvant treatment of gastric cancer.

Radiation therapy plays an integral role in the treatment of gastric cancer in the postsurgery setting, the inoperable/palliative setting, and, as in the case of the current report, in the setting of neoadjuvant therapy prior to surgery. Typically, anterior-posterior/posterior-anterior (AP/PA) or 3-field techniques are used. In this report, we explore the use of intensity-modulated radiotherapy (IMRT) treatment in a patient whose care was transferred to our institution after 3-field radiotherapy (RT) was given to a dose of 30 Gy at an outside institution. If the 3-field plan were continued to 50 Gy, the volume of irradiated liver receiving greater than 30 Gy would have been unacceptably high. To deliver the final 20 Gy, an opposed parallel AP/PA plan and an IMRT plan were compared to the initial 3-field technique for coverage of the target volume as well as dose to the kidneys, liver, small bowel, and spinal cord. Comparison of the 3 treatment techniques to deliver the final 20 Gy revealed reduced median and maximum dose to the whole kidney with the IMRT plan. For this 20-Gy boost, the volume of irradiated liver was lower for both the IMRT plan and the AP/PA plan vs. the 3-field plan. Comparing the IMRT boost plan to the AP/PA boost-dose range (<10 Gy) in comparison to the AP/PA plan; however, the IMRT plan irradiated a smaller liver volume within the higher dose region (>10 Gy) in comparison to the AP/PA plan. The IMRT boost plan also irradiated a smaller volume of the small bowel compared to both the 3-field plan and the AP/PA plan, and also delivered lower dose to the spinal cord in comparison to the AP/PA plan. Comparison of the composite plans revealed reduced dose to the whole kidney using IMRT. The V20 for the whole kidney volume for the composite IMRT plan was 30% compared to approximately 60% for the composite AP/PA plan. Overall, the dose to the liver receiving greater than 30 Gy was lower for the composite IMRT plan and was well below acceptable limits. In conclusion, our study suggests a dosimetric benefit of IMRT over conventional planning, and suggests an important role for IMRT in the neoadjuvant treatment of gastric cancer.

Intensity-modulated radiation therapy (IMRT) in the treatment of anal cancer: toxicity and clinical outcome.

PURPOSE: To assess survival, local control, and toxicity of intensity modulated radiation therapy (IMRT) in squamous cell carcinoma of the anal canal. METHODS AND MATERIALS: Seventeen patients were treated with nine-field IMRT plans. Thirteen received concurrent 5-fluorouracil and mitomycin C, whereas 1 patient received 5-fluorouracil alone. Seven patients were planned with three-dimensional anteroposterior/posterior-anterior (AP/PA) fields for dosimetric comparison to IMRT. RESULTS: Compared with AP/PA, IMRT reduced the mean and threshold doses to small bowel, bladder, and genitalia. Treatment was well tolerated, with no Grade > or =3 acute nonhematologic toxicity. There were no treatment breaks attributable to gastrointestinal or skin toxicity. Of patients who received mitomycin C, 38% experienced Grade 4 hematologic toxicity. IMRT did not afford bone marrow sparing, possibly resulting from the clinical decision to prescribe 45 Gy to the whole pelvis in most patients, vs. the Radiation Therapy Oncology Group-recommended 30.6 Gy whole pelvic dose. Three of 17 patients, who did not achieve a complete response, proceeded to an abdominoperineal resection and colostomy. At a median follow-up of 20.3 months, there were no other local failures. Two-year overall survival, disease-free survival, and colostomy-free survival are: 91%, 65%, and 82% respectively. CONCLUSIONS: In this hypothesis-generating analysis, the acute toxicity and clinical outcome with IMRT in the treatment of anal cancer is encouraging. Compared with historical controls, local control is not compromised despite efforts to increase conformality and reduce normal structure dose.

Analysis of acute toxicity with use of transabdominal ultrasonography for prostate positioning during intensity-modulated radiotherapy.

OBJECTIVES: To analyze the effects of the B-mode ultrasound acquisition and targeting (BAT) system for positioning of patients with prostate cancer receiving intensity-modulated radiotherapy on acute gastrointestinal (GI) and genitourinary (GU) toxicity. METHODS: The records of 50 consecutive patients treated using the BAT system were reviewed. Additionally, a comparison (no-BAT) group (ie, a group without a BAT study) treated in a similar manner was identified. The no-BAT group consisted of 49 patients treated immediately before the BAT group. For the two groups, the target definitions and dose prescriptions were identical, the treatment plan acceptance criteria were identical, and intensity-modulated radiotherapy was used for all patients. The daily BAT movements were charted in each of the three principal directions. Acute toxicity was scored for all patients according to the Radiation Therapy Oncology Group GI and GU acute toxicity scales. RESULTS: The GU toxicity rates for the BAT versus no-BAT groups were grade 0 in 20% versus 14%; grade 1 in 38% versus 47%; grade 2 in 38% versus 39%; and grade 3 in 4% versus 0%, respectively (P = 0.284). The corresponding GI toxicity rates were grade 0 in 42% versus 27%; grade 1 in 28% versus 29%; and grade 2 in 30% versus 45% (P = 0.040). The incidence of GU and GI toxicity did not correlate with the directions or size of the BAT moves. Regression analysis revealed that for acute GI toxicity, the only variable reaching statistical significance was BAT use; no variable, including BAT use, reached statistical significance for acute GU toxicity. CONCLUSIONS: The use of the BAT system did not change the rate of acute GU toxicity but did reduce the rate of acute GI toxicity.