Advertisement

Towards spatial representations of dose distributions to predict risk of normal tissue morbidity after radiotherapy

  • Oscar Casares-Magaz
    Affiliations
    Department of Medical Physics – Oncology, Aarhus University/Aarhus University Hospital, Aarhus, Denmark
    Search for articles by this author
  • Vitali Moiseenko
    Affiliations
    Department of Radiation Medicine and Applied Science, University of California San Diego, La Jolla, CA, United States
    Search for articles by this author
  • Marnix Witte
    Affiliations
    Cluster Radiation Oncology, The Netherlands Cancer Institute, Amsterdam, The Netherlands
    Search for articles by this author
  • Tiziana Rancati
    Affiliations
    Prostate Cancer Program, Fondazione IRCCS Istituto Nazionale dei Tumori di Milano, Milan, Italy
    Search for articles by this author
  • Author Footnotes
    1 Ludvig Muren, a co-author of this paper, is an Editor-in-Chief of Physics & Imaging in Radiation Oncology. The editorial process for this manuscript was managed independently from Prof. Muren.
    Ludvig P. Muren
    Footnotes
    1 Ludvig Muren, a co-author of this paper, is an Editor-in-Chief of Physics & Imaging in Radiation Oncology. The editorial process for this manuscript was managed independently from Prof. Muren.
    Affiliations
    Department of Medical Physics – Oncology, Aarhus University/Aarhus University Hospital, Aarhus, Denmark
    Search for articles by this author
  • Author Footnotes
    1 Ludvig Muren, a co-author of this paper, is an Editor-in-Chief of Physics & Imaging in Radiation Oncology. The editorial process for this manuscript was managed independently from Prof. Muren.
Open AccessPublished:August 27, 2020DOI:https://doi.org/10.1016/j.phro.2020.08.002
      Design of clinical trials and definition of new standards of care in radiotherapy have historically been driven by the balance between doses delivered to the target volume against doses delivered to the healthy organs surrounding the tumour. State-of-the-art high-precision radiotherapy, together with other advances in cancer therapy, have led to improved clinical outcomes [
      • Bray F.
      • Ferlay J.
      • Soerjomataram I.
      • Siegel R.L.
      • Torre L.A.
      • Jemal A.
      Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries.
      ].
      In the setting of conventionally fractionated radiotherapy, dose distributions within target volumes are reasonably uniform [

      International Commission on Radiation Units and Measurements (ICRU). Prescribing, Recording, and Reporting Intensity-Modulated Photon-Beam Therapy (IMRT)(ICRU Report 83). https://icru.org/testing/reports/prescribing-recording-and-reporting-intensity-modulated-photon-beam-therapy-imrt-icru-report-83.

      ]. In contrast, normal tissue dose distributions are highly non-uniform and their shape can be tailored while still achieving planning goals. These planning goals are typically formulated in terms of dose-volume constraints which drive optimization towards treatment plans deemed best for each patient. Visual inspection of how isodose lines are shaped around target volumes and into normal tissues is an integral part of plan checking and approval. Most radiotherapy plan optimization, evaluation and outcome analysis have been based on summarizing 3D dose distributions into dose-volume histograms (DVH). This approach has been successfully used to describe and utilize dose-volume-response relationships [
      • Olsson C.E.
      • Jackson A.
      • Deasy J.O.
      • Thor M.
      A systematic post-QUANTEC review of tolerance doses for late toxicity after prostate cancer radiation therapy.
      ].
      Limitations of DVH-based outcome assessments are well known. The DVH strips away geometry information, i.e. every region of normal tissue is seen as equally important, and no cross-talk between regions is considered. Clinical data support existence of regional effects, although explanations for these effects remain elusive [
      • Heemsbergen W.D.
      • Incrocci L.
      • Pos F.J.
      • Heijmen B.J.M.
      • Witte M.G.
      Local Dose effects for late gastrointestinal toxicity after hypofractionated and conventionally fractionated modern radiotherapy for prostate cancer in the HYPRO trial.
      ,
      • Gagliardi G.
      • Constine L.S.
      • Moiseenko V.
      • Correa C.
      • Pierce L.J.
      • Allen A.M.
      • et al.
      Radiation dose-volume effects in the heart.
