An Explainable MRI-Radiomic Quantum Neural Network to Differentiate Between Large Brain Metastases and High-Grade Glioma Using Quantum Annealing for Feature Selection.

Solitary large brain metastases (LBM) and high-grade gliomas (HGG) are sometimes hard to differentiate on MRI. The management differs significantly between these two entities, and non-invasive methods that help differentiate between them are eagerly needed to avoid potentially morbid biopsies and surgical procedures. We explore herein the performance and interpretability of an MRI-radiomics variational quantum neural network (QNN) using a quantum-annealing mutual-information (MI) feature selection approach. We retrospectively included 423 patients with HGG and LBM (> 2 cm) who had a contrast-enhanced T1-weighted (CE-T1) MRI between 2012 and 2019. After exclusion, 72 HGG and 129 LBM were kept. Tumors were manually segmented, and a 5-mm peri-tumoral ring was created. MRI images were pre-processed, and 1813 radiomic features were extracted. A set of best features based on MI was selected. MI and conditional-MI were embedded into a quadratic unconstrained binary optimization (QUBO) formulation that was mapped to an Ising-model and submitted to D'Wave's quantum annealer to solve for the best combination of 10 features. The 10 selected features were embedded into a 2-qubits QNN using PennyLane library. The model was evaluated for balanced-accuracy (bACC) and area under the receiver operating characteristic curve (ROC-AUC) on the test set. The model performance was benchmarked against two classical models: dense neural networks (DNN) and extreme gradient boosting (XGB). Shapley values were calculated to interpret sample-wise predictions on the test set. The best 10-feature combination included 6 tumor and 4 ring features. For QNN, DNN, and XGB, respectively, training ROC-AUC was 0.86, 0.95, and 0.94; test ROC-AUC was 0.76, 0.75, and 0.79; and test bACC was 0.74, 0.73, and 0.72. The two most influential features were tumor Laplacian-of-Gaussian-GLRLM-Entropy and sphericity. We developed an accurate interpretable QNN model with quantum-informed feature selection to differentiate between LBM and HGG on CE-T1 brain MRI. The model performance is comparable to state-of-the-art classical models.

Time-Dependent Density-Functional Theory for Determining the Electron-Capture Cross Section for Protons Impacting on Atoms and Molecules.

The use of the Time-Dependent Density-Functional Theory (TDDFT) has increased in the atomic collision field. Calculating the electron-capture cross section (ECCS) for protons is an important question in hadrontherapy and plasma physics, among other areas. In previous studies, it was shown that the approach based on the Local Density Approximation (LDA) fails in the 1-50 keV region, requiring the use of the Optimized Effective Potential (OEP) method. In this work, the ECCS values for 1-50 keV protons impacting on isolated hydrogen, carbon, nitrogen, oxygen, and nitrogenous atoms were determined using the TDDFT. It is shown that adding the Self Interaction Correction to the LDA (LDA-Sic) allows obtaining results close to those provided by the OEP and experiments, with the advantage that the LDA-Sic consumes less computational time. In addition, it was demonstrated that it is imperative to include the spin correction for the specific helium and oxygen cases, in order to get good results for the ECCS using the TDDFT. Theoretical results obtained in this work show excellent agreement with experimental values.

Prediction of DNA rejoining kinetics and cell survival after proton irradiation for V79 cells using Geant4-DNA.

PURPOSE: Track structure Monte Carlo (MC) codes have achieved successful outcomes in the quantitative investigation of radiation-induced initial DNA damage. The aim of the present study is to extend a Geant4-DNA radiobiological application by incorporating a feature allowing for the prediction of DNA rejoining kinetics and corresponding cell surviving fraction along time after irradiation, for a Chinese hamster V79 cell line, which is one of the most popular and widely investigated cell lines in radiobiology. METHODS: We implemented the Two-Lesion Kinetics (TLK) model, originally proposed by Stewart, which allows for simulations to calculate residual DNA damage and surviving fraction along time via the number of initial DNA damage and its complexity as inputs. RESULTS: By optimizing the model parameters of the TLK model in accordance to the experimental data on V79, we were able to predict both DNA rejoining kinetics at low linear energy transfers (LET) and cell surviving fraction. CONCLUSION: This is the first study to demonstrate the implementation of both the cell surviving fraction and the DNA rejoining kinetics with the estimated initial DNA damage, in a realistic cell geometrical model simulated by full track structure MC simulations at DNA level and for various LET. These simulation and model make the link between mechanistic physical/chemical damage processes and these two specific biological endpoints.

