Matthew J. Nyflot

Quantitative radiomics: impact of stochastic effects on textural feature analysis implies the need for standards

Abstract. Image heterogeneity metrics such as textural features are an active area of research for evaluating clinical outcomes with positron emission tomography (PET) imaging and other modalities. However, the effects of stochastic image acquisition noise on these metrics are poorly understood. We performed a simulation study by generating 50 statistically independent PET images of the NEMA IQ phantom with realistic noise and resolution properties. Heterogeneity metrics based on gray-level intensity histograms, co-occurrence matrices, neighborhood difference matrices, and zone size matrices were evaluated within regions of interest surrounding the lesions. The impact of stochastic variability was evaluated with percent difference from the mean of the 50 realizations, coefficient of variation and estimated sample size for clinical trials. Additionally, sensitivity studies were performed to simulate the effects of patient size and image reconstruction method on the quantitative performance of these metrics. Complex trends in variability were revealed as a function of textural feature, lesion size, patient size, and reconstruction parameters. In conclusion, the sensitivity of PET textural features to normal stochastic image variation and imaging parameters can be large and is feature-dependent. Standards are needed to ensure that prospective studies that incorporate textural features are properly designed to measure true effects that may impact clinical outcomes.

Correlation of PET images of metabolism, proliferation and hypoxia to characterize tumor phenotype in patients with cancer of the oropharynx

UNLABELLED Spatial organization of tumor phenotype is of great interest to radiotherapy target definition and outcome prediction. We characterized tumor phenotype in patients with cancers of the oropharynx through voxel-based correlation of PET images of metabolism, proliferation, and hypoxia. METHODS Patients with oropharyngeal cancer received (18)F-fluorodeoxyglucose (FDG) PET/CT, (18)F-fluorothymidine (FLT) PET/CT, and (61)Cu-diacetyl-bis(N4-methylthiosemicarbazone) (Cu-ATSM) PET/CT. Images were co-registered and standardized uptake values (SUV) were calculated for all modalities. Voxel-based correlation was evaluated with Pearsons correlation coefficient in tumor regions. Additionally, sensitivity studies were performed to quantify the effects of image segmentation, registration, noise, and segmentation on R. RESULTS On average, FDG PET and FLT PET images were most highly correlated (R(FDG:FLT) = 0.76, range 0.53-0.85), while Cu-ATSM PET showed greater heterogeneity in correlation to other tracers (R(FDG:Cu-ATSM) = 0.64, range 0.51-0.79; R(FLT:Cu-ATSM) = 0.61, range 0.21-0.80). Of the tested parameters, correlation was most sensitive to image registration. Misregistration of one voxel lead to ΔR(FDG) = 0.25, ΔR(FLT) = 0.39, and ΔR(Cu-ATSM) = 0.27. Image noise and reconstruction also had quantitative effects on correlation. No significant quantitative differences were found between GTV, expanded GTV, or CTV regions. CONCLUSIONS Voxel-based correlation represents a first step into understanding spatial organization of tumor phenotype. These results have implications for radiotherapy target definition and provide a framework to test outcome prediction based on pretherapy distribution of phenotype.

Metrics of success: Measuring impact of a departmental near-miss incident learning system