      ]. Commonly used rationale to justify more sophisticated approaches relies on one or more of the following, and to a certain extent overlapping, arguments: 1) functional burden is not uniformly distributed in normal tissues; 2) stem cells capable of rescuing normal tissue function are primarily concentrated in particular parts of normal tissue; 3) organs which are contoured as a whole are in fact anatomically/geometrically sub-structured, e.g., lobes in lung or tracts in brain, and risk of morbidity does depend on which substructures receive dose; 4) parts of an organ are differentially sensitive to radiation; 5) morbidity is related to radiation-induced damage to organs other than ones for which DVHs are evaluated; 6) spatial distribution matters, e.g. small hot spots spread out on the rectal surface might have different consequences compared to larger, spatially clustered patterns.
      Dose-volume-response relationships that also incorporate the geometrical information of dose distributions have therefore received considerable interest. The first studies including geometrical information in normal tissue complication probability (NTCP) models utilized dose surface maps (DSMs), and were focused on hollow organs, in particular to study bladder and rectum dose-volume-response relationships following radiotherapy for prostate cancer [
      • Tucker S.L.
      • Zhang M.
      • Dong L.
      • Mohan R.
      • Kuban D.
      • Thames H.D.
      Cluster model analysis of late rectal bleeding after IMRT of prostate cancer: a case-control study.
      ,
      • Buettner F.
      • Gulliford S.L.
      • Webb S.
      • Partridge M.
      Using dose-surface maps to predict radiation-induced rectal bleeding: a neural network approach.
      ,
      • Buettner F.
      • Gulliford S.L.
      • Webb S.
      • Sydes M.R.
      • Dearnaley D.P.
      • Partridge M.
      Assessing correlations between the spatial distribution of the dose to the rectal wall and late rectal toxicity after prostate radiotherapy: an analysis of data from the MRC RT01 trial (ISRCTN 47772397).
      ]. Therefore, utilization of DSMs has helped to demonstrate that the dose delivered to the caudal part together with shape and extension of the hot spots over rectal wall are related to an exacerbation of morbidity [
      • Shelley L.E.A.
      • Scaife J.E.
      • Romanchikova M.
      • Harrison K.
      • Forman J.R.
      • Bates A.M.
      • et al.
      Delivered dose can be a better predictor of rectal toxicity than planned dose in prostate radiotherapy.
      ,
      • Shelley L.E.A.
      • Sutcliffe M.P.F.
      • Thomas S.J.
      • Noble D.J.
      • Romanchikova M.
      • Harrison K.
      • et al.
      Associations between voxel-level accumulated dose and rectal toxicity in prostate radiotherapy.
      ,
      • Wortel R.C.
      • Witte M.G.
      • van der Heide U.A.
      • Pos F.J.
      • Lebesque J.V.
      • van Herk M.
      • et al.
      Dose-surface maps identifying local dose-effects for acute gastrointestinal toxicity after radiotherapy for prostate cancer.
      ,
      • Casares-Magaz O.
      • Bülow S.
      • Pettersson N.J.
      • Moiseenko V.
      • Pedersen J.
      • Thor M.
      • et al.
      High accumulated doses to the inferior rectum are associated with late gastro-intestinal toxicity in a case-control study of prostate cancer patients treated with radiotherapy.
      ,
      • Onjukka E.
      • Fiorino C.
      • Cicchetti A.
      • Palorini F.
      • Improta I.
      • Gagliardi G.
      • et al.
      Patterns in ano-rectal dose maps and the risk of late toxicity after prostate IMRT.
      ]. Similarly, doses delivered to the trigone area and the urethra were found to play a major role in the overall worsening of genito-urinary symptoms following radiotherapy for prostate cancer [
      • Palorini F.
      • Cozzarini C.
      • Gianolini S.
      • Botti A.
      • Carillo V.
      • Iotti C.
      • et al.
      First application of a pixel-wise analysis on bladder dose-surface maps in prostate cancer radiotherapy.
      ,
      • Mylona E.
      • Acosta O.
      • Lizee T.
      • Lafond C.
      • Crehange G.
      • Magné N.
      • et al.
      Voxel-based analysis for identification of urethrovesical subregions predicting urinary toxicity after prostate cancer radiation therapy.
      ,
      • Yahya N.
      • Ebert M.A.
      • House M.J.
      • Kennedy A.
      • Matthews J.
      • Joseph D.J.
      • et al.
      Modeling urinary dysfunction after external beam radiation therapy of the prostate using bladder dose-surface maps: evidence of spatially variable response of the bladder surface.
      ,
      • Ghadjar P.
      • Zelefsky M.J.
      • Spratt D.E.
      • Munck P.
