-
Sharp dose profiles for high precision proton therapy using focused proton beams
Authors:
Fardous Reaz,
Kyrre Ness Sjobak,
Eirik Malinen,
Nina Frederike Jeppesen Edin,
Erik Adli
Abstract:
Proton beam therapy has been developed to irradiate the tumor with higher precision and dose conformity compared to conventional X-ray irradiation. The dose conformity of this treatment modality may be further improved if narrower proton beams are used. Still, this is limited by multiple Coulomb scattering of protons through tissue. The primary aim of this work was to develop techniques to produce…
▽ More
Proton beam therapy has been developed to irradiate the tumor with higher precision and dose conformity compared to conventional X-ray irradiation. The dose conformity of this treatment modality may be further improved if narrower proton beams are used. Still, this is limited by multiple Coulomb scattering of protons through tissue. The primary aim of this work was to develop techniques to produce narrow proton beams and investigate the resulting dose profiles. We introduced and assessed three different proton beam sha** techniques: 1) metal collimators (100/150~MeV), 2) focusing of conventional- (100/150~MeV), and 3) focusing of high-energy (350~MeV, shoot-through) proton beams. Focusing was governed by the initial value of the Twiss parameter $α$~($α_0$), and can be implemented with magnetic particle accelerator optics. The dose distributions in water were calculated by Monte Carlo simulations using Geant4, and evaluated by target to surface dose ratio (TSDR) in addition to the transverse beam size~($σ_T$) at the target. The target was defined as the location of the Bragg peak or the focal point. The different techniques showed greatly differing dose profiles, where focusing gave pronouncedly higher relative target dose and efficient use of primary protons. Metal collimators with radii < 2~mm gave low TSDRs (<~0.7) and large $σ_T$(>~3.6~mm). In contrast, a focused beam of conventional (150~MeV) energy produced a very high TSDR (>~80) with similar $σ_T$ as a collimated beam. High-energy focused beams were able to produce TSDRs > 100 and $σ_T$ around 1.5~mm. From this study, it appears very attractive to implement magnetically focused proton beams in radiotherapy of small lesions or tumors in close vicinity to healthy organs at risk. This can also lead to a paradigm change in spatially fractionated radiotherapy.
△ Less
Submitted 2 September, 2022;
originally announced September 2022.
-
2D map** of radiation dose and clonogenic survival for accurate assessment of in vitro X-ray GRID irradiation effects
Authors:
D. Arous,
J. L. Lie,
B. V. Håland,
M. Børsting,
N. F. J. Edin,
E. Malinen
Abstract:
Spatially fractionated radiation therapy (SFRT or GRID) is an approach to deliver high local radiation doses in an 'on-off' pattern. To better appraise the radiobiological effects from GRID, a framework to link local radiation dose to clonogenic survival needs to be developed. A549 (lung) cancer cells were irradiated in T25 cm$^2$ flasks using 220 kV X-rays with an open field or through a tungsten…
▽ More
Spatially fractionated radiation therapy (SFRT or GRID) is an approach to deliver high local radiation doses in an 'on-off' pattern. To better appraise the radiobiological effects from GRID, a framework to link local radiation dose to clonogenic survival needs to be developed. A549 (lung) cancer cells were irradiated in T25 cm$^2$ flasks using 220 kV X-rays with an open field or through a tungsten GRID collimator with periodical 5 mm openings and 10 mm blockings. Delivered nominal doses were 2, 5, and 10 Gy. A novel approach for image segmentation was used to locate the centroid of surviving colonies in scanned images of the cell flasks. GafchromicTM film dosimetry (GFD) and FLUKA Monte Carlo (MC) simulations were employed to map the dose distribution in the flasks at each surviving colony centroid. Fitting the linear-quadratic (LQ) function to clonogenic survival data for open field irradiation, the expected survival level at a given dose level was calculated. The expected survival level was then mapped together with the observed levels in the GRID-irradiated flasks. GFD and FLUKA MC gave similar dose distributions, with a mean peak-to-valley dose ratio of about 5. LQ-parameters for open field irradiation gave $α= 0.16 \pm 0.04$ Gy$^{-1}$ and $β= 0.001 \pm 0.004$ Gy$^{-2}$. Using the image segmentation method, the mean absolute percentage deviation between observed and predicted survival in the (peak; valley) dose regions was (8; 10) %, (4; 41) %, and (3; 138) % for 2, 5 and 10 Gy, respectively. In conclusion, a framework for map** of surviving colonies following GRID irradiation together with predicted survival levels from homogeneous irradiation was presented. For the given cell line, our findings indicate that GRID irradiation, especially at high peak doses, causes reduced survival compared to an open field configuration.
