Surface scanning for 3D dose calculation in intraoperative electron radiation therapy
Dose calculations in intraoperative electron radiation therapy (IOERT) rely on the conventional assumption of water-equivalent tissues at the applicator end, which defines a flat irradiation surface. However, the shape of the irradiation surface modifies the dose distribution. Our study explores, for the first time, the use of surface scanning methods for three-dimensional dose calculation of IOERT.
Two different three-dimensional scanning technologies were evaluated in a simulated IOERT scenario: a tracked conoscopic holography sensor (ConoProbe) and a structured-light three-dimensional scanner (Artec). Dose distributions obtained from computed tomography studies of the surgical field (gold standard) were compared with those calculated under the conventional assumption or from pseudo-computed tomography studies based on surfaces.
In the simulated IOERT scenario, the conventional assumption led to an average gamma pass rate of 39.9% for dose values greater than 10% (two configurations, with and without blood in the surgical field). Results improved when considering surfaces in the dose calculation (88.5% for ConoProbe and 92.9% for Artec).
More accurate three-dimensional dose distributions were obtained when considering surfaces in the dose calculation of the simulated surgical field. The structured-light three-dimensional scanner provided the best results in terms of dose distributions. The findings obtained in this specific experimental setup warrant further research on surface scanning in the IOERT context owing to the clinical interest of improving the documentation of the actual IOERT scenario.
KeywordsIOERT Intraoperative radiotherapy Surface scanning Conoscopic holography Structured-light 3D scanner Dose distribution
Pseudo-CT study obtained from Artec surface
Pseudo-CT study obtained from ConoProbe surface
CT study of the surgical field and the markers acquired after surface scanning
CT study of the surgical field and the markers acquired before surface scanning
Pseudo-CT study obtained from reproSurf
Pseudo-CT study obtained after converting CTpreSurf data to only two values, tissue and air
3D dose distribution calculated from CTArtec
3D dose distribution calculated from CTConoProbe
3D dose distribution calculated from CTpreSurf
3D dose distribution calculated from CTreproSurf
3D dose distribution calculated from CTtissue&air
3D dose distribution calculated under the conventional assumption of water-equivalent tissues at the applicator end
Fiducial registration error
Intraoperative electron radiation therapy
Surface generated from CTtissue&air
Treatment planning system
Target registration error
Intraoperative electron radiation therapy (IOERT) refers to the delivery of a single-fraction, high-energy electron beam during surgery with the goal of promoting local cancer control [1, 2, 3]. The target volume (namely a post-resection tumour bed or the macroscopic residue after partial resection) is irradiated by using a specific applicator that pushes aside healthy tissues and collimates the electron beam generated by a linear accelerator (LINAC) [1, 4, 5].
IOERT dosimetry relies on the conventional assumption of water-equivalent tissues in both stopping and scattering power at the applicator end, which defines a flat irradiation surface. Previous literature reported small differences between prescribed doses under previous assumption and measured doses using in-vivo dosimetry in the case of breast IOERT scenarios [6, 7]. However, Costa et al. , by placing radiochromic films on irradiation surfaces, showed that clinical two-dimensional (2D) dose distributions in pelvic IOERT scenarios frequently differ from the expected ones. In a previous study, Costa et al.  simulated these IOERT scenarios with solid water slabs and a radiotherapy bolus, highlighting that the shape of the irradiation surface (namely step-like or curved surfaces) modified the dose distribution. Documentation of the actual scenario is relevant to the quality assurance of IOERT and the proper assessment of clinical results. However, the applicator position, the angle of beam incidence in relation to the patient’s anatomy and the shape of the irradiation surface are not available in IOERT records. Intraoperative computed tomography (CT) studies with the applicator placed on the tumour bed would allow the calculation of three-dimensional (3D) dose distributions of the actual scenario before irradiation since these images include tissue heterogeneities, surface irregularities of the irradiated volume and the air gap from the applicator end to the tumour bed. A portable CT scanner inside the operating room or a LINAC with on-board kV cone beam CT may be used for that purpose . However, a low number of IOERT interventions per week may not support the installation costs of in-room imaging devices . In addition, intraoperative CT studies with metal artefacts (for example, owing to shielding discs and surgical retractors) cannot be directly used for dose calculation since CT values are substantially altered.