      • af
      • Rosenschöld
      • Oh J.H.
      • Hunt M.
      • et al.
      Impact of dose to the bladder trigone on long-term urinary function after high-dose intensity modulated radiation therapy for localized prostate cancer.
      ,
      • Heemsbergen W.D.
      • Al-Mamgani A.
      • Witte M.G.
      • van Herk M.
      • Pos F.J.
      • Lebesque J.V.
      Urinary obstruction in prostate cancer patients from the Dutch trial (68 Gy vs. 78 Gy): relationships with local dose, acute effects, and baseline characteristics.
      ]. Overall, DSMs are often used in an empirical approach, exploiting significant differences in dose distribution patterns between patients with and without morbidity to identify regions which might correspond to substructures governing the dose response. In the search for methods to include geometrical information into NTCP models, a more theoretical or deterministic approach can also be considered. A deterministic approach hypothesizes that an organ/tissue is composed by subunits with different radiosensitivity, with one (or more) playing a major role in the manifestation of morbidity. Therefore, damage to an identified substructure(s) may trigger manifestation of the observed morbidity.
      This issue of the journal presents the first study describing the impact of dose to heart substructures on the overall survival rate [
      • McWilliam A.
      • Dootson C.
      • Graham L.
      • Banfill K.
      • Abravan A.
      • van Herk M.
      Dose surface maps of the heart can identify regions associated with worse survival for lung cancer patients treated with radiotherapy.
      ]. Cardiac disease associated with radiation has been observed in cancer patients treated to thoracic and abdominal regions, i.e. breast cancer and lymphoma patients, because they are typically long-term survivors, providing most of the outcome data. The QUANTEC report on the heart provided dose-volume guidelines for two endpoints: pericarditis and long-term cardiac mortality [
      • Gagliardi G.
      • Constine L.S.
      • Moiseenko V.
      • Correa C.
      • Pierce L.J.
      • Allen A.M.
      • et al.
      Radiation dose-volume effects in the heart.
      ]. For the former, mean dose < 26 Gy and V30Gy < 46% were suggested. Cardiac mortality, in contrast to pericarditis, appears to exhibit a weak volume effect. This observation has led to substantial changes in how radiotherapy is delivered to left-sided breast cancer patients. Heart blocks including the “no heart in beams-eye-view” policies have been adopted, deep inspiration breath hold and treating prone techniques have also been used to spare the heart. The QUANTEC heart paper proposed a V25Gy < 10% constraint to control the “tail” of the DVH [
      • Gagliardi G.
      • Constine L.S.
      • Moiseenko V.
      • Correa C.
      • Pierce L.J.
      • Allen A.M.
      • et al.
      Radiation dose-volume effects in the heart.
      ], however, this constraint has been later shown to not fully protect against the probability of cardiac mortality < 1% in breast cancer patients [
      • Moiseenko V.
      • Einck J.
      • Murphy J.
      • Ödén J.
      • Bjöhle J.
      • Uzan J.
      • et al.
      Clinical evaluation of QUANTEC guidelines to predict the risk of cardiac mortality in breast cancer patients.
      ]. This also underlines another deficiency in using DVH constraints. Various endpoints exhibit different dose-volume response. Consequently optimizing, e.g. mean dose, will reduce probability of pericarditis but is not sufficient to control the risk of cardiac mortality. Identifying an endpoint of the greatest concern or with the tightest constraints is one possible way but it may become too restrictive.
      For the heart also the initial definition and delineation of the organ at risk is challenging. Specifically, the whole heart can be contoured, while also the pericardium (often defined as a “shell” expansion of the heart contour), the left ventricle, coronary vessels or specifically the left anterior descending coronary artery are used in different protocols and institutions [
      • Gagliardi G.
      • Constine L.S.
      • Moiseenko V.
      • Correa C.
      • Pierce L.J.
      • Allen A.M.
      • et al.
      Radiation dose-volume effects in the heart.
      ]. Different endpoints have been connected to anatomical features and functionality, for example the pericardium is likely appropriate for pericardial effusion and the left anterior descending artery for ischemic heart disease [
      • Moiseenko V.
      • Einck J.
      • Murphy J.
      • Ödén J.
      • Bjöhle J.
      • Uzan J.
      • et al.
      Clinical evaluation of QUANTEC guidelines to predict the risk of cardiac mortality in breast cancer patients.
      ]. However, dose-volume parameters for different structures behave in a correlated manner which complicates search for the “guilty” party, and a full mechanistic understanding of how morbidity develops is lacking.