PURPOSE There is a growing interest in the application of incident learning systems (ILS) to radiation oncology. The purpose of the present study is to define statistical metrics that may serve as benchmarks for successful operation of an incident learning system. METHODS AND MATERIALS A departmental safety and quality ILS was developed to monitor errors, near-miss events, and process improvement suggestions. Event reports were reviewed by a multiprofessional quality improvement committee. Events were scored by a near-miss risk index (NMRI) and categorized by event point of origination and discovery. Reporting trends were analyzed over a 2-year period, including total number and rates of events reported, users reporting, NMRI, and event origination and discovery. RESULTS A total of 1897 reports were evaluated (1.0 reports/patient, 0.9 reports/unique treatment course). Participation in the ILS increased as demonstrated by total events (2.1 additional reports/month) and unique users (0.5 new users/month). Sixteen percent of reports had an NMRI of 0 (none), 42% had an NMRI of 1 (mild), 25% had an NMRI of 2 (moderate), 12% had an NMRI of 3 (severe), and 5% had an NMRI of 4 (critical). Event NMRI showed a significant decrease in the first 6 months (1.68-1.42, P < .001). Trends in origination and discovery of reports were broadly distributed between radiation therapy process steps and staff groups. The highest risk events originated in imaging for treatment planning (NMRI = 2.0 ± 1.1; P < .0001) and were detected in on-treatment quality management (NMRI = 1.7 ± 1.1; P = .003). CONCLUSIONS Over the initial 2-year period of ILS operation, rates of reporting increased, staff participation increased, and NMRI of reported events declined. These data mirror previously reported findings of improvement in safety culture endpoints. These metrics may be useful for other institutions seeking to create or evaluate their own ILS.

Measurable improvement in patient safety culture: A departmental experience with incident learning

PURPOSE Rigorous use of departmental incident learning is integral to improving patient safety and quality of care. The goal of this study was to quantify the impact of a high-volume, departmental incident learning system on patient safety culture. METHODS AND MATERIALS A prospective, voluntary, electronic incident learning system was implemented in February 2012 with the intent of tracking near-miss/no-harm incidents. All incident reports were reviewed weekly by a multiprofessional team with regular department-wide feedback. Patient safety culture was measured at baseline with validated patient safety culture survey questions. A repeat survey was conducted after 1 and 2 years of departmental incident learning. Proportional changes were compared by χ(2) or Fisher exact test, where appropriate. RESULTS Between 2012 and 2014, a total of 1897 error/near-miss incidents were reported, representing an average of 1 near-miss report per patient treated. Reports were filed by a cross section of staff, with the majority of incidents reported by therapists, dosimetrists, and physicists. Survey response rates at baseline and 1 and 2 years were 78%, 80%, and 80%, respectively. Statistically significant and sustained improvements were noted in several safety metrics, including belief that the department was openly discussing ways to improve safety, the sense that reports were being used for safety improvement, and the sense that changes were being evaluated for effectiveness. None of the surveyed dimensions of patient safety culture worsened. Fewer punitive concerns were noted, with statistically significant decreases in the worry of embarrassment in front of colleagues and fear of getting colleagues in trouble. CONCLUSIONS A comprehensive incident learning system can identify many areas for improvement and is associated with significant and sustained improvements in patient safety culture. These data provide valuable guidance as incident learning systems become more widely used in radiation oncology.

Differential hepatic avoidance radiation therapy: Proof of concept in hepatocellular carcinoma patients

PURPOSE To evaluate the feasibility of a novel planning concept that differentially redistributes RT dose away from functional liver regions as defined by (99m)Tc-sulphur colloid (SC) uptake on patient SPECT/CT images. MATERIALS AND METHODS Ten HCC patients with different Child-Turcotte-Pugh scores (A5-B9) underwent SC SPECT/CT scans in treatment position prior to RT that were registered to planning CT scans. Proton pencil beam scanning (PBS) therapy plans were optimized to deliver 37.5-60.0Gy (RBE) over 5-15 fractions using single field uniform dose technique robust to range and setup uncertainty. Photon volumetrically modulated arc therapy (VMAT) plans were optimized to the same prescribed dose and minimum target coverage. For both treatment modalities, differential hepatic avoidance RT (DHART) plans were generated to decrease dose to functional liver volumes (FLV) defined by a range of thresholds relative to maximum SC uptake (43-90%) in the tumor-subtracted liver. Radiation dose was redistributed away from regions of increased SC uptake in each FLV by linearly scaling mean dose objectives during PBS or VMAT optimization. DHART planning feasibility was assessed by a significantly negative Spearmans rank correlation (RS) between dose difference and SC uptake. Patient, tumor, and treatment planning characteristics were tested for association to DHART planning feasibility using non-parametric Kruskal-Wallis ANOVA. RESULTS Compared to conventional plans, DHART plans achieved a 3% FLV dose reduction for every 10% SC uptake increase. DHART planning was feasible in the majority of patients with 60% of patients having RS<-0.5 (p<0.01, range -1.0 to 0.2) and was particularly effective in 30% of patients (RS<-0.9). Mean dose to FLV was reduced by up to 20% in these patients. Only fractionation regimen was associated with DHART planning feasibility: 15 fraction courses were more feasible than 5-6 fraction courses (RS<-0.93 vs. RS>-0.60, p<0.02). CONCLUSION Differential avoidance of functional liver regions defined on sulphur colloid SPECT/CT is achievable with either photon VMAT or proton PBS therapy. Further investigation with phantom studies and in a larger cohort of patients may validate the utility of DHART planning for HCC radiotherapy.