An intermediate approach between the IOERT conventional assumption and the use of intraoperative CT images would involve 3D scanning the surface of the tumour bed and the inclusion of these data in the dose calculation. In a preliminary study , the authors evaluated four different technologies to scan surfaces, selecting a custom-made structured-light 3D scanner as the most appropriate for IOERT in terms of resolution, accuracy, acquisition time and cost. However, this statement was based on scanning a half body mannequin, not simulating an actual IOERT scenario and calculating dose distributions. Brudfors et al.  presented an open-source software for scanning surfaces with a tracked conoscopic holography sensor that could also be used in IOERT. This system has been previously evaluated for clinical usage (for example, to scan human cadaver kidneys [14, 15] and resection cavities during neurosurgeries ).
In this study we aimed to explore, for the first time, the use of surface scanning methods for 3D dose calculation in IOERT. Two different 3D scanning technologies were evaluated in a simulated IOERT scenario, a tracked conoscopic holography sensor and a structured-light 3D scanner.
In this section, we describe the 3D scanning systems assessed (section 3D scanning systems), the process of creating 3D images from surfaces (section Surface to pseudo-CT study for IOERT dose calculation) and the experiment designed to evaluate these non-contact devices for 3D dose calculation in IOERT (section Experiment). A simulated surgical field was scanned with both 3D scanning systems. 3D dose distributions obtained from CT studies of the surgical field (gold standard) were compared with those calculated under the conventional assumption of water-equivalent tissues at the applicator end and with those calculated from pseudo-CT studies based on surfaces.
3D scanning systems
Tracked conoscopic holography sensor
A conoscopic holography sensor measures distances to objects with an interferometric technique based on the double refraction of uniaxial crystals . The collinearity between the laser beam emitted and the cone of light returned from the scanned object enables measurements of sharp edges and inside narrow cavities. Distances are converted into 3D coordinates of the scanned surface by tracking the sensor .
The configuration selected in this study was that presented in : a ConoProbe Mark 10 sensor (Optimet, Optical Metrology Ltd) (working distance from 155 mm to 336 mm with the 250-mm objective lens as the origin, laser wavelength 655 nm, laser spot size 107 μm, laser power lower than 1 mW and weight 0.72 kg ) was tracked by a multi-camera optical tracking system OptiTrack (NaturalPoint, Inc.) with three optical cameras FLEX:V100R2 (data rate 50 Hz).
The data acquired with the tracked ConoProbe sensor (abbreviated hereafter to “ConoProbe”) is an unorganised point set that is converted to a triangle mesh by calculating its oriented surface normals and applying Poisson surface reconstruction  by means of open-source software MeshLab .
Structured-light 3D scanner
This 3D scanning technology is based on an incoherent light source that projects structured 2D patterns onto the scanned object, and a camera at a different viewpoint that records the patterns geometrically distorted by the object surface.
The specifications of the handheld 3D scanner selected in this study, Artec Eva (abbreviated hereafter to “Artec”), are: 3D resolution up to 0.5 mm, working distance from 0.4 m to 1 m, frame rate up to 16 fps and weight 0.85 kg . This solution includes another camera to capture colour texture. The software Artec Studio 10 manages the capturing process, data post-processing, 3D model (triangle mesh) creation and texture mapping.
Surface to pseudo-CT study for IOERT dose calculation
Dose distributions are calculated with a treatment planning system (TPS) specifically developed for IOERT (radiance, GMV) [22, 23] based on a Monte Carlo algorithm that takes account of the LINAC phase space  and tissue heterogeneities  (data obtained from CT studies after converting Hounsfield units (HU) values to physical density [10, 25]).
A surgical field was simulated with a large piece of beef (weight 6 kg) placed at the bottom of a plastic box (volume 28 × 39 × 57 cm), on the CT table, as shown in Fig. 2. This large container simulated the volume of a sterile field with surgical retractors that may hinder 3D scanning. Six metallic nipple markers (diameter 1 mm, SL-10, The Suremark Company) were placed on cuboids of different heights around the phantom (Fig. 2). These markers were used as landmarks for mapping surfaces and CT studies into a common coordinate space.
A CT study of the surgical field and the markers (CTpreSurf or CT study acquired before surface scanning) was acquired with the following parameters: voltage 120 kVp, exposure 37 ± 10 mAs (mean ± standard deviation), matrix size 512 × 512 × 961 and voxel size 0.729 × 0.729 × 0.500 mm.
