Image-based Data Mining Research Articles

Development of age-specific population-based paediatric computational phantoms for image-based data mining and other radiotherapy applications

Abstract Objective. Computational anatomical models have many applications in paediatric radiotherapy. Age-specific computational anatomical models were historically developed to represent average and/or healthy individuals, where cancer patients may present with anatomical variations caused by the disease and/or treatment effects. We developed RT-PAL, a library of computational age-specific voxelized anatomical models tailored to represent the paediatric radiotherapy population. Approach. Data from patients undergoing craniospinal irradiation (CSI) were used (n = 74, median age 7.3y, range: 1–17y). The RT-PAL phantoms were generated using groupwise deformable image registration to spatially normalize and average a sub-set of twenty clinical CTs and contours (n = 74, median age 7.7y, range: 3–14 y). To assess their anatomical and age-dependency plausibility, the RT-PAL models were compared against clinical cancer patient data and two healthy population based libraries of phantoms: the International Commission on Radiological Protection (ICRP) pediatric reference computational phantoms (n = 8, median age 7.5y, range: 1–15y) and a range of 4D paediatric extended cardiac torso (XCAT) phantoms (n = 75, median age 9.1y, range: 1–18y). For each dataset, nineteen organs were segmented on all age models to determine their volume. Each set was evaluated through a linear fit of organ volume with age, where comparisons were made relative to the linear fit of the clinical data. Main Results. Overall good anatomical plausibility was found for the RT-PAL phantoms. The age-dependency reported was comparable to both the clinical data and other phantoms, demonstrating their efficacy as a library of age-specific phantoms. Larger discrepancies with the clinical, ICRP and XCAT organ data were attributable to differences in organ filling, segmentation strategy and age distribution of the datasets, limitations of RT-PAL generation methodology, and/or possible anatomical differences between healthy and cancer populations. Significance. The RT-PAL models showed potential in representing the paediatric radiotherapy cohort, who are most likely to benefit from dedicated, age-specific anatomical phantoms.

CT- and MR-based image-based data mining are consistent in the brain

PurposeImage-based data mining (IBDM) is a voxel-based analysis technique to investigate dose–response. Most often, IBDM uses radiotherapy planning CTs because of their broad accessibility, however, it was unknown whether CT provided sufficient soft tissue contrast for brain IBDM. This study evaluates whether MR-based IBDM improves upon CT-based IBDM for studies of children with brain tumours. MethodsWe compared IBDM pipelines using either CT- or MRI-based spatial normalisation in 128 children (ages 3.3–19.7 years) who received photon radiotherapy for primary brain tumours at a single institution. We quantified spatial-normalisation accuracy using contour comparison measures (centre-of-mass separation, average contour distance-to-agreement (DTavg), and Hausdorff distance) at multiple anatomic loci. We performed an end-to-end test of CT- and MRI-IBDM using modified clinical dose distributions and simulated effect labels to detect associations in pre-defined anatomic loci. Accuracy was assessed via sensitivity and specificity. ResultsSpatial normalisation accuracy was comparable for both modalities, with a significant but small improvement for MRI compared to CT in all structures except the brainstem. The median (range) difference between the DTavg for the two pipelines was 0.37 (0.00–2.91) mm. The end-to-end test revealed no significant difference in sensitivity of the IBDM-identified regions for the two pipelines. Specificity slightly improved for MR-IBDM at the 99% significance level. ConclusionCT-based IBDM was comparable to MR-based IBDM, despite a small advantage in spatial normalisation accuracy with MRI. The use of CT-IBDM over MR-IBDM is useful for multi-institutional retrospective IBDM studies, where the availability of standardised MRI data can be limited.

