Improving the Quality of Care in Radiation Oncology using Artificial Intelligence.

Radiation therapy is a complex process involving multiple professionals and steps from simulation to treatment planning to delivery, and these procedures are prone to error. Additionally, the imaging and treatment delivery equipment in radiotherapy is highly complex and interconnected and represents another risk point in the quality of care. Numerous quality assurance tasks are carried out to ensure quality and to detect and prevent potential errors in the process of care. Recent developments in artificial intelligence provide potential tools to the radiation oncology community to improve the efficiency and performance of quality assurance efforts. Targets for artificial intelligence enhancement include the quality assurance of treatment plans, target and tissue structure delineation used in the plans, delivery of the plans and the radiotherapy delivery equipment itself. Here we review recent developments of artificial intelligence applications that aim to improve quality assurance processes in radiation therapy and discuss some of the challenges and limitations that require further development work to realise the potential of artificial intelligence for quality assurance.

Quantifying the performance of in vivo portal dosimetry in detecting four types of treatment parameter variations.

PURPOSE: To quantify the ability of electronic portal imaging device (EPID) dosimetry used during treatment (in vivo) in detecting variations that can occur in the course of patient treatment. METHODS: Images of transmitted radiation from in vivo EPID measurements were converted to a 2D planar dose at isocenter and compared to the treatment planning dose using a prototype software system. Using the treatment planning system (TPS), four different types of variability were modeled: overall dose scaling, shifting the positions of the multileaf collimator (MLC) leaves, shifting of the patient position, and changes in the patient body contour. The gamma pass rate was calculated for the modified and unmodified plans and used to construct a receiver operator characteristic (ROC) curve to assess the detectability of the different parameter variations. The detectability is given by the area under the ROC curve (AUC). The TPS was also used to calculate the impact of the variations on the target dose-volume histogram. RESULTS: Nine intensity modulation radiation therapy plans were measured for four different anatomical sites consisting of 70 separate fields. Results show that in vivo EPID dosimetry was most sensitive to variations in the machine output, AUC = 0.70 - 0.94, changes in patient body habitus, AUC = 0.67 - 0.88, and systematic shifts in the MLC bank positions, AUC = 0.59 - 0.82. These deviations are expected to have a relatively small clinical impact [planning target volume (PTV) D99 change <7%]. Larger variations have even higher detectability. Displacements in the patient's position and random variations in MLC leaf positions were not readily detectable, AUC < 0.64. The D99 of the PTV changed by up to 57% for the patient position shifts considered here. CONCLUSIONS: In vivo EPID dosimetry is able to detect relatively small variations in overall dose, systematic shifts of the MLC's, and changes in the patient habitus. Shifts in the patient's position which can introduce large changes in the target dose coverage were not readily detected.

Consensus recommendations for incident learning database structures in radiation oncology.

PURPOSE: Incident learning plays a key role in improving quality and safety in a wide range of industries and medical disciplines. However, implementing an effective incident learning system is complex, especially in radiation oncology. One current barrier is the lack of technical standards to guide users or developers. This report, the product of an initiative by the Work Group on Prevention of Errors in Radiation Oncology of the American Association of Physicists in Medicine, provides technical recommendations for the content and structure of incident learning databases in radiation oncology. METHODS: A panel of experts was assembled and tasked with developing consensus recommendations in five key areas: definitions, process maps, severity scales, causality taxonomy, and data elements. Experts included representatives from all major North American radiation oncology organizations as well as users and developers of public and in-house reporting systems with over two decades of collective experience. Recommendations were developed that take into account existing incident learning systems as well as the requirements of outside agencies. RESULTS: Consensus recommendations are provided for the five major topic areas. In the process mapping task, 91 common steps were identified for external beam radiation therapy and 88 in brachytherapy. A novel feature of the process maps is the identification of "safety barriers," also known as critical control points, which are any process steps whose primary function is to prevent errors or mistakes from occurring or propagating through the radiotherapy workflow. Other recommendations include a ten-level medical severity scale designed to reflect the observed or estimated harm to a patient, a radiation oncology-specific root causes table to facilitate and regularize root-cause analyses, and recommendations for data elements and structures to aid in development of electronic databases. Also presented is a list of key functional requirements of any reporting system. CONCLUSIONS: Incident learning is recognized as an invaluable tool for improving the quality and safety of treatments. The consensus recommendations in this report are intended to facilitate the implementation of such systems within individual clinics as well as on broader national and international scales.