Laser beam (ConoProbe) sweeping of the surgical field and the markers (Fig. 2, top right) (minimum signal-to-noise ratio set to 40%).
Surface scanning of the surgical field and the markers (Fig. 2, top right) with Artec.
Steps 2 and 3 were carried out from only one viewpoint to simulate the difficulty of moving around the patient in the operating room with our wired 3D scanners connected to both the power supply and the computer.
Step 1 was repeated with the same initial position of the patient table to obtain CTpostSurf (CT study acquired after surface scanning), which was used to check the stability of the whole setting (surgical field and markers) by comparing CTpostSurf and CTpreSurf. This verification was done because the surface irregularities of the simulated surgical field and the position of the markers could vary in steps 2 and 3 from the original setting shown in CTpreSurf due to the force of gravity, the movements of the CT table after CTpreSurf acquisition or a wrong attachment of the markers. The comparison was performed by superimposing CTpostSurf to CTpreSurf and calculating the root-mean-square error [RMSE] between the intensities from both studies without applying any alignment.
Steps 1 through 4 were repeated after adding 40 ml of pig blood with heparin (anticoagulant) to the surgical field (Fig. 2, bottom right) in order to simulate accumulation of biological fluid before irradiation.
Scanned surfaces were mapped to CTpreSurf to set a common coordinate space for comparisons in each configuration (with and without blood). The transformation matrix was calculated with a landmark rigid registration algorithm based on singular value decomposition  and the 3D coordinates of each marker (its centroid in the CT study and a point on its 3D surface) using 3D Slicer software  and SlicerIGT extension . Colour texture (visible light picture data from the surface) was essential for obtaining the marker location in the Artec surfaces since the markers were not properly identified in the triangle meshes.
IOERT dose distributions
Registered surfaces were converted to pseudo-CT studies of 1-mm isotropic voxel size by applying the method described in section Surface to pseudo-CT study for IOERT dose calculation (obtaining CTConoProbe and CTArtec for ConoProbe and Artec surfaces respectively). CTpreSurf studies were also resampled to that voxel size.
D_CTpreSurf (considered as the gold standard) was calculated with CTpreSurf.
D_water was calculated with CTpreSurf after selecting Water option in the TPS to obtain the dose distribution under the conventional assumption of water-equivalent tissues at the applicator end. D_water was obtained to assess the dose deviation when not including tissue heterogeneities, surface irregularities of the irradiated volume and the air gap from the applicator end to the tumour bed in the dose calculation.
D_CTtissue&air was calculated after converting CTpreSurf to another pseudo-CT study (CTtissue&air) that had only two different CT values (namely water [0 HU] and air [− 1000 HU]). D_CTtissue&air was used to assess the dose deviation when applying the same simplification of CTConoProbe and CTArtec in the dose calculation (namely not taking account of tissue heterogeneities but including surface irregularities of the irradiated volume and the air gap from the applicator end to the tumour bed). CTtissue&air was obtained by first segmenting the air in CTpreSurf using intensity thresholding method (maximum limit − 500 HU), and then setting those voxels to the CT value of air and the remaining voxels (specifically, tissue) to the CT value of water.
D_CTConoProbe and D_CTArtec were calculated with CTConoProbe and CTArtec respectively.
The output of the Monte Carlo algorithm was a 3D dose distribution in percentage where 100% corresponded to the maximum dose along the clinical axis measured in a water phantom for the selected applicator diameter and energy of the electron beam (in this study, 70 mm and 9 MeV respectively), the bevel angle 0°, and the source-to-surface distance set when conducting the measurements for modelling the LINAC.
D_water, D_CTtissue&air, D_CTConoProbe and D_CTArtec were compared with D_CTpreSurf using global normalisation , and a 3D gamma criteria of 3%/3 mm (thresholds accepted clinically) for dose values greater than 10% or 70% (to focus on high-dose regions) . 3D gamma analyses did not take account of dose differences in voxels not belonging to tissue in CTpreSurf.
Evaluation of “Surface to pseudo-CT study” conversion process
A source of error when using surface scanning methods for 3D IOERT dose calculation is the conversion from a 3D surface into a pseudo-CT study. The surface created to assess this process was reproSurf, which was created by first segmenting the voxels with the CT value of water in CTtissue&air (configuration without blood), then generating a watertight surface (both steps done with 3D Slicer software), and finally selecting a non-manifold surface of the surgical field like that obtained with ConoProbe and Artec (step done with Meshmixer software). The configuration chosen was that without blood since its surface was more irregular than that of the configuration with blood. A pseudo-CT study (CTreproSurf) was generated by applying the method detailed in section Surface to pseudo-CT study for IOERT dose calculation to reproSurf.