      Searching for regional effects may provide us with a guide to selective sparing of either structures, or geometrically defined sections of the organ. Because multiple anatomically/geometrically defined structures/regions must be analysed for different endpoints, automation might be required. Auto-contouring not only allows us to tackle a laborious task of delineating structures which are not routinely delineated in clinical practice, it may also improve consistency [
      • Haq R.
      • Hotca A.
      • Apte A.
      • Rimner A.
      • Deasy J.O.
      • Thor M.
      Cardio-pulmonary substructure segmentation of radiotherapy computed tomography images using convolutional neural networks for clinical outcomes analysis.
      ]. Additionally, auto-contouring tools included in retrospective analysis will pave the way to interrogating outcomes data and will further promote prospective use for selective regional sparing [
      • Jung J.W.
      • Lee C.
      • Mosher E.G.
      • Mille M.M.
      • Yeom Y.S.
      • Jones E.C.
      • et al.
      Automatic segmentation of cardiac structures for breast cancer radiotherapy.
      ].
      An overall aim of dose-surface based outcome studies is to identify the substructures of the organ that play a key role in dose–response relationships, and from there provide robust metrics to be used in plan optimisation. Most studies in this field have applied a voxel-wise analysis to quantify significant dose differences [
      • Shelley L.E.A.
      • Sutcliffe M.P.F.
      • Thomas S.J.
      • Noble D.J.
      • Romanchikova M.
      • Harrison K.
      • et al.
      Associations between voxel-level accumulated dose and rectal toxicity in prostate radiotherapy.
      ,
      • Wortel R.C.
      • Witte M.G.
      • van der Heide U.A.
      • Pos F.J.
      • Lebesque J.V.
      • van Herk M.
      • et al.
      Dose-surface maps identifying local dose-effects for acute gastrointestinal toxicity after radiotherapy for prostate cancer.
      ,
      • Casares-Magaz O.
      • Bülow S.
      • Pettersson N.J.
      • Moiseenko V.
      • Pedersen J.
      • Thor M.
      • et al.
      High accumulated doses to the inferior rectum are associated with late gastro-intestinal toxicity in a case-control study of prostate cancer patients treated with radiotherapy.
      ,
      • Onjukka E.
      • Fiorino C.
      • Cicchetti A.
      • Palorini F.
      • Improta I.
      • Gagliardi G.
      • et al.
      Patterns in ano-rectal dose maps and the risk of late toxicity after prostate IMRT.
      ,
      • Palorini F.
      • Cozzarini C.
      • Gianolini S.
      • Botti A.
      • Carillo V.
      • Iotti C.
      • et al.
      First application of a pixel-wise analysis on bladder dose-surface maps in prostate cancer radiotherapy.
      ,
      • Mylona E.
      • Acosta O.
      • Lizee T.
      • Lafond C.
      • Crehange G.
      • Magné N.
      • et al.
      Voxel-based analysis for identification of urethrovesical subregions predicting urinary toxicity after prostate cancer radiation therapy.
      ]. Voxel-based data mining of radiotherapy doses relies on the comparison of delivered dose distributions, aggregated over subgroups of patients with and without the morbidity endpoint of interest. Such aggregation requires a common frame of reference, which is chosen based on the anatomy of the organ and the nature of the endpoint of interest. For hollow organs simplified spherical or cylindrical geometries are common, allowing for an easy anatomical interpretation of results. In case such simplified geometries are unable to capture the relevant effects, more sophisticated methods based on validated deformable registration techniques may offer increased statistical power [
      • Mylona E.
      • Acosta O.
      • Lizee T.
      • Lafond C.
      • Crehange G.
      • Magné N.
      • et al.
      Voxel-based analysis for identification of urethrovesical subregions predicting urinary toxicity after prostate cancer radiation therapy.
      ]. As these typically have much larger degrees of freedom, careful validations should be performed to prevent overfitting. A next challenge is the estimation of the actually delivered dose to tissues in the chosen frame of reference; often only treatment planning geometries and dose distributions are available, which in the case of a highly mobile organ may only offer a rough estimate. Leveraging treatment room imaging modalities (e.g. cone-beam computed tomography or magnetic resonance imaging guided systems) may improve these estimates and hence strengthen the statistical associations between doses and the morbidity endpoints [
      • Shelley L.E.A.
      • Scaife J.E.
      • Romanchikova M.
      • Harrison K.
      • Forman J.R.