RADT-03. IDENTIFYING BRAIN STRUCTURES LINKED TO RADIATION-INDUCED ENDOCRINE AND COGNITIVE SEQUELAE: A MULTI-CENTRE VOXEL-BASED ANALYSIS

Abstract BACKGROUND Radiotherapy (RT) is a common treatment for paediatric brain tumours. Despite advancements in techniques to mitigate RT-induced toxicity, almost two-thirds of patients develop long-lasting endocrine and cognitive sequalae. Image-based data mining (IBDM) is a voxel-based analysis tool that can identify dose-sensitive regions, supporting development of tailored radiation treatments for future patients that avoids these regions. We present the first IBDM study localising endocrine and cognitive effects in paediatric brain tumour patients. METHODS Imaging (MRI and CT) and clinical data were available for 261 children (146 medulloblastoma patients in centre A, 115 patients with mixed diagnoses in centre B). Outcomes were reported as Growth Hormone Deficiency (GHD, presence/absence, both centres) and processing speed (PS, continuous slope derived from a linear mixed-effects model, centre A). IBDM evaluated where dose was associated with outcome (p-value<0.01). Mean dose to this region was included in multivariable analysis adjusting for age, risk status, hydrocephalus status/management, and chemotherapy. RESULTS GHD was associated with dose to a region overlapping the pituitary and brainstem. In centre A, multivariable analysis revealed a 13(2-27)% average increase in GHD odds per 1Gy increase in dose, with an inverse effect for age. Center B showed a 8(4-13)% average increase for every 1Gy, with no other predictors. For PS, a central frontal region from the medial frontal lobes to the hypothalamus was identified (139 patients in centre A). Every 1Gy increase led to a -0.08(-0.20 – 0.04) PS change, with age and hydrocephalus management as contributing factors. CONCLUSION We identified a region (encompassing the pituitary and brainstem) associated with increased risk of GHD, that was consistent in two independent cohort analyses. Furthermore, PS decline was linked to dose to a frontal region. Minimising dose to these regions for future patients could reduce the risk of cognitive and endocrine sequelae.

Dysphagia at 1 Year is Associated With Mean Dose to the Inferior Section of the Brain Stem

Dysphagia is a common toxicity after head and neck (HN) radiation therapy that negatively affects quality of life. We explored the relationship between radiation therapy dose to normal HN structures and dysphagia 1 year after treatment using image-based datamining (IBDM), a voxel-based analysis technique. We used data from 104 patients with oropharyngeal cancer treated with definitive (chemo)radiation therapy. Swallow function was assessed pretreatment and 1 year posttreatment using 3 validated measures: MD Anderson Dysphagia Inventory (MDADI), performance status scale for normalcy of diet (PSS-HN), and water swallowing test (WST). For IBDM, we spatially normalized all patients' planning dose matrices to 3 reference anatomies. Regions where the dose was associated with dysphagia measures at 1 year were found by performing voxel-wise statistics and permutation testing. Clinical factors, treatment variables, and pretreatment measures were used in multivariable analysis to predict each dysphagia measure at 1 year. Clinical baseline models were found using backward stepwise selection. Improvement in model discrimination after adding the mean dose to the identified region was quantified using the Akaike information criterion. We also compared the prediction performance of the identified region with a well-established association: mean doses to the pharyngeal constrictor muscles. IBDM revealed highly significant associations between dose to distinct regions and the 3 outcomes. These regions overlapped around the inferior section of the brain stem. All clinical models were significantly improved by including mean dose to the overlap region (P ≤ .006). Including pharyngeal dosimetry significantly improved WST (P=.04) but not PSS-HN or MDADI (P ≥ .06). In this hypothesis-generating study, we found that mean dose to the inferior section of the brain stem is strongly associated with dysphagia 1 year posttreatment. The identified region includes the swallowing centers in the medulla oblongata, providing a possible mechanistic explanation. Further work including validation in an independent cohort is required.

Dose outside of the prostate is associated with improved outcomes for high-risk prostate cancer patients treated with brachytherapy boost.