The evaluation consisted in comparing the 3D dose distribution calculated with CTreproSurf (D_CTreproSurf) and the reference 3D dose distribution D_CTtissue&air, which was calculated with the CT study used to generate reproSurf. CTreproSurf included both the error of generating the surface from CTtissue&air and the error when converting that surface into a pseudo-CT study. D_CTreproSurf was obtained with the same IOERT parameters detailed in section IOERT dose distributions and both dose distributions were compared using the same method described in section IOERT dose distributions. This comparison enabled us to assess the influence of the conversion from a 3D surface into a pseudo-CT study on the dose calculation.
Summary of CT/pseudo-CT studies and 3D dose distributions obtained in the experiment
CT acquisition before surface scanninga
Surface scan with Conoprobea
Surface scan with Arteca
CT acquisition after surface scanninga
Surface from CTpreSurf b
Generated 3D images
3D dose distributions
Acquisition data of both 3D scanning systems and registration results
Tracked ConoProbe sensor
Laser power (%)
Scanning time (s)
No. of points
Scanning time (s)
No. of vertexes
In each configuration, CTpostSurf was superimposed to CTpreSurf to check the stability of the whole setting (surgical field and markers). Table 2 shows the RMSE between both studies without applying any alignment. The RMSE was lower than 26 HU in both configurations. In addition, the only difference found when comparing CTpreSurf in both configurations was the blood at the base of the surgical field. The position and orientation difference between the four CT studies was neglectable (lower than 0.30 mm and 0.05° respectively). These values were obtained from the transformation matrix calculated when mapping CTpreSurf (configuration with blood) and CTpostSurf (both configurations) to CTpreSurf (configuration without blood) with the landmark rigid registration algorithm explained in section Surface registration and the 3D coordinates of the markers.
Percentage of voxels fulfilling the 3D gamma criteriaa
Dose > 10%
Dose > 70%
D_CTpreSurf v. D_water
D_CTpreSurf v. D_CTtissue&air
D_CTpreSurf v. D_CTConoProbe
D_CTpreSurf v. D_CTArtec
D_CTpreSurf v. D_water
D_CTpreSurf v. D_CTtissue&air
D_CTpreSurf v. D_CTConoProbe
D_CTpreSurf v. D_CTArtec
Two different 3D scanning technologies were evaluated in the context of 3D dose calculation in IOERT. Both devices can scan at a distance greater than 30 cm (minimum safety margin between unsterile personnel and the sterile field ), suitable for the IOERT theatre. Advantages of Artec over ConoProbe are a larger working distance (up to 1 m) and a shorter scanning time. Nevertheless, ConoProbe can obtain better measurements inside narrow cavities since its laser beam can access them more easily than the structured 2D patterns generated by the Artec projector. However, neither device could completely scan the deeper hole of the simulated surgical field (Fig. 3, red arrows), although ConoProbe was slightly better in this task.
CT studies (namely CTpreSurf and CTpostSurf) were used to check the stability of the whole setting (surgical field and markers) and to obtain reference surfaces/studies and dose distributions. The results of the comparison between CTpreSurf and CTpostSurf showed a similar scenario in each configuration (with and without blood). In addition, a similar IOERT case was simulated in both configurations except for blood accumulation. The air gap of the configuration without blood (maximum distance from the applicator end to the surface of the simulated tumour bed 33.5 mm) was within the range seen in limb sarcomas (air gaps up to 5 cm ). However, the IOERT scenario simulated is not realistic for breast IOERT, where the smoothness of the irradiation surface and a minimum air gap from the applicator end to the tumour bed can be ensured by other means .
The use of surface scanning for 3D dose calculation in IOERT is based on assuming water-equivalent tissues at the surface of the tumour bed, not at the applicator end as in the conventional assumption. The gamma pass rates obtained with D_CTtissue&air validated this assumption for our experiment (average value 99.7% for dose values greater than 10%) and therefore lower gamma pass rates are due to surface defects in the pseudo-CT studies. This statement may not be valid when bone is included in the irradiation volume, as for instance in a rectal cancer case where the tumour bed or the high-risk area is very close to the sacrum.