Localized CT-guided irradiation inhibits neurogenesis in specific regions of the adult mouse brain.

Radiation is used in the study of neurogenesis in the adult mouse both as a model for patients undergoing radiation therapy for CNS malignancies and as a tool to interrupt neurogenesis. We describe the use of a dedicated CT-guided precision device to irradiate specific sub-regions of the adult mouse brain. Improved CT visualization was accomplished with intrathecal injection of iodinated contrast agent, which enhances the lateral ventricles. T2-weighted MRI images were also used for target localization. Visualization of delivered beams (10 Gy) in tissue was accomplished with immunohistochemical staining for the protein γ-H2AX, a marker of DNA double-strand breaks. γ-H2AX stains showed that the lateral ventricle wall could be targeted with an accuracy of 0.19 mm (n = 10). In the hippocampus, γ-H2AX staining showed that the dentate gyrus can be irradiated unilaterally with a localized arc treatment. This resulted in a significant decrease of proliferative neural progenitor cells as measured by Ki-67 staining (P < 0.001) while leaving the contralateral side intact. Two months after localized irradiation, neurogenesis was significantly inhibited in the irradiated region as seen with EdU/NeuN double labeling (P < 0.001). Localized radiation in the rodent brain is a promising new tool for the study of neurogenesis.

Respiration-correlated spiral CT: a method of measuring respiratory-induced anatomic motion for radiation treatment planning.

We describe a method for generating CT images at multiple respiratory phases with a single spiral CT scan, referred to as respiratory-correlated spiral CT (RCCT). RCCT relies on a respiration wave form supplied by an external patient monitor. During acquisition this wave form is recorded along with the initiation time of the CT scan, so as to "time stamp" each reconstructed slice with the phase of the respiratory cycle. By selecting the appropriate slices, a full CT image set is generated at several phases, typically 7-11 per cycle. The CT parameters are chosen to optimize the temporal resolution while minimizing the spatial gap between slices at successive respiratory cycles. Using a pitch of 0.5, a gantry rotation period of 1.5 s, and a 180 degrees reconstruction algorithm results in approximately 5 mm slice spacing at a given phase for typical respiration periods, and a respiratory motion within each slice that is acceptably small, particularly near end expiration or end inspiration where gated radiotherapy is to occur. We have performed validation measurements on a phantom with a moving sphere designed to simulate respiration-induced tumor motion. RCCT scans of the phantom at respiratory periods of 4, 5, and 6 s show good agreement of the sphere's motion with that observed under fluoroscopic imaging. The positional deviations in the sphere's centroid between RCCT and fluoroscopy are 1.1+/-0.9 mm in the transaxial direction (average over all scans at all phases +/-1 s.d.) and 1.2+/-1.0 mm in the longitudinal direction. Reconstructed volumes match those expected on the basis of stationary-phantom scans to within 5% in all cases. The surface distortions of the reconstructed sphere, as quantified by deviations from a mathematical reference sphere, are similar to those from a stationary phantom scan and are correlated with the speed of the phantom. A RCCT scan of the phantom undergoing irregular motion, demonstrates that successful reconstruction can be achieved even with irregular respiration. Limitations from x-ray tube heating in our current CT unit restrict the length of the scan region to 9 cm for the RCCT settings used, though this will not be a limitation for a multislice scanner. RCCT offers an alternative to the current method of respiration-triggered axial scans. Multiple phases of respiration are imaged with RCCT in approximately the same scanning time required to image a single phase with a triggered axial scan. RCCT scans can be used in connection with respiratory-gated treatment to identify the patient-specific phase of minimum tumor motion, determine residual tumor motion within the gate interval, and compare treatment plans at different phases.

Evaluation of respiratory movement during gated radiotherapy using film and electronic portal imaging.