One in three high-risk prostate cancer patients treated with radiotherapy recur. Detection of lymph node metastasis and microscopic disease spread using conventional imaging is poor, and many patients are under-treated due to suboptimal seminal vesicle or lymph node irradiation. We use Image Based Data Mining (IBDM) to investigate association between dose distributions, and prognostic variables and biochemical recurrence (BCR) in prostate cancer patients treated with radiotherapy. We further test whether including dose information in risk-stratification models improves performance. Planning CTs, dose distributions and clinical information were collected for 612 high-risk prostate cancer patients treated with conformal hypo-fractionated radiotherapy, intensity modulated radiotherapy (IMRT), or IMRT plus a single fraction high dose rate (HDR) brachytherapy boost. Dose distributions (including HDR boost) of all studied patients were mapped to a reference anatomy using the prostate delineations. Regions where dose distributions significantly differed between patients that did and did-not experience BCR were assessed voxel-wise using 1) a binary endpoint of BCR at four-years (dose only) and 2) Cox-IBDM (dose and prognostic variables). Regions where dose was associated with outcome were identified. Cox proportional-hazard models with and without region dose information were produced and the Akaike Information Criterion (AIC) was used to assess model performance. No significant regions were observed for patients treated with hypo-fractionated radiotherapy or IMRT. Regions outside the target where higher dose was associated with lower BCR were observed for patients treated with brachytherapy boost. Cox-IBDM revealed that dose response was influenced by age and T-stage. A region at the seminal vesicle tips was identified in binary- and Cox-IBDM. Including the mean dose in this region in a risk-stratification model (hazard ratio=0.84, p=0.005) significantly reduced AIC values (p=0.019), indicating superior performance, compared with prognostic variables only. The region dose was lower in the brachytherapy boost patients compared with the external beam cohorts supporting the occurrence of marginal misses. Association was identified between BCR and dose outside of the target region in high-risk prostate cancer patients treated with IMRT plus brachytherapy boost. We show, for the first-time, that the importance of irradiating this region is linked to prognostic variables.

OC-0776 Are MRI scans needed for precise image-based data mining in paediatric brain tumour cohorts?

OC-0776 Are MRI scans needed for precise image-based data mining in paediatric brain tumour cohorts?

Spatial normalisation for novel MR-image based data mining in children with brain tumours

<h3>Objectives</h3> Image-based data mining (IBDM) is a novel technique to identify dose-response relationships (DRRs) in radiotherapy. IBDM comprises two steps: spatial normalisation of patient data to a reference anatomy using deformable image registration, and ‘per-voxel' dose comparison. The main advantage of IBDM is the possibility of identifying DRRs at the sub-organ level and in structures that are not routinely contoured, depending on the image registration accuracy. Computed Tomography (CT)-based IBDM has been successfully applied in adults; however, for children with brain tumours, Magnetic Resonance (MR) imaging is indicated due to superior anatomical definition. We evaluate the accuracy of spatial normalisation using MR images to determine the feasibility of MR-based IBDM in children with brain tumours. <h3>Methods</h3> We used MR images and radiotherapy contours from 129 children previously treated on the SJMB03 protocol for medulloblastoma. The patient with the median brain volume was selected as the reference anatomy for deformable image registration (NiftyReg software). We analysed registration accuracy, comparing each patient's deformed contours to the reference contours, in 5 brain structures (Table 1) via 95%-Hausdorff (HD95%), average-contour (MeanDTA) and centre-of-mass (CoM) distances. Based on previous results in CT data, meanDTA accuracy <3mm was considered acceptable. <h3>Results</h3> 16/129 cases required manual adjustments to fix registration errors, commonly arising due to enlarged ventricles or large rotations compared to the reference anatomy. MeanDTA for all structures was <2.1mm and all metrics <6mm (Table 1). <h3>Conclusion</h3> Accurate spatial normalisation was achieved using MR images from paediatric brain tumour patients, suggesting that MR-based IBDM is feasible in this patient group.