In the simulated IOERT scenario, the conventional assumption of water-equivalent tissues at the applicator end led to inaccurate dose distributions in both configurations (average gamma pass rate 39.9% for dose values greater than 10%). The configuration with blood presented a different surface of the simulated tumour bed (irradiation surface flatter than that without blood) and a different air gap compared with the configuration without blood. Results improved when surfaces were considered in the dose calculation although there were regions in the surgical field that did not completely match, such as the 14-mm-width hole in the configuration without blood (Fig. 3, axial view of CTConoProbe and CTArtec). This hole could not be properly scanned since a small piece of tissue protruded and covered part of the hole. On the other hand, both devices could scan a surgical field with blood.
The match between CTtissue&air and the pseudo-CT studies CTConoProbe and CTArtec in the target volume was better in the case of Artec compared with ConoProbe (Fig. 3). Similarly, gamma pass rates with Artec were better than those with ConoProbe (average values 92.9% and 88.5% for dose values greater than 10% respectively). Possible sources of error in the ConoProbe measurements were the surface reconstruction from noisy points (PSE around 2 mm, Table 2) and larger error in the location of the markers compared with Artec (fiducial registration error [FRE] Table 2).
The pseudo-CT study CTreproSurf fitted the CT study used to generate reproSurf (namely CTtissue&air). These CT studies were utilised to assess the influence of the conversion from a 3D surface into a pseudo-CT study on the dose calculation. The results showed that the influence was negligible for the configuration without blood (the worst case of those presented in this study). Therefore, the error in the dose distributions D_CTConoProbe and D_CTArtec was caused by the inaccuracies of the scanned surfaces, and the registration between the 3D surfaces and CTpreSurf. 3D scanning from multiple viewpoints could improve the acquired data and thus increase the gamma pass rates, although it would further alter the surgical workflow.
Other studies have proposed in-vivo dosimetry with micro metal oxide semiconductor field-effect transistors, radiochromic films or thermoluminescence radiation detectors, which enables point and 2D dose measurements in IOERT [8, 36, 37]. Our approach focuses only on obtaining 3D dose distributions and some issues could still be improved. Applicator position and angle of beam incidence (applicator rotation) in relation to the patient’s anatomy are required for dose calculations. After scanning the surface of the tumour bed, the applicator position and rotation could be collected with an optical tracking system , and then combined with that scanned surface to also obtain the air gap from the applicator end to the surface of the irradiated volume. The applicator data would be in the same coordinate space as the scanned surface when using ConoProbe (same optical tracking system) while a calibration tool visible to the tracking system and the Artec scanner should be added to the scenario to map the applicator pose to the scanned surface. Finally, it would be necessary to reduce the data post-processing steps detailed in section Surface to pseudo-CT study for IOERT dose calculation and section Results in order to obtain a real-time implementation in the clinical practice, although this is not crucial since the main interest was to obtain the actual dose distribution for quality assurance of IOERT, but not immediately before irradiation.
This is the first study that explores the use of surface scanning for 3D dose calculation in IOERT. In the simulated IOERT scenario, the conventional assumption of water-equivalent tissues at the applicator end led to inaccurate dose distributions. More accurate dose distributions were obtained when considering surfaces in the dose calculation of our simulated surgical field. The structured-light 3D scanner provided the best results in terms of dose distributions. Surface scanning is a promising method that could be easily included in the clinical practice, since the acquisition process is simple and does not require much time, and the calculated dose is comparable to the one that could be obtained with CT imaging. The findings obtained in this specific experimental setup warrant further research on surface scanning in the IOERT context owing to the clinical interest of improving the documentation of the actual IOERT scenario.
The authors extend their gratitude to GMV SA (Madrid, Spain), Alexandra de Francisco López, Yolanda Sierra Palomares, and medical physicists and radiotherapy technicians at Hospital General Universitario Gregorio Marañón (Madrid, Spain) for their valuable technical support.
This study was supported by Ministerio de Ciencia, Innovación y Universidades [grant number TEC2013–48251-C2–1-R]; by Ministerio de Ciencia, Innovación y Universidades,, Instituto de Salud Carlos III and European Regional Development Fund (FEDER) Funds from the European Commission, “A way of making Europe” [grant numbers DTS14/00192, PI15/02121]; and by Comunidad de Madrid [grant number TOPUS-CM S2013/MIT-3024]. The CNIC is supported by the Ministerio de Ciencia, Innovación y Universidades and the Pro CNIC Foundation, and is a Severo Ochoa Center of Excellence (SEV-2015-0505).