      • Bates A.M.
      • et al.
      Delivered dose can be a better predictor of rectal toxicity than planned dose in prostate radiotherapy.
      ,
      • Casares-Magaz O.
      • Bülow S.
      • Pettersson N.J.
      • Moiseenko V.
      • Pedersen J.
      • Thor M.
      • et al.
      High accumulated doses to the inferior rectum are associated with late gastro-intestinal toxicity in a case-control study of prostate cancer patients treated with radiotherapy.
      ,
      • Thor M.
      • Bentzen L.
      • Hysing L.B.
      • Ekanger C.
      • Helle S.-I.
      • Karlsdóttir Á.
      • et al.
      Prediction of rectum and bladder morbidity following radiotherapy of prostate cancer based on motion-inclusive dose distributions.
      ,
      • Alam S.
      • Thor M.
      • Rimner A.
      • Tyagi N.
      • Zhang S.Y.
      • Cheng Kuo L.
      • et al.
      Quantification of accumulated dose and associated anatomical changes of esophagus using weekly Magnetic Resonance Imaging acquired during radiotherapy of locally advanced lung cancer.
      ].
      Estimation of the significance of an observed difference in population average doses over the outcome subgroups (controlling the multiple testing issues introduced by the large numbers of voxels analysed) is generally established using a permutation-based approach, randomly re-shuffling patients between the morbidity and non-morbidity subgroups [
      • Chen C.
      • Witte M.G.
      • Heemsbergen W.D.
      • van Herk M.
      Multiple comparisons permutation test for image based data mining in radiotherapy.
      ]. The associated p-value indicates per voxel significant dose differences. One generally (ad-hoc) selects a seemingly relevant threshold of the test statistic (e.g. local dose difference divided by standard deviation) and selects voxels in the observation which exceed this value. It should be noted that all subsequent (statistical) modelling using these selected voxels should be interpreted with care; even for a subset of voxels which have a large dose difference merely by chance, subsequent modelling would lead to erroneous, but seemingly significant dose–effect relations and survival analyses.
      Attempts can be made to interpret the observed dose difference patterns in terms of an underlying anatomical or biological cause; a crucial, and likely the most difficult step of the dose data mining methodology. Suppose one would analyse a group of prostate cancer patients who had all been “ideally” irradiated, so with an optimal compromise between tumour control probability and risk of morbidity, and one would construct rectum morbidity dose difference maps. Patients with less favourable anatomy would have received high dose to a larger fraction of the rectum wall, leading to higher morbidity rates, and dose distributions might appear slightly shifted compared to those from patients with more favourable anatomy. These dose difference regions illustrate the well-known fact that dose causes morbidity, and merely pin-point the locations where dose variations exist due to patient specific anatomies and limitations of the applied external beam radiotherapy modality [
      • Wortel R.C.
      • Witte M.G.
      • van der Heide U.A.
      • Pos F.J.
      • Lebesque J.V.
      • van Herk M.
      • et al.
      Dose-surface maps identifying local dose-effects for acute gastrointestinal toxicity after radiotherapy for prostate cancer.
      ,
      • Palorini F.
      • Cozzarini C.
      • Gianolini S.
      • Botti A.
      • Carillo V.
      • Iotti C.
      • et al.
      First application of a pixel-wise analysis on bladder dose-surface maps in prostate cancer radiotherapy.
      ,
      • Mylona E.
      • Acosta O.
      • Lizee T.
      • Lafond C.
      • Crehange G.
      • Magné N.
      • et al.
      Voxel-based analysis for identification of urethrovesical subregions predicting urinary toxicity after prostate cancer radiation therapy.
      ]. Misinterpretation of these locations as dose avoidance regions, e.g. loading these patterns into a treatment planning system to steer dose elsewhere, could deteriorate the (already optimal) treatment plans and lead to less favourable clinical outcomes. Rather, observed patterns should serve as the basis for an informed hypothesis of the mechanisms underlying the clinical manifestation of morbidity; e.g. in the way these patterns align with potentially critical substructures of the organ. Next, prospective tests should be devised to validate these findings before being incorporated into NTCP models which may then improve radiotherapy for future patients.
      Further attention should be paid to the possible impact of differences between planned and delivered doses, mainly driven by organ motion. In many cases statistically significant local dose effects are found in sub-regions where the positioning of organs is more stable [
      • Palorini F.
      • Cozzarini C.
      • Gianolini S.
      • Botti A.
      • Carillo V.