PURPOSE: To evaluate the effectiveness of a commercial system(1) in reducing respiration-induced treatment uncertainty by gating the radiation delivery. METHODS AND MATERIALS: The gating system considered here measures respiration from the position of a reflective marker on the patient's chest. Respiration-triggered planning CT scans were obtained for 8 patients (4 lung, 4 liver) at the intended phase of respiration (6 at end expiration and 2 at end inspiration). In addition, fluoroscopic movies were recorded simultaneously with the respiratory waveform. During the treatment sessions, gated localization films were used to measure the position of the diaphragm relative to the vertebral bodies, which was compared to the reference digitally reconstructed radiograph derived from the respiration-triggered planning CT. Variability was quantified by the standard deviation about the mean position. We also assessed the interfraction variability of soft tissue structures during gated treatment in 2 patients using an amorphous silicon electronic portal imaging device. RESULTS: The gated localization films revealed an interfraction patient-averaged diaphragm variability of 2.8 +/- 1.0 mm (error bars indicate standard deviation in the patient population). The fluoroscopic data yielded a patient-averaged intrafraction diaphragm variability of 2.6 +/- 1.7 mm. With no gating, this intrafraction excursion became 6.9 +/- 2.1 mm. In gated localization films, the patient-averaged mean displacement of the diaphragm from the planning position was 0.0 +/- 3.9 mm. However, in 4 of the 8 patients, the mean (over localization films) displacement was >4 mm, indicating a systematic displacement in treatment position from the planned one. The position of soft tissue features observed in portal images during gated treatments over several fractions showed a mean variability between 2.6 and 5.7 mm. The intrafraction variability, however, was between 0.6 and 1.4 mm, indicating that most of the variability was due to patient setup errors rather than to respiratory motion. CONCLUSIONS: The gating system evaluated here reduces the intra- and interfraction variability of anatomy due to respiratory motion. However, systematic displacements were observed in some cases between the location of an anatomic feature at simulation and its location during treatment. Frequent monitoring is advisable with film or portal imaging.

Cone-beam CT with megavoltage beams and an amorphous silicon electronic portal imaging device: potential for verification of radiotherapy of lung cancer.

We investigate the potential of megavoltage (MV) cone-beam CT with an amorphous silicon electronic portal imaging device (EPID) as a tool for patient position verification and tumor/organ motion studies in radiation treatment of lung tumors. We acquire 25 to 200 projection images using a 22 x 29 cm EPID. The acquisition is automatic and requires 7 minutes for 100 projections; it can be synchronized with respiratory gating. From these images, volumetric reconstruction is accomplished with a filtered backprojection in the cone-beam geometry. Several important prereconstruction image corrections, such as detector sag, must be applied. Tests with a contrast phantom indicate that differences in electron density of 2% can be detected with 100 projections, 200 cGy total dose. The contrast-to-noise ratio improves as the number of projections is increased. With 50 projections (100 cGy), high contrast objects are visible, and as few as 25 projections yield images with discernible features. We identify a technique of acquiring projection images with conformal beam apertures, shaped by a multileaf collimator, to reduce the dose to surrounding normal tissue. Tests of this technique on an anthropomorphic phantom demonstrate that a gross tumor volume in the lung can be accurately localized in three dimensions with scans using 88 monitor units. As such, conformal megavoltage cone-beam CT can provide three-dimensional imaging of lung tumors and may be used, for example, in verifying respiratory gated treatments.

Radioimmunoassay and characterization of woodchuck hepatitis virus core antigen and antibody.

Solid-phase radioimmunoassays for woodchuck hepatitis virus core antigen (WHcAg) and antibody (anti-WHc) were developed. WHcAg in woodchuck liver homogenates was characterized by ultracentrifugation in CsCl gradients; both heavy (1.35 g/cm3) and light (1.31 g/cm3) cores were obtained from the liver of an animal during acute WHV infection, which is consistent with observations in hepatitis B virus infection in man. Endpoint titers of anti-WHc were higher in chronic WHV carriers than in animals recovered from acute infections. Both IgM and IgG anti-WHc antibodies were produced by infected woodchucks. A survey of colony woodchucks demonstrated that 88/89 animals having one or more markers of past or ongoing WHV infection were positive for anti-WHc. Thus, serum anti-WHc appears to be a sensitive marker of WHV infection.

Core antigen and antibody in woodchucks after infection with woodchuck hepatitis virus.

The woodchuck hepatitis virus is a naturally occurring hepatitis B-like virus that infects the eastern woodchuck. Direct immunofluorescence staining for woodchuck hepatitis virus core antigen in liver biopsies demonstrated the presence of this antigen in 14 of 17 chronically infected woodchucks, and in 8 of 10 woodchucks undergoing acute infections. Fluorescent localization of woodchuck hepatitis virus core antigen was typically cytoplasmic, and this was confirmed further by electron microscopy. Experimental infection with woodchuck hepatitis virus was achieved in four of four woodchucks inoculated with serum from chronic carrier woodchucks. All infected animals developed a self-limited disease characterized by seroconversion to antibodies against the major viral antigens (core and surface antigens); naturally acquired acute infection demonstrated a similar course. A chimpanzee seronegative for all markers of hepatitis B virus developed a subclinical infection after inoculation with woodchuck hepatitis virus.