Study sponsors had no involvement in the study design, in the collection, analysis and interpretation of data; in the writing of the manuscript; and in the decision to submit the manuscript for publication.
Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
VGV, BSL, FAC and JP contributed in the study design. VGV and BSL participated in data collection. VGV and BSL contributed to data analysis. VGV, BSL, FAC, JJV, MD and JP provided data interpretation. VGV wrote the draft manuscript. VGV, BSL, FAC, JJV, MD and JP revised the manuscript. JP supervised this study. All authors read and approved the final manuscript.
Ethics approval and consent to participate
Consent for publication
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
- 2.Nag S, Willett CG, Gunderson LL, Harrison LB, Calvo FA, Biggs P. IORT with electron-beam, high-dose-rate brachytherapy or low-kV/electronic brachytherapy: methodological comparisons. In: Gunderson LL, Willett CG, Calvo FA, Harrison LB, editors. Intraoperative irradiation: techniques and results. New York: Humana Press; 2011. p. 99–115.CrossRefGoogle Scholar
- 4.Biggs P, Willett CG, Rutten H, Ciocca M, Gunderson LL, Calvo FA. Intraoperative electron beam irradiation: physics and techniques. In: Gunderson LL, Willett CG, Calvo FA, Harrison LB, editors. Intraoperative irradiation: techniques and results. New York: Humana Press; 2011. p. 51–72.CrossRefGoogle Scholar
- 18.OEM Manual for OPTIMET’S Mark10/10HD (P/N 3J06009, Rev 2): Optimet Optical Metrology Ltd. http://www.optimet.com. Accessed 31 Nov 2017.
- 19.Kazhdan M, Bolitho M, Hoppe H. Poisson surface reconstruction. Symp Geom Process. 2006:61–70.Google Scholar
- 20.Cignoni P, Callieri M, Corsini M, Dellepiane M, Ganovelli F, Ranzuglia G. MeshLab: an open-source mesh processing tool. Sixth Eurographics Italian Chapter Conference. 2008:129-36.Google Scholar
- 21.Professional 3D scanning solutions (Artec scanners brochure, 001-05/2017-ENG): Artec 3D. https://www.artec3d.com. Accessed 31 Nov 2017.
- 22.Pascau J, Santos Miranda JA, Calvo FA, Bouché A, Morillo V, González-San Segundo C, et al. An innovative tool for intraoperative electron beam radiotherapy simulation and planning: description and initial evaluation by radiation oncologists. Int J Radiat Oncol Biol Phys. 2012;83:e287–e95.CrossRefPubMedGoogle Scholar
- 23.Valdivieso-Casique MF, Rodríguez R, Rodríguez-Bescós S, Lardíes D, Guerra P, Ledesma MJ, et al. RADIANCE – a planning software for intra-operative radiation therapy. Transl Cancer Res. 2015;4:196–209.Google Scholar
- 26.Autodesk Meshmixer. http://www.meshmixer.com. Accessed 31 Nov 2017.
- 27.Aitkenhead AH. Mesh voxelisation. https://es.mathworks.com/matlabcentral/fileexchange/27390-mesh-voxelisation. Accessed 1 Feb 2017.
- 28.Patil S, Ravi B. Voxel-based representation, display and thickness analysis of intricate shapes. Proceedings of the 9th international conference on computer aided design and. Computer Graphics. 2005.Google Scholar
- 33.Phillips N. Berry & Kohn's operating room technique. St. Louis: Elsevier Health Sciences; 2016.Google Scholar
- 35.Takanen S, Gambirasio A, Gritti G, Källi M, Andreoli S, Fortunato M, et al. Breast cancer electron intraoperative radiotherapy: assessment of preoperative selection factors from a retrospective analysis of 758 patients and review of literature. Breast Cancer Res Treat. 2017;165:261–71.CrossRefPubMedGoogle Scholar
- 37.López-Tarjuelo J, Bouché-Babiloni A, Morillo-Macías V, de Marco-Blancas N, Santos-Serra A, Quirós-Higueras JD, et al. In vivo dosimetry in intraoperative electron radiotherapy: microMOSFETs, radiochromic films and a general-purpose linac. Strahlenther Onkol. 2014;190:1060–5.CrossRefPubMedGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.