Cardiac Function Modifies the Impact of Heart Base Dose on Survival: A Voxel-Wise Analysis of Patients With Lung Cancer From the PET-Plan Trial

IntroductionHeart dose has emerged as an independent predictor of overall survival in patients with NSCLC treated with radiotherapy. Several studies have identified the base of the heart as a region of enhanced dose sensitivity and a potential target for cardiac sparing. We present a dosimetric analysis of overall survival in the multicenter, randomized PET-Plan trial (NCT00697333) and for the first time include left ventricular ejection fraction (EF) at baseline as a metric of cardiac function. MethodsA total of 205 patients with inoperable stage II or III NSCLC treated with 60 to 72 Gy in 2 Gy fractions were included in this study. A voxel-wise image-based data mining methodology was used to identify anatomical regions where higher dose was significantly associated with worse overall survival. Univariable and multivariable Cox proportional hazards models tested the association of survival with dose to the identified region, established prognostic factors, and baseline cardiac function. ResultsA total of 172 patients remained after processing and censoring for follow-up. At 2-years posttreatment, a highly significant region was identified within the base of the heart (p < 0.005), centered on the origin of the left coronary artery and the region of the atrioventricular node. In multivariable analysis, the number of positron emission tomography–positive nodes (p = 0.02, hazard ratio = 1.13, 95% confidence interval: 1.02–1.25) and mean dose to the cardiac subregion (p = 0.02, hazard ratio = 1.11 Gy−1, 95% confidence interval: 1.02–1.21) were significantly associated with overall survival. There was a significant interaction between EF and region dose (p = 0.04) for survival, with contrast plots revealing a larger effect of region dose on survival in patients with lower EF values. ConclusionsThis work validates previous image-based data mining studies by revealing a strong association between dose to the base of the heart and overall survival. For the first time, an interaction between baseline cardiac health and heart base dose was identified, potentially suggesting preexisting cardiac dysfunction exacerbates the impact of heart dose on survival.

Image-based data mining applies to data collected from children

PurposeImage-based data mining (IBDM) is a novel voxel-based method for analyzing radiation dose responses that has been successfully applied in adult data. Because anatomic variability and side effects of interest differ for children compared to adults, we investigated the feasibility of IBDM for pediatric analyses. MethodsWe tested IBDM with CT images and dose distributions collected from 167 children (aged 10 months to 20 years) who received proton radiotherapy for primary brain tumors. We used data from four reference patients to assess IBDM sensitivity to reference selection. We quantified spatial-normalization accuracy via contour distances and deviations of the centers-of-mass of brain substructures. We performed dose comparisons with simplified and modified clinical dose distributions with a simulated effect, assessing their accuracy via sensitivity, positive predictive value (PPV) and Dice similarity coefficient (DSC). ResultsSpatial normalizations and dose comparisons were insensitive to reference selection. Normalization discrepancies were small (average contour distance < 2.5 mm, average center-of-mass deviation < 6 mm). Dose comparisons identified differences (p < 0.01) in 81% of simplified and all modified clinical dose distributions. The DSCs for simplified doses were high (peak frequency magnitudes of 0.9–1.0). However, the PPVs and DSCs were low (maximum 0.3 and 0.4, respectively) in the modified clinical tests. ConclusionsIBDM is feasible for childhood late-effects research. Our findings may inform cohort selection in future studies of pediatric radiotherapy dose responses and facilitate treatment planning to reduce treatment-related toxicities and improve quality of life among childhood cancer survivors.

PO-1780 Image-based data mining for radiation outcomes research applies to data from children

PO-1780 Image-based data mining for radiation outcomes research applies to data from children

Radial Data Mining to Identify Density-Dose Interactions That Predict Distant Failure Following SABR.