Transmission of the hepatitis B virus-associated delta agent to the eastern woodchuck.

delta agent of human origin was inoculated into four woodchucks chronically infected with woodchuck hepatitis virus (WHV). The animals developed delta infections with serologic patterns similar to those previously observed in human and chimpanzee infections. delta antigen was detected transiently in serum and liver and was followed by seroconversion to anti-delta antibody. Analogous to the chimpanzee model of delta infection, serum and hepatocyte markers of WHV were suppressed in the woodchuck during acute delta infection. The suppression of WHV DNA in serum was evident only during the time of delta-antigen positivity, while the inhibition of other WHV markers was more protracted. The delta antigen in woodchuck sera circulated as an internal component of a particle similar in size to the human delta particle (36-nm diameter) and was encapsidated by the woodchuck hepatitis virus surface antigen; delta antigen from infected woodchuck and chimpanzee livers had similar biophysical properties. Histologic analysis showed that experimental delta infection is associated with a transient acute hepatitis in woodchucks and loss of hepatocytes carrying WHV antigens. The lesions differed from the conspicuous hepatitis associated with reappearance of WHV replication. Hepatitis B-like viruses, therefore, appear to provide the requisite helper functions for delta replication and the woodchuck represents a useful model for study of the virology and pathology of the delta agent.

Monoclonal antibodies to respiratory syncytial virus: detection of virus neutralization and other antigen-antibody systems using infected human and murine cells.

Monoclonal antibodies to human respiratory syncytial (RS) virus-specific antigens can be obtained without preliminary recourse to large-scale culture and purification of the virion. Lytically infected human and persistently infected murine cultured cells expressing RS virus-specific cell surface and cytoplasmic antigens were substituted as priming immunogens and as substrates in solid-phase antibody radioimmunoassays. Seven hybridoma clones secreting murine IgG of either the gamma 1 or the gamma 2a subclass bearing kappa light chains were isolated. Two of the antibodies were specific for cell surface viral antigens, but only one was able to neutralize RS virus infectivity. The five remaining antibodies did not neutralize virus infectivity and were specific for viral antigens associated with large cytoplasmic inclusions as judged by indirect immunofluorescence (IF) analysis on fixed infected cells. Similar IF analysis using live cells revealed that those antigens, associated with the cytoplasmic inclusions in both the human and murine infected cells, were not expressed on the cell surface of the live infected human cells, but were expressed on the cell surface of the live infected murine cells. Monoclonal antibodies generated via the present system will prove useful in the immunological analysis of viral components which are specific pathogenic functions, such as infectivity, and those which may be abnormally exposed at the surface of persistently infected cells.

Demonstration of subpopulations of Dane particles.

Two populations of Dane particles were isolated from the plasma of individuals carrying hepatitis B surface antigen. These populations had densities in CsCl of 1.22 and 1.20 g/ml. Endogenous DNA polymerase activity was found to be associated only with the heavier of these two populations. Using a positive stain, electron microscopic examination of these particles suggested that the heavier the particle contained nucleic acid in its core whereas the lighter particle appeared empty. Cores isolated from Dane particles with densities of 1.22 and 1.20 g/ml banded in CsCl at densities of 1.36 and 1.30 g/ml, respectively. Endogenous DNA polymerase activity was associated only with the higher density core particles.

Biochemical characterization of Australia antigen. Evidence for defective particles of hepatitis B virus.

Australia antigen exists in the sera of chronic carriers in several particulate forms, one of which may represent the virion of hepatitis B. This report describes the existence of subpopulations of these 43-nm particles, the Dane particles, on the basis of the staining properties of their internal cores and banding characteristics in cesium chloride (CsCl) density gradients. These data suggested that only a minor proportion of Dane particles contained an intact viral genome and represent the standard infectious virus of hepatitis B. The bulk of the Dane particles appeared to be deficient in viral nucleic acid and, as defective interfering particles, may specifically interfere with the growth of standard virus. Such defective interfering particles could thereby play a role in the persistence of HBV infection in man.