Impact of CT attenuation correction method on quantitative respiratory‐correlated (4D) PET/CT imaging

PURPOSE Respiratory-correlated positron emission tomography (PET/CT) 4D PET/CT is used to mitigate errors from respiratory motion; however, the optimal CT attenuation correction (CTAC) method for 4D PET/CT is unknown. The authors performed a phantom study to evaluate the quantitative performance of CTAC methods for 4D PET/CT in the ground truth setting. METHODS A programmable respiratory motion phantom with a custom movable insert designed to emulate a lung lesion and lung tissue was used for this study. The insert was driven by one of five waveforms: two sinusoidal waveforms or three patient-specific respiratory waveforms. 3DPET and 4DPET images of the phantom under motion were acquired and reconstructed with six CTAC methods: helical breath-hold (3DHEL), helical free-breathing (3DMOT), 4D phase-averaged (4DAVG), 4D maximum intensity projection (4DMIP), 4D phase-matched (4DMATCH), and 4D end-exhale (4DEXH) CTAC. Recovery of SUV(max), SUV(mean), SUV(peak), and segmented tumor volume was evaluated as RC(max), RC(mean), RC(peak), and RC(vol), representing percent difference relative to the static ground truth case. Paired Wilcoxon tests and Kruskal-Wallis ANOVA were used to test for significant differences. RESULTS For 4DPET imaging, the maximum intensity projection CTAC produced significantly more accurate recovery coefficients than all other CTAC methods (p < 0.0001 over all metrics). Over all motion waveforms, ratios of 4DMIP CTAC recovery were 0.2 ± 5.4, -1.8 ± 6.5, -3.2 ± 5.0, and 3.0 ± 5.9 for RC(max), RC(peak), RC(mean), and RC(vol). In comparison, recovery coefficients for phase-matched CTAC were -8.4 ± 5.3, -10.5 ± 6.2, -7.6 ± 5.0, and -13.0 ± 7.7 for RC(max), RC(peak), RC(mean), and RC(vol). When testing differences between phases over all CTAC methods and waveforms, end-exhale phases were significantly more accurate (p = 0.005). However, these differences were driven by the patient-specific respiratory waveforms; when testing patient and sinusoidal waveforms separately, patient waveforms were significantly different between phases (p < 0.0001) while the sinusoidal waveforms were not significantly different (p = 0.98). When considering only the subset of 4DMATCH images that corresponded to the end-exhale image phase, 4DEXH, mean and interquartile range were similar to 4DMATCH but variability was considerably reduced. CONCLUSIONS Comparative advantages in accuracy and precision of SUV metrics and segmented volumes were demonstrated with the use of the maximum intensity projection and end-exhale CT attenuation correction. While respiratory phase-matched CTAC should in theory provide optimal corrections, image artifacts and differences in implementation of 4DCT and 4DPET sorting can degrade the benefit of this approach. These results may be useful to guide the implementation, analysis, and development of respiratory-correlated thoracic PET/CT in the radiation oncology and diagnostic settings.