PurposeLower dose outside the planned treatment area in lung stereotactic radiotherapy has been linked to increased risk of distant metastasis (DM) possibly due to underdosage of microscopic disease (MDE). Independently, tumour density on pretreatment computed tomography (CT) has been linked to risk of MDE. No studies have investigated the interaction between imaging biomarkers and incidental dose. The interaction would showcase whether the impact of dose on outcome is dependent on imaging and, hence, if imaging could inform which patients require dose escalation outside the gross tumour volume (GTV). We propose an image-based data mining methodology to investigate density–dose interactions radially from the GTV to predict DM with no a priori assumption on location.MethodsDose and density were quantified in 1-mm annuli around the GTV for 199 patients with early-stage lung cancer treated with 60 Gy in 5 fractions. Each annulus was summarised by three density and three dose parameters. For parameter combinations, Cox regressions were performed including a dose–density interaction in independent annuli. Heatmaps were created that described improvement in DM prediction due to the interaction. Regions of significant improvement were identified and studied in overall outcome models.ResultsDose–density interactions were identified that significantly improved prediction for over 50% of bootstrap resamples. Dose and density parameters were not significant when the interaction was omitted. Tumour density variance and high peritumour density were associated with DM for patients with more cold spots (less than 30-Gy EQD2) and non-uniform dose about 3 cm outside of the GTV. Associations identified were independent of the mean GTV dose.ConclusionsPatients with high tumour variance and peritumour density have increased risk of DM if there is a low and non-uniform dose outside the GTV. The dose regions are independent of tumour dose, suggesting that incidental dose may play an important role in controlling occult disease. Understanding such interactions is key to identifying patients who will benefit from dose-escalation. The methodology presented allowed spatial dose–density interactions to be studied at the exploratory stage for the first time. This could accelerate the clinical implementation of imaging biomarkers by demonstrating the impact of incidental dose for tumours of varying characteristics in routine data.

Anatomical Association of Dose Distribution With Radiotherapy-Related Lymphopenia in Oropharynx Cancer

<h3>Purpose/Objective(s)</h3> Lymphopenia during radiotherapy (RT) may be related to adverse outcome for head and neck cancer and is a theoretical concern for effectiveness of adjuvant immunotherapy. Prevention may be possible via plan adaptation if the relationship between RT dose distribution and incidence of lymphopenia is known. This work utilized image-based data mining to identify anatomical regions associated with development of lymphopenia during RT. <h3>Materials/Methods</h3> 639 oropharyngeal cancer patients treated from 2005-2016 at a single institution with curative intent RT and baseline lymphocytes ≥ 0.5 × 10⁹/L were studied. To control for co-founders and improve data mining sensitivity, two matched groups were defined - patients with and without lymphopenia (defined as Grade 3 or higher based on lymphocytes at nadir during RT < 0.5 × 10⁹/L according to CTCAE v5.0) - matched on target volume, baseline lymphocytes, and prescribed dose (propensity score matching, M = 1:1, caliper = 0.2). Following matching, 309 patients remained, and image-based data mining was used to identify regions where dose correlates with lymphopenia. EQD2 dose matrices (α/β = 10) were mapped onto a reference patient using deformable registration of the planning CT images. L-R mirrored images were included in the cohort to compensate for potential laterality bias. Mean dose distributions for the groups with/without lymphopenia were calculated and regions with significant differences between the mean dose of the two groups detected using permutation testing with 1000 permutations. Multivariable logistic analysis was conducted for the full cohort including mean dose to the identified region, along with age, gender, smoking, performance status, HPV status, stage, histology, target volume, prescribed dose, RT modality, RT duration, baseline lymphocytes, and chemotherapy delivery to generate a predictive model for development of lymphopenia. <h3>Results</h3> 74% of the patients developed lymphopenia ≥ G3 during RT. An association with dose to a discrete anatomical region was found. Of interest, this location corresponds to dense lymphoid tissue, bone marrow reserve, and numerous blood vessels. Following elastic-net least absolute shrinkage and selection operator regression, bootstrapped 100 times, a model including baseline lymphocytes (HR = 0.43, <i>P</i> < 0.001) and mean dose to the identified region (HR = 1.02 Gy⁻¹, <i>P</i> = 0.006) adjusted for age (HR = 0.98, <i>P</i> = 0.09) and target volume (HR = 1.83, <i>P</i> = 0.07) was proposed. The Hosmer-Lemeshow test indicated that the model was a good fit, C-statistics 94%. <h3>Conclusion</h3> In this large oropharynx cohort, RT related lymphopenia ≥ G3 was associated with dose to a discrete anatomical region including structures containing blood, lymphoid tissue, and bone marrow that may offer a mechanistic explanation. While hypothesis generating, this suggests the possibility that treatment related lymphopenia may be mitigated by prioritizing dose reduction to this region. Further confirmatory studies are indicated.