      • Iotti C.
      • et al.
      First application of a pixel-wise analysis on bladder dose-surface maps in prostate cancer radiotherapy.
      ,
      • Palorini F.
      • Botti A.
      • Carillo V.
      • Gianolini S.
      • Improta I.
      • Iotti C.
      • et al.
      Bladder dose–surface maps and urinary toxicity: Robustness with respect to motion in assessing local dose effects.
      ]. This enhances the chance of finding statistically significant differences in the doses between patients with and without morbidity because in those more stable sub-regions the planned dose is similar to the delivered dose. An explicative situation is given by the non-isotropic motion of the bladder due to its variable filling, with the base being relatively stable while the cranial portion showing large variations. This fact entails lower systematic errors (planned vs. delivered dose) in the high dose region which corresponds to the bladder base [
      • Palorini F.
      • Botti A.
      • Carillo V.
      • Gianolini S.
      • Improta I.
      • Iotti C.
      • et al.
      Bladder dose–surface maps and urinary toxicity: Robustness with respect to motion in assessing local dose effects.
      ]. Results on local dose effects for the bladder base/trigone could thus also be due to the higher reliability of the investigated features (doses at the pixel level) in the motion-stable region, with doses to the bladder dome coming out as not significant due to the limited correspondence between planned and delivered doses. Systematic studies on organ motion and on differences between planned and delivered dose could help in discriminating between biologically relevant local dose effects and effects driven by the inaccurate description of delivered doses in some regions [
      • Simon A.
      • Nassef M.
      • Cazoulat G.
      • Acosta O.
      • Lafond C.
      • Haigron P.
      • et al.
      Quantification of the Differences Between Planned and Delivered Doses in the Bladder in Prostate IGRT.
      ,
      • Casares-Magaz O.
      • Moiseenko V.
      • Hopper A.
      • Pettersson N.J.
      • Thor M.
      • Knopp R.
      • et al.
      Associations between volume changes and spatial dose metrics for the urinary bladder during local versus pelvic irradiation for prostate cancer.
      ].
      DSMs might increase predictive power of normal tissue dose response relationships. New findings from voxel-based data mining should be accompanied by studies on organ motion, on estimation of differences between planned and delivered dose and on the possible biological role of the regions identified as significantly associated with an increased risk of radiation-induced morbidity. Finally, generalisability studied on a suitable number of independent cohorts will pave the way to clinical applications of the results.

      Declaration of Competing Interest

      The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

      References

        • Bray F.
        • Ferlay J.
        • Soerjomataram I.
        • Siegel R.L.
        • Torre L.A.
        • Jemal A.
        Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries.
        CA Cancer J Clin. 2018; 68: 394-424https://doi.org/10.3322/caac.21492
      1. International Commission on Radiation Units and Measurements (ICRU). Prescribing, Recording, and Reporting Intensity-Modulated Photon-Beam Therapy (IMRT)(ICRU Report 83). https://icru.org/testing/reports/prescribing-recording-and-reporting-intensity-modulated-photon-beam-therapy-imrt-icru-report-83.

        • Olsson C.E.
        • Jackson A.
        • Deasy J.O.
        • Thor M.
        A systematic post-QUANTEC review of tolerance doses for late toxicity after prostate cancer radiation therapy.
        Int J Radiat Oncol Biol Phys. 2018; 102: 1514-1532https://doi.org/10.1016/j.ijrobp.2018.08.015
        • Heemsbergen W.D.
        • Incrocci L.
        • Pos F.J.
        • Heijmen B.J.M.
        • Witte M.G.
        Local Dose effects for late gastrointestinal toxicity after hypofractionated and conventionally fractionated modern radiotherapy for prostate cancer in the HYPRO trial.
        Front Oncol. 2020; 10https://doi.org/10.3389/fonc.2020.00469
        • Gagliardi G.
        • Constine L.S.
        • Moiseenko V.
        • Correa C.
        • Pierce L.J.
        • Allen A.M.
        • et al.
        Radiation dose-volume effects in the heart.
        Int J Radiat Oncol Biol Phys. 2010; 76: S77-S85https://doi.org/10.1016/j.ijrobp.2009.04.093
        • Tucker S.L.
        • Zhang M.
        • Dong L.
        • Mohan R.
        • Kuban D.
        • Thames H.D.
        Cluster model analysis of late rectal bleeding after IMRT of prostate cancer: a case-control study.