Improved accuracy for noncoplanar radiotherapy: an EPID-based method for submillimeter alignment of linear accelerator table rotation with MV isocenter

Accurate alignment of linear accelerator table rotational axis with radiation isocenter is critical for noncoplanar radiotherapy applications. The purpose of the present study is to develop a method to align the table rotation axis and the MV isocenter to submillimeter accuracy. We developed a computerized method using electronic portal imaging device (EPID) and measured alignment stability over time. Mechanical and radiation isocenter coincidence was measured by placing a steel ball bearing at radiation isocenter using existing EPID techniques. Then, EPID images were acquired over the range of table rotation. A MATLAB script was developed to calculate the center of rotation, as well as the necessary adjustment to move the table rotational axis to MV isocenter. Adjustment was applied via torque to screws at the base of the linac table. Stability of rotational alignment was measured with 49 measurements over 363 days on four linacs. Initial rotational misalignment from radiation isocenter ranged from 0.91−2.11 mm on the four tested linacs. Linac‐A had greatest error (>2 mm) and was adjusted with the described method. After adjustment, the error was significantly decreased to 0.40±0.12 mm. The adjustment was stable over the course of 15 measurements over 231 days. Linac‐B was not adjusted, but tracked from time of commissioning with 27 measurements over 363 days. No discernible shift in couch characteristics was observed (mean error 1.40±0.22 mm). The greater variability for Linac‐B may relate to the interchangeable two‐piece couch, which allows more lateral movement than the one‐piece Linac‐A couch. Submillimeter isocenter alignment was achieved by applying a precision correction to the linac table base. Table rotational characteristics were shown to be stable over the course of twelve months. The accuracy and efficiency of this method may make it suitable for acceptance testing, annual quality assurance, or commissioning of highly‐conformal noncoplanar radiotherapy programs. PACS number: 87

Interrater reliability of a near-miss risk index for incident learning systems in radiation oncology

PURPOSE Tools for assessing the severity and risk of near-miss events in radiation oncology are few and needed. Recent work has described guidelines for the use of a 5-tier near-miss risk index (NMRI) for the classification of near-miss events. The purpose of this study was to assess the reliability of the NMRI among users in a radiation oncology department. METHODS AND MATERIALS Reliability of the NMRI was assessed using an online survey distributed to members of a radiation oncology department. The survey contained 70 events extracted from the departments incident learning system (ILS). Survey participants rated each event using the NMRI guidelines, reported their attendance to weekly ILS meetings (used as a surrogate for familiarity with the ILS), and indicated their familiarity with the radiation oncology workflow. Interrater reliability was determined using Krippendorffs alpha. Use of the NMRI to rate actual events during 5 weekly ILS meetings was also assessed and interrater reliability determined. RESULTS Twenty-eight survey respondents represented a wide variety of care providers. Krippendorffs alpha was calculated for the whole respondent cohort to be 0.376, indicating fair agreement among raters. Respondents who had the most participation at ILS meetings (n = 4) had moderate agreement with an alpha of 0.501. Interestingly, there were significant differences in reliability and median NMRI scores between professions. NMRI use during weekly NMRI meetings (80 events rated), participants showed moderate reliability (alpha = 0.607). CONCLUSIONS Using the NMRI guidelines, raters from a wide variety of professions were able to assess the severity of near-miss incidents with fair agreement. Those experienced with the ILS showed better agreement, and higher agreement was seen during multidisciplinary ILS meetings. These data support the use the indices such as the NMRI for near-miss risk assessment in patient safety and prioritization of process improvements in radiation oncology.

Can emergent treatments result in more severe errors?: An analysis of a large institutional near-miss incident reporting database.

