Georgi Nalbantov

Predicting outcomes in radiation oncology —multifactorial decision support systems

With the emergence of individualized medicine and the increasing amount and complexity of available medical data, a growing need exists for the development of clinical decision-support systems based on prediction models of treatment outcome. In radiation oncology, these models combine both predictive and prognostic data factors from clinical, imaging, molecular and other sources to achieve the highest accuracy to predict tumour response and follow-up event rates. In this Review, we provide an overview of the factors that are correlated with outcome—including survival, recurrence patterns and toxicity—in radiation oncology and discuss the methodology behind the development of prediction models, which is a multistage process. Even after initial development and clinical introduction, a truly useful predictive model will be continuously re-evaluated on different patient datasets from different regions to ensure its population-specific strength. In the future, validated decision-support systems will be fully integrated in the clinic, with data and knowledge being shared in a standardized, instant and global manner.

‘Rapid Learning health care in oncology’ – An approach towards decision support systems enabling customised radiotherapy’ ☆ ☆☆

PURPOSE An overview of the Rapid Learning methodology, its results, and the potential impact on radiotherapy. MATERIAL AND RESULTS Rapid Learning methodology is divided into four phases. In the data phase, diverse data are collected about past patients, treatments used, and outcomes. Innovative information technologies that support semantic interoperability enable distributed learning and data sharing without additional burden on health care professionals and without the need for data to leave the hospital. In the knowledge phase, prediction models are developed for new data and treatment outcomes by applying machine learning methods to data. In the application phase, this knowledge is applied in clinical practice via novel decision support systems or via extensions of existing models such as Tumour Control Probability models. In the evaluation phase, the predictability of treatment outcomes allows the new knowledge to be evaluated by comparing predicted and actual outcomes. CONCLUSION Personalised or tailored cancer therapy ensures not only that patients receive an optimal treatment, but also that the right resources are being used for the right patients. Rapid Learning approaches combined with evidence based medicine are expected to improve the predictability of outcome and radiotherapy is the ideal field to study the value of Rapid Learning. The next step will be to include patient preferences in the decision making.

The effect of SUV discretization in quantitative FDG-PET Radiomics: the need for standardized methodology in tumor texture analysis

FDG-PET-derived textural features describing intra-tumor heterogeneity are increasingly investigated as imaging biomarkers. As part of the process of quantifying heterogeneity, image intensities (SUVs) are typically resampled into a reduced number of discrete bins. We focused on the implications of the manner in which this discretization is implemented. Two methods were evaluated: (1) RD, dividing the SUV range into D equally spaced bins, where the intensity resolution (i.e. bin size) varies per image; and (2) RB, maintaining a constant intensity resolution B. Clinical feasibility was assessed on 35 lung cancer patients, imaged before and in the second week of radiotherapy. Forty-four textural features were determined for different D and B for both imaging time points. Feature values depended on the intensity resolution and out of both assessed methods, RB was shown to allow for a meaningful inter- and intra-patient comparison of feature values. Overall, patients ranked differently according to feature values–which was used as a surrogate for textural feature interpretation–between both discretization methods. Our study shows that the manner of SUV discretization has a crucial effect on the resulting textural features and the interpretation thereof, emphasizing the importance of standardized methodology in tumor texture analysis.

Cardiac comorbidity is an independent risk factor for radiation-induced lung toxicity in lung cancer patients

PURPOSE To test the hypothesis that cardiac comorbidity before the start of radiotherapy (RT) is associated with an increased risk of radiation-induced lung toxicity (RILT) in lung cancer patients. MATERIAL AND METHODS A retrospective analysis was performed of a prospective cohort of 259 patients with locoregional lung cancer treated with definitive radio(chemo)therapy between 2007 and 2011 (ClinicalTrials.gov Identifiers: NCT00572325 and NCT00573040). We defined RILT as dyspnea CTCv.3.0 grade ≥2 within 6 months after RT, and cardiac comorbidity as a recorded treatment of a cardiac pathology at a cardiology department. Univariate and multivariate analyses, as well as external validation, were performed. The model-performance measure was the area under the receiver operating characteristic curve (AUC). RESULTS Prior to RT, 75/259 (28.9%) patients had cardiac comorbidity, 44% of whom (33/75) developed RILT. The odds ratio of developing RILT for patients with cardiac comorbidity was 2.58 (p<0.01). The cross-validated AUC of a model with cardiac comorbidity, tumor location, forced expiratory volume in 1s, sequential chemotherapy and pretreatment dyspnea score was 0.72 (p<0.001) on the training set, and 0.67 (p<0.001) on the validation set. CONCLUSION Cardiac comorbidity is an important risk factor for developing RILT after definite radio(chemo)therapy of lung cancer patients.

Intravascular enhancement with identical iodine delivery rate using different iodine contrast media in a circulation phantom.