        Int J Radiat Oncol Biol Phys. 2006; 64: 1255-1264https://doi.org/10.1016/j.ijrobp.2005.10.029
        • Buettner F.
        • Gulliford S.L.
        • Webb S.
        • Partridge M.
        Using dose-surface maps to predict radiation-induced rectal bleeding: a neural network approach.
        Phys Med Biol. 2009; 54: 5139-5153https://doi.org/10.1088/0031-9155/54/17/005
        • Buettner F.
        • Gulliford S.L.
        • Webb S.
        • Sydes M.R.
        • Dearnaley D.P.
        • Partridge M.
        Assessing correlations between the spatial distribution of the dose to the rectal wall and late rectal toxicity after prostate radiotherapy: an analysis of data from the MRC RT01 trial (ISRCTN 47772397).
        Phys Med Biol. 2009; 54: 6535-6548https://doi.org/10.1088/0031-9155/54/21/006
        • Shelley L.E.A.
        • Scaife J.E.
        • Romanchikova M.
        • Harrison K.
        • Forman J.R.
        • Bates A.M.
        • et al.
        Delivered dose can be a better predictor of rectal toxicity than planned dose in prostate radiotherapy.
        Radiother Oncol. 2017; 123: 466-471https://doi.org/10.1016/j.radonc.2017.04.008
        • Shelley L.E.A.
        • Sutcliffe M.P.F.
        • Thomas S.J.
        • Noble D.J.
        • Romanchikova M.
        • Harrison K.
        • et al.
        Associations between voxel-level accumulated dose and rectal toxicity in prostate radiotherapy.
        Phys Imag Radiat Oncol. 2020; 14: 87-94https://doi.org/10.1016/j.phro.2020.05.006
        • Wortel R.C.
        • Witte M.G.
        • van der Heide U.A.
        • Pos F.J.
        • Lebesque J.V.
        • van Herk M.
        • et al.
        Dose-surface maps identifying local dose-effects for acute gastrointestinal toxicity after radiotherapy for prostate cancer.
        Radiother Oncol. 2015; 117: 515-520https://doi.org/10.1016/j.radonc.2015.10.020
        • Casares-Magaz O.
        • Bülow S.
        • Pettersson N.J.
        • Moiseenko V.
        • Pedersen J.
        • Thor M.
        • et al.
        High accumulated doses to the inferior rectum are associated with late gastro-intestinal toxicity in a case-control study of prostate cancer patients treated with radiotherapy.
        Acta Oncol. 2019; 58: 1543-1546https://doi.org/10.1080/0284186X.2019.1632476
        • Onjukka E.
        • Fiorino C.
        • Cicchetti A.
        • Palorini F.
        • Improta I.
        • Gagliardi G.
        • et al.
        Patterns in ano-rectal dose maps and the risk of late toxicity after prostate IMRT.
        Acta Oncol. 2019; 58: 1757-1764https://doi.org/10.1080/0284186X.2019.1635267
        • Palorini F.
        • Cozzarini C.
        • Gianolini S.
        • Botti A.
        • Carillo V.
        • Iotti C.
        • et al.
        First application of a pixel-wise analysis on bladder dose-surface maps in prostate cancer radiotherapy.
        Radiother Oncol. 2016; 119: 123-128https://doi.org/10.1016/j.radonc.2016.02.025
        • Mylona E.
        • Acosta O.
        • Lizee T.
        • Lafond C.
        • Crehange G.
        • Magné N.
        • et al.
        Voxel-based analysis for identification of urethrovesical subregions predicting urinary toxicity after prostate cancer radiation therapy.
        Int J Radiat Oncol Biol Phys. 2019; 104: 343-354https://doi.org/10.1016/j.ijrobp.2019.01.088
        • Yahya N.
        • Ebert M.A.
        • House M.J.
        • Kennedy A.
        • Matthews J.
        • Joseph D.J.
        • et al.
        Modeling urinary dysfunction after external beam radiation therapy of the prostate using bladder dose-surface maps: evidence of spatially variable response of the bladder surface.
        Int J Radiat Oncol Biol Phys. 2017; 97: 420-426https://doi.org/10.1016/j.ijrobp.2016.10.024
        • Ghadjar P.
        • Zelefsky M.J.
        • Spratt D.E.
        • Munck P.
        • af
        • Rosenschöld
        • Oh J.H.
        • Hunt M.
        • et al.
        Impact of dose to the bladder trigone on long-term urinary function after high-dose intensity modulated radiation therapy for localized prostate cancer.