PURPOSE Emergent radiation treatments may be subject to more errors because of the compressed time frame. Few data exist on the magnitude of this problem or how to guide safety improvement interventions. The purpose of this study is to examine patterns of near-miss events in emergent treatments using a large institutional incident reporting system. METHODS AND MATERIALS Events in the incident reporting database from February 2012 to October 2013 were reviewed prospectively by a multidisciplinary team to identify emergent treatments. Reports were scored for potential near-miss risk index (NMRI) on a 0 to 4 scale. Workflow steps of where events originated and were detected were analyzed. Events were categorized by use of the causal factor system from the Radiation Oncology Incident Learning System. Mann-Whitney U tests were used to compare mean NMRI score, and Fisher exact tests were performed to compare the proportion of high-risk events between emergent and nonemergent treatments and between emergent treatments on weekdays and weekends or holidays. RESULTS Over the study period, approximately 1600 patients were treated, 190 of them emergently. Seventy-one incident reports were submitted for 55 unique patients. Fewer events were reported for emergent treatments than for nonemergent treatments (0.37 events per new treatment vs 0.86; P < .01). Mean risk index for emergent reports was 1.90 versus 1.48 for nonemergent reports (P < .01). Rate of NMRI 4 was 10% for emergent treatments versus 4% for nonemergent treatments (P < .01). Emergent treatments started on a weekend or holiday had a higher proportion of critical near-miss events than emergent treatments started during the week (37% vs 7.9%, P = .034). CONCLUSIONS In this study, fewer near-miss incidents were reported per treatment course for emergent treatments. This may be attributable to reporting bias. More importantly, when emergent near misses occur, they are of greater severity.

Assessment of functional liver reserve: old and new in 99mTc-sulfur colloid scintigraphy

Purpose A semiquantitative assessment of hepatic reticuloendothelial system function using colloidal particles scintigraphy has been proposed previously as a surrogate for liver function evaluation. In this article, we present an updated method for the overall assessment of technetium-99m (99mTc)-sulfur colloid (SC) biodistribution that combines information from planar and attenuation-corrected 99mTc-SC single-photon emission computed tomography (SPECT) images. The imaging protocol described here was developed as an easy-to-implement method to assess overall and regional liver function changes associated with chronic liver disease. Patients and methods Thirty patients with chronic liver disease and primary liver cancers underwent 99mTc-SC whole-body planar imaging and upper-abdomen SPECT/computed tomography (CT) imaging before external beam radiation therapy. Liver plus spleen and bone marrow counts as a fraction of whole-body total counts were calculated from SC planar imaging. Attenuation correction 99mTc-SC images were rigidly coregistered with treatment planning CT images that contained liver and spleen regions-of-interest. Ratios of total liver counts to total spleen counts were obtained from the aligned 99mTc-SC SPECT and CT images, and were subsequently used to separate liver plus spleen counts obtained on the planar images. This hybrid SPECT/CT and planar scintigraphy approach yielded an updated estimation of whole-body SC distribution. These biodistribution estimates were compared with historical data for reference. Statistical associations of 99mTc-SC biodistribution to liver function parameters and liver disease scoring systems (Child–Pugh) were evaluated by Spearman rank correlation. Results Percentages of 99mTc-SC uptake ranged from 19.3 to 77.3% for the liver; 3.4 to 40.7% for the spleen; and 19.0 to 56.7% for the bone marrow. Spearman’s correlation coefficient showed a significant statistical association between Child–Pugh score and bone marrow uptake at 0.55 (P⩽0.05), liver uptake at 0.71 (P⩽0.001), spleen uptake at 0.56 (P⩽0.05), and spleen plus bone marrow uptake at 0.71 (P⩽0.001). There was also a good correlation of SC uptake percentages with individual quantitative liver function components such as albumin and total bilirubin, and qualitative liver function components (varices, portal hypertension, ascites). For albumin: r=0.64 (P<0.001) compared with liver uptake percentage from the whole-body counts, r=0.49 (P<0.001) compared with splenic uptake percentage, and r=0.45 (P⩽0.05) compared with bone marrow uptake percentage. Conclusion We describe a novel liver function quantitative assessment method that combines whole-body planar images and SPECT/CT attenuation-corrected images of 99mTc-SC distribution. Attenuation-corrected SC images provide valuable regional liver function information, which is a unique feature compared with other imaging methods available. The results of our study indicate that the 99mTc-SC uptake by the liver, spleen, and bone marrow correlates with liver function parameters in patients with diffuse liver disease and the correlation with liver disease severity is slightly better for liver uptake percentages than for individual values of bone marrow and spleen uptake percentages.