OC-0639 Towards personalised treatment for prostate cancer: biology improves image-based data mining models

OC-0639 Towards personalised treatment for prostate cancer: biology improves image-based data mining models

PO-1822 Feasibility of spatial normalisation for image-based data mining in breast cancer radiotherapy

PO-1822 Feasibility of spatial normalisation for image-based data mining in breast cancer radiotherapy

Flogging a Dead Salmon? Reduced Dose Posterior to Prostate Correlates With Increased PSA Progression in Voxel-Based Analysis of 3 Randomized Phase 3 Trials

Flogging a Dead Salmon? Reduced Dose Posterior to Prostate Correlates With Increased PSA Progression in Voxel-Based Analysis of 3 Randomized Phase 3 Trials

Image Based Data Mining Using Per-voxel Cox Regression.

Image Based Data Mining (IBDM) is a novel analysis technique allowing the interrogation of large amounts of routine radiotherapy data. Using this technique, unexpected correlations have been identified between dose close to the prostate and biochemical relapse, and between dose to the base of the heart and survival in lung cancer. However, most analyses to date have considered only dose when identifying a region of interest, with confounding variables accounted for post-hoc, most often using a multivariate Cox regression. In this work, we introduce a novel method to account for confounding variables directly in the analysis, by performing a Cox regression in every voxel of the dose distribution, and apply it in the analysis of a large cohort of lung cancer patients. Our method produces three-dimensional maps of hazard for clinical variables, accounting for dose at each spatial location in the patient. Results confirm that a region of interest exists in the base of the heart where those patients with poor performance status (PS), PS > 1, have a stronger adverse reaction to incidental dose, but that the effect changes when considering other clinical variables, with patient age becoming dominant. Analyses such as this will help shape future clinical trials in which hypotheses generated by the analysis will be tested.

Radiotherapy-Related Lymphopenia Affects Overall Survival in Patients With Lung Cancer.

Lymphopenia after radiotherapy has an adverse effect on the patient's outcome. However, the relationship between radiotherapy dose delivery and lymphopenia is not fully understood. This work used image-based data mining to identify anatomical regions where the received dose is correlated with severe lymphopenia. A total of 901 patients with lung cancer were analyzed. A Cox model was used to assess prognostic factors of overall survival (OS). Two matched groups were defined-patients with lymphopenia of grade 3 or higher and patients without lymphopenia of grade 3-based on tumor volume, baseline lymphocytes, and prescribed dose. Then, data mining was used to identify regions where dose correlates significantly with lymphopenia of grade 3 or higher. For this, dose matrices were aligned using registration of the computed tomography images to one reference patient. Mean dose distributions were obtained for the two groups, and organs of significance were detected. Dosimetric parameters from the identified organs that had the highest correlation with lymphocytes at nadir were selected. Multivariable analysis was conducted for lymphopenia of grade 3 or higher on the full lung cohort, and the model was tested on 305 patients with esophageal cancer. Adjusted Cox regression revealed that lymphopenia of grade 3 or higher is an independent factor of OS. The anatomical regions identified were the heart, lung, and thoracic vertebrae. Dosimetric parameters for lymphopenia included thoracic vertebrae V20, mean lung dose, and mean heart dose, which were further validated in the esophageal cancer cohort. We report that severe lymphopenia during radiotherapy is a poor prognostic factor for OS in patients with lung cancer and could be mitigated by minimizing thoracic vertebrae V20, mean lung dose, and mean heart dose to limit the irradiation of stem cells and blood pool.