PurposeBoth iodine delivery rate (IDR) and iodine concentration are decisive factors for vascular enhancement in computed tomographic angiography. It is unclear, however, whether the use of high–iodine concentration contrast media is beneficial to lower iodine concentrations when IDR is kept identical. This study evaluates the effect of using different iodine concentrations on intravascular attenuation in a circulation phantom while maintaining a constant IDR. Materials and MethodsA circulation phantom with a low-pressure venous compartment and a high-pressure arterial compartment simulating physiological circulation parameters was used (heart rate, 60 beats per minute; stroke volume, 60 mL; blood pressure, 120/80 mm Hg). Maintaining a constant IDR (2.0 g/s) and a constant total iodine load (20 g), prewarmed (37°C) contrast media with differing iodine concentrations (240–400 mg/mL) were injected into the phantom using a double-headed power injector. Serial computed tomographic scans at the level of the ascending aorta (AA), the descending aorta (DA), and the left main coronary artery (LM) were obtained. Total amount of contrast volume (milliliters), iodine delivery (grams of iodine), peak flow rate (milliliter per second), and intravascular pressure (pounds per square inch) were monitored using a dedicated data acquisition program. Attenuation values in the AA, the DA, and the LM were constantly measured (Hounsfield unit [HU]). In addition, time-enhancement curves, aortic peak enhancement, and time to peak were determined. ResultsAll contrast injection protocols resulted in similar attenuation values: the AA (516 [11] to 531 [37] HU), the DA (514 [17] to 531 [32] HU), and the LM (490 [10] to 507 [17] HU). No significant differences were found between the AA, the DA, and the LM for either peak enhancement (all P > 0.05) or mean time to peak (AA, 19.4 [0.58] to 20.1 [1.05] seconds; DA, 21.1 [1.0] to 21.4 [1.15] seconds; LM, 19.8 [0.58] to 20.1 [1.05] seconds). ConclusionsThis phantom study demonstrates that constant injection parameters (IDR, overall iodine load) lead to robust enhancement patterns, regardless of the contrast material used. Higher iodine concentration itself does not lead to higher attenuation levels. These results may stimulate a shift in paradigm toward clinical usage of contrast media with lower iodine concentrations (eg, 240 mg iodine/mL) in individual tailored contrast protocols. The use of low–iodine concentration contrast media is desirable because of the lower viscosity and the resulting lower injection pressure.

Infrastructure and distributed learning methodology for privacy-preserving multi-centric rapid learning health care: euroCAT

Graphical abstract

Preclinical assessment of efficacy of radiation dose painting based on intratumoral FDG-PET uptake

Purpose: We tested therapeutic efficacy of two dose painting strategies of applying higher radiation dose to tumor subvolumes with high FDG uptake (biologic target volume, BTV): dose escalation and dose redistribution. We also investigated whether tumor response was determined by the highest dose in BTV or the lowest dose in gross tumor volume (GTV). Experimental Design: FDG uptake was evaluated in rat rhabdomyosarcomas prior to irradiation. BTV was defined as 30% of GTV with the highest (BTVhot) or lowest (BTVcold) uptake. To test efficacy of dose escalation, tumor response (time to reach two times starting tumor volume, TGTV2) to Hot Boost irradiation (40% higher dose to BTVhot) was compared with Cold Boost (40% higher dose to BTVcold), while mean dose to GTV remained 12 Gy. To test efficacy of dose redistribution, TGTV2 after Hot Boost was compared with uniform irradiation with the same mean dose (8 or 12 Gy). Results: TGTV2 after 12 Gy delivered heterogeneously (Hot and Cold Boost) or uniformly were not significantly different: 20.2, 19.5, and 20.6 days, respectively. Dose redistribution (Hot Boost) with 8 Gy resulted in faster tumor regrowth as compared with uniform irradiation (13.3 vs. 17.1 days; P = 0.026). Further increase in dose gradient to 60% led to a more pronounced decrease in TGTV2 (10.9 days; P < 0.0001). Conclusions: Dose escalation effect was independent of FDG uptake in target tumor volume, while dose redistribution was detrimental in this tumor model for dose levels applied here. Our data are consistent with the hypothesis that tumor response depends on the minimum intratumoral dose. Clin Cancer Res; 21(24); 5511–8. ©2015 AACR.

Nonlinear Support Vector Machines Through Iterative Majorization and I-Splines

To minimize the primal support vector machine (SVM) problem, we propose to use iterative majorization. To allow for nonlinearity of the predictors, we use (non)monotone spline transformations. An advantage over the usual kernel approach in the dual problem is that the variables can be easily interpreted. We illustrate this with an example from the literature.

Equity style timing using support vector regressions

The disappointing performance of value and small cap strategies shows that style consistency may not provide the long-term benefits often assumed in the literature. In this study we examine whether the short-term variation in the U.S. size and value premium is predictable. We document style-timing strategies based on technical and (macro-)economic predictors using a recently developed artificial intelligence tool called Support Vector Regressions (SVR). SVR are known for their ability to tackle the standard problem of overfitting, especially in multivariate settings. Our findings indicate that both premiums are predictable under fair levels of transaction costs and various forecasting horizons.

Quantification of CT-assessed radiation-induced lung damage in lung cancer patients treated with or without chemotherapy and cetuximab

Abstract Background and Purpose: Prediction models for radiation-induced lung damage (RILD) are still unsatisfactory, with clinical toxicity endpoints that are difficult to quantify objectively. We therefore evaluated RILD more objectively, quantitatively and on a continuous scale measuring the lung tissue density changes per voxel. Material and methods: Patients treated with radiotherapy (RT) alone, sequential and concurrent chemo-RT with and without the addition of cetuximab were studied. Follow-up computed tomography (CT) scans were co-registered using deformable registration to baseline CT scans. CT density changes were correlated to the RT dose delivered in every part of the lungs. Results: One hundred and seventeen lung cancer patients were included. Mean dose to tumor was 60 Gy (range 45–79.2 Gy). Dose response curves showed a linear increase in the dose region between 0 and 65 Gy having a slope (based on coefficients of the multilevel model) expressed as a lung density increase per dose of 0.86 (95% CI 0.73–0.99), 1.31 (95% CI 1.19–1.43), 1.39 (95% CI 1.28–1.50) and 2.07 (95% CI 1.93–2.21) for patients treated only with RT (N=19), sequential chemo-RT (N=30), concurrent chemo-RT (N=49), and concurrent chemo-RT with cetuximab (N=19), respectively. Conclusions: CT density changes allow quantitative assessment of lung damage after fractionated RT, giving complementary information to standard used clinical endpoints. Patients receiving cetuximab showed a significantly larger dose response compared with other treatments.