Electron tracks simulation in water: Performance comparison between GPU CPU and the EUMED grid installation.

PURPOSE: We explored different technologies to minimize simulation time of the Monte-Carlo method for track generation following the Geant4-DNA processes for electrons in water. METHODS: A GPU software tool is developed for electron track simulations. A similar CPU version is also developed using the same collision models. CPU simulations were carried out on a single user desktop computer and on the computing grid France Grilles using 10 and 100 computing nodes. Computing time results for CPU, GPU, and grid simulations are compared with those using Geant4-DNA processes. RESULTS: The CPU simulations better performs when the number of electrons is less than 104 with 100 eV initial energy, this number decreases as the energy increases. The GPU simulations gives better results when the number of electrons is more than 104 with initial energy of 100 eV, this number decreases to 103 for electrons with 10KeV and increases back with higher energy. The use of the grid introduces an additional queuing time which slows down the overall simulation performance. Thus, the Grid gives better performance when the number of electrons is over 105 with initial energy of 10KeV, and this number decreases as the energy increases. CONCLUSIONS: The CPU is best suited for small numbers of primary incident electrons. The GPU is best suited when the number of primary incident particles occupies sufficient resources on GPU card in order to get an important computing power. The grid is best suited for simulations with high number of primary incident electrons with high initial energy.

A Geant4-DNA Evaluation of Radiation-Induced DNA Damage on a Human Fibroblast.

Accurately modeling the radiobiological mechanisms responsible for the induction of DNA damage remains a major scientific challenge, particularly for understanding the effects of low doses of ionizing radiation on living beings, such as the induction of carcinogenesis. A computational approach based on the Monte Carlo technique to simulate track structures in a biological medium is currently the most reliable method for calculating the early effects induced by ionizing radiation on DNA, the primary cellular target of such effects. The Geant4-DNA Monte Carlo toolkit can simulate not only the physical, but also the physico-chemical and chemical stages of water radiolysis. These stages can be combined with simplified geometric models of biological targets, such as DNA, to assess direct and indirect early DNA damage. In this study, DNA damage induced in a human fibroblast cell was evaluated using Geant4-DNA as a function of incident particle type (gammas, protons, and alphas) and energy. The resulting double-strand break yields as a function of linear energy transfer closely reproduced recent experimental data. Other quantities, such as fragment length distribution, scavengeable damage fraction, and time evolution of damage within an analytical repair model also supported the plausibility of predicting DNA damage using Geant4-DNA.The complete simulation chain application "molecularDNA", an example for users of Geant4-DNA, will soon be distributed through Geant4.

Monte Carlo Simulation of SARS-CoV-2 Radiation-Induced Inactivation for Vaccine Development.

Immunization with an inactivated virus is one of the strategies currently being tested towards developing a SARS-CoV-2 vaccine. One of the methods used to inactivate viruses is exposure to high doses of ionizing radiation to damage their nucleic acids. While gamma (γ) rays effectively induce lesions in the RNA, envelope proteins are also highly damaged in the process. This in turn may alter their antigenic properties, affecting their capacity to induce an adaptive immune response able to confer effective protection. Here, we modeled the effect of sparsely and densely ionizing radiation on SARS-CoV-2 using the Monte Carlo toolkit Geant4-DNA. With a realistic 3D target virus model, we calculated the expected number of lesions in the spike and membrane proteins, as well as in the viral RNA. Our findings showed that γ rays produced significant spike protein damage, but densely ionizing charged particles induced less membrane damage for the same level of RNA lesions, because a single ion traversal through the nuclear envelope was sufficient to inactivate the virus. We propose that accelerated charged particles produce inactivated viruses with little structural damage to envelope proteins, thereby representing a new and effective tool for developing vaccines against SARS-CoV-2 and other enveloped viruses.

Determination of fast neutron RBE using a fully mechanistic computational model.