PyHIVE, a health-related image visualization and engineering system using Python

BackgroundImaging is one of the major biomedical technologies to investigate the status of a living object. But the biomedical image based data mining problem requires extensive knowledge across multiple disciplinaries, e.g. biology, mathematics and computer science, etc.ResultspyHIVE (a Health-related Image Visualization and Engineering system using Python) was implemented as an image processing system, providing five widely used image feature engineering algorithms. A standard binary classification pipeline was also provided to help researchers build data models immediately after the data is collected. pyHIVE may calculate five widely-used image feature engineering algorithms efficiently using multiple computing cores, and also featured the modules of Principal Component Analysis (PCA) based preprocessing and normalization.ConclusionsThe demonstrative example shows that the image features generated by pyHIVE achieved very good classification performances based on the gastrointestinal endoscopic images. This system pyHIVE and the demonstrative example are freely available and maintained at http://www.healthinformaticslab.org/supp/resources.php.

P1.17-01 Robustness of an Image-Based Data Mining Approach in Lung Cancer Patients

Image-based data mining (IBDM) enables exploring the correlation of dose distributions and outcomes in large cohorts of patients without the requirement of additional contouring. IBDM has recently identified the dose to the base of the heart as an important predictor for overall survival (OS) in lung cancer patients receiving radiotherapy [McWilliam et al EJC 2017]. IBDM relies on non-rigid registration to set inter-patient dosimetric data into a common reference anatomy or reference patient. Here, we investigated the uncertainties associated with the choice of reference patient, and their influence on the correlation between incidental dose to the base of the heart and OS. In previous work, 1101 NSCLC patients (55Gy / 20 fractions) were randomly selected, and their planning CT images non-rigidly registered to a reference patient CT scan using NiftyReg (http://cmictig.cs.ucl.ac.uk/wiki/) as part of IBDM process. In this work, 5 additional patients with small cell lung cancer (i.e. without a large tumour burden) were used as “reference patients” and the IBDM analysis in the whole cohort was repeated for each reference patient. Permutation testing with 100 iterations was applied to assess statistical significance. Figure 1 shows the regions of highly significant correlation between dose and OS for each reference patient. In spite of large variations in anatomy between the reference patients, each analysis identified similar anatomical regions as significantly associated with OS (t>5). Moreover, permutation testing was consistent with the original findings. IBDM is a robust approach and, in this analysis, does not appear to be sensitive to the choice of reference patient for the investigated dose-effect correlation. Prospective studies are necessary to confirm the correlation between dose to the base of the heart and OS in NSCLC patients. Methodological studies are needed to determine the level of effect strength and region size that this general technique can identify.

OA072] Per-voxel cox regression with interaction terms

Purpose Image based data mining (IBDM) is a novel technique in which dose response relationships are extracted from routine clinical data by relating the dose at each voxel in the planned dose to the eventual outcome for the patient. In IBDM many patients are registered to a reference patient to spatially normalise their dose distributions for analysis, eliminating the need for clinician contours and removing assumptions implicit in DVH based analyses. Recently, IBDM has been used to identify a dose sensitive region in the heart where patients receiving excess dose have poorer overall survival [1] . Methods In this work, a subset (n = 872) of the cohort of patients in [1] is used, but a more sophisticated IBDM technique is applied. A Cox regression, considering tumour size, age, performance status, dose and a dose-PS interaction is calculated in each voxel of the planned dose distribution. Performance status is binarised using a threshold of PS > = 1 to account for inaccuracy in its determination. This analysis leads to a map of hazard ratio across the reference patient anatomy. Statistical significance is assessed using a permutation test in which only survival labels are permuted. Results When performing the analysis with an interaction between performance status and dose present in the Cox model, a region is identified in the base of the heart where excess dose is detrimental to survival. This region overlaps with that found by other analyses, and suggests that patients with a performance status of 2 or greater suffer a more profound effect from dose in this region. In the present analysis statistical significance is not reached in any voxel for either the dose, PS or dose-PS interaction. Conclusions A region in the base of the heart is identified, consistent with other analyses, in which excess dose is related to poorer overall survival, particularly for those patients with poor performance. Statistical significance may be reached by including more data, or modifying the permutation technique.