△ Less
Submitted 30 May, 2022;
originally announced May 2022.
-
Principal component-based image segmentation: a new approach to outline in vitro cell colonies
Authors:
Delmon Arous,
Stefan Schrunner,
Ingunn Hanson,
Nina F. J. Edin,
Eirik Malinen
Abstract:
The in vitro clonogenic assay is a technique to study the ability of a cell to form a colony in a culture dish. By optical imaging, dishes with stained colonies can be scanned and assessed digitally. Identification, segmentation and counting of stained colonies play a vital part in high-throughput screening and quantitative assessment of biological assays. Image processing of such pictured/scanned…
▽ More
The in vitro clonogenic assay is a technique to study the ability of a cell to form a colony in a culture dish. By optical imaging, dishes with stained colonies can be scanned and assessed digitally. Identification, segmentation and counting of stained colonies play a vital part in high-throughput screening and quantitative assessment of biological assays. Image processing of such pictured/scanned assays can be affected by image/scan acquisition artifacts like background noise and spatially varying illumination, and contaminants in the suspension medium. Although existing approaches tackle these issues, the segmentation quality requires further improvement, particularly on noisy and low contrast images. In this work, we present an objective and versatile machine learning procedure to amend these issues by characterizing, extracting and segmenting inquired colonies using principal component analysis, k-means clustering and a modified watershed segmentation algorithm. The intention is to automatically identify visible colonies through spatial texture assessment and accordingly discriminate them from background in preparation for successive segmentation. The proposed segmentation algorithm yielded a similar quality as manual counting by human observers. High F1 scores (>0.9) and low root-mean-square errors (around 14%) underlined good agreement with ground truth data. Moreover, it outperformed a recent state-of-the-art method. The methodology will be an important tool in future cancer research applications.
△ Less
Submitted 10 March, 2021;
originally announced March 2021.
-
Deep learning for automatic tumour segmentation in PET/CT images of patients with head and neck cancers
Authors:
Yngve Mardal Moe,
Aurora Rosvoll Groendahl,
Martine Mulstad,
Oliver Tomic,
Ulf Indahl,
Einar Dale,
Eirik Malinen,
Cecilia Marie Futsaether
Abstract:
An automatic segmentation algorithm for delineation of the gross tumour volume and pathologic lymph nodes of head and neck cancers in PET/CT images is described. The proposed algorithm is based on a convolutional neural network using the U-Net architecture. Several model hyperparameters were explored and the model performance in terms of the Dice similarity coefficient was validated on images from…
▽ More
An automatic segmentation algorithm for delineation of the gross tumour volume and pathologic lymph nodes of head and neck cancers in PET/CT images is described. The proposed algorithm is based on a convolutional neural network using the U-Net architecture. Several model hyperparameters were explored and the model performance in terms of the Dice similarity coefficient was validated on images from 15 patients. A separate test set consisting of images from 40 patients was used to assess the generalisability of the algorithm. The performance on the test set showed close-to-oncologist level delineations as measured by the Dice coefficient (CT: $0.65 \pm 0.17$, PET: $0.71 \pm 0.12$, PET/CT: $0.75 \pm 0.12$).
△ Less
Submitted 2 August, 2019;
originally announced August 2019.