        Int J Radiat Oncol Biol Phys. 2014; 88: 339-344https://doi.org/10.1016/j.ijrobp.2013.10.042
        • Heemsbergen W.D.
        • Al-Mamgani A.
        • Witte M.G.
        • van Herk M.
        • Pos F.J.
        • Lebesque J.V.
        Urinary obstruction in prostate cancer patients from the Dutch trial (68 Gy vs. 78 Gy): relationships with local dose, acute effects, and baseline characteristics.
        Int J Radiat Oncol Biol Phys. 2010; 78: 19-25https://doi.org/10.1016/j.ijrobp.2009.07.1680
        • McWilliam A.
        • Dootson C.
        • Graham L.
        • Banfill K.
        • Abravan A.
        • van Herk M.
        Dose surface maps of the heart can identify regions associated with worse survival for lung cancer patients treated with radiotherapy.
        Phys Imag Radiat Oncol. 2020; 15: 46-51https://doi.org/10.1016/j.phro.2020.07.002
        • Moiseenko V.
        • Einck J.
        • Murphy J.
        • Ödén J.
        • Bjöhle J.
        • Uzan J.
        • et al.
        Clinical evaluation of QUANTEC guidelines to predict the risk of cardiac mortality in breast cancer patients.
        Acta Oncol. 2016; 55: 1506-1510https://doi.org/10.1080/0284186X.2016.1234067
        • Haq R.
        • Hotca A.
        • Apte A.
        • Rimner A.
        • Deasy J.O.
        • Thor M.
        Cardio-pulmonary substructure segmentation of radiotherapy computed tomography images using convolutional neural networks for clinical outcomes analysis.
        Phys Imag Radiat Oncol. 2020; 14: 61-66https://doi.org/10.1016/j.phro.2020.05.009
        • Jung J.W.
        • Lee C.
        • Mosher E.G.
        • Mille M.M.
        • Yeom Y.S.
        • Jones E.C.
        • et al.
        Automatic segmentation of cardiac structures for breast cancer radiotherapy.
        Phys Imag Radiat Oncol. 2019; 12: 44-48https://doi.org/10.1016/j.phro.2019.11.007
        • Thor M.
        • Bentzen L.
        • Hysing L.B.
        • Ekanger C.
        • Helle S.-I.
        • Karlsdóttir Á.
        • et al.
        Prediction of rectum and bladder morbidity following radiotherapy of prostate cancer based on motion-inclusive dose distributions.
        Radiother Oncol. 2013; 107: 147-152https://doi.org/10.1016/j.radonc.2013.03.029
        • Alam S.
        • Thor M.
        • Rimner A.
        • Tyagi N.
        • Zhang S.Y.
        • Cheng Kuo L.
        • et al.
        Quantification of accumulated dose and associated anatomical changes of esophagus using weekly Magnetic Resonance Imaging acquired during radiotherapy of locally advanced lung cancer.
        Phys Imag Radiat Oncol. 2020; 13: 36-43https://doi.org/10.1016/j.phro.2020.03.002
        • Chen C.
        • Witte M.G.
        • Heemsbergen W.D.
        • van Herk M.
        Multiple comparisons permutation test for image based data mining in radiotherapy.
        Radiat Oncol. 2013; 8: 293https://doi.org/10.1186/1748-717X-8-293
        • Palorini F.
        • Botti A.
        • Carillo V.
        • Gianolini S.
        • Improta I.
        • Iotti C.
        • et al.
        Bladder dose–surface maps and urinary toxicity: Robustness with respect to motion in assessing local dose effects.
        Phys Med. 2016; 32: 506-511https://doi.org/10.1016/j.ejmp.2016.03.006
        • Simon A.
        • Nassef M.
        • Cazoulat G.
        • Acosta O.
        • Lafond C.
        • Haigron P.
        • et al.
        Quantification of the Differences Between Planned and Delivered Doses in the Bladder in Prostate IGRT.
        Int J Radiat Oncol. 2014; 90: S435https://doi.org/10.1016/j.ijrobp.2014.05.1370
        • Casares-Magaz O.
        • Moiseenko V.
        • Hopper A.
        • Pettersson N.J.
        • Thor M.
        • Knopp R.
        • et al.
        Associations between volume changes and spatial dose metrics for the urinary bladder during local versus pelvic irradiation for prostate cancer.
        Acta Oncol. 2017; 56: 884-890https://doi.org/10.1080/0284186X.2017.1312014