This work presents a model previously developed for estimating relative biological effectiveness (RBE) associated with high-LET particles. It is based on the combination of Monte Carlo simulations of particle interactions when traversing an atomic resolution DNA geometrical model. In addition, the model emulates the induction of lethal damage from the interaction of two sublethal lesions, taken as double-strand breaks. The Geant4-DNA package was used for simulations with liquid water as the transport medium. The RBE of neutron beams with energies ranging from 0.1 MeV up to 14 MeV was studied. The model succeeded in reproducing the general behavior of RBE as a function of neutron energy, including the RBE peak reported by experiments at approximately 0.4 MeV. Furthermore, the results of the model agree rather well with some experimental works. However, our results underestimate RBE for neutron energies above approximately 5 MeV due to the current limitations of Geant4-DNA for the tracking of heavy ions below 0.5 MeV/u.

Assessment of Radio-Induced Damage in Endothelial Cells Irradiated with 40 kVp, 220 kVp, and 4 MV X-rays by Means of Micro and Nanodosimetric Calculations.

The objective of this work was to study the differences in terms of early biological effects that might exist between different X-rays energies by using a mechanistic approach. To this end, radiobiological experiments exposing cell monolayers to three X-ray energies were performed in order to assess the yields of early DNA damage, in particular of double-strand breaks (DSBs). The simulation of these irradiations was set in order to understand the differences in the obtained experimental results. Hence, simulated results in terms of microdosimetric spectra and early DSB induction were analyzed and compared to the experimental data. Human umbilical vein endothelial cells (HUVECs) were irradiated with 40, 220 kVp, and 4 MV X-rays. The Geant4 Monte Carlo simulation toolkit and its extension Geant4-DNA were used for the simulations. Microdosimetric calculations aiming to determine possible differences in the variability of the energy absorbed by the irradiated cell population for those photon spectra were performed on 10,000 endothelial cell nuclei representing a cell monolayer. Nanodosimetric simulations were also carried out using a computation chain that allowed the simulation of physical, physico-chemical, and chemical stages on a single realistic endothelial cell nucleus model including both heterochromatin and euchromatin. DNA damage was scored in terms of yields of prompt DSBs per Gray (Gy) and per giga (109) base pair (Gbp) and DSB complexity was derived in order to be compared to experimental data expressed as numbers of histone variant H2AX (γ-H2AX) foci per cell. The calculated microdosimetric spread in the irradiated cell population was similar when comparing between 40 and 220 kVp X-rays and higher when comparing with 4 MV X-rays. Simulated yields of induced DSB/Gy/Gbp were found to be equivalent to those for 40 and 220 kVp but larger than those for 4 MV, resulting in a relative biological effectiveness (RBE) of 1.3. Additionally, DSB complexity was similar between the considered photon spectra. Simulated results were in good agreement with experimental data obtained by IRSN (Institut de radioprotection et de sûreté nucléaire) radiobiologists. Despite differences in photon energy, few differences were observed when comparing between 40 and 220 kVp X-rays in microdosimetric and nanodosimetric calculations. Nevertheless, variations were observed when comparing between 40/220 kVp and 4 MV X-rays. Thanks to the simulation results, these variations were able to be explained by the differences in the production of secondary electrons with energies below 10 keV.

Computational approach to determine the relative biological effectiveness of fast neutrons using the Geant4-DNA toolkit and a DNA atomic model from the Protein Data Bank.

This study proposes an innovative approach to estimate relative biological effectiveness (RBE) of fast neutrons using the Geant4 toolkit. The Geant4-DNA version cannot track heavy ions below 0.5 MeV/nucleon. In order to explore the impact of this issue, secondary particles are simulated instead of the primary low-energy neutrons. The Evaluated Nuclear Data File library is used to determine the cross sections for the elastic and inelastic interactions of neutrons with water and to find the contribution of each secondary particle spectrum. Two strategies are investigated in order to find the best possible approach and results. The first one takes into account only light particles, protons produced from elastic scattering, and α particles from inelastic scattering. Geantino particles are shot instead of heavy ions; hence all heavy ions are considered in the simulations, though their physical effects on DNA not. The second strategy takes into account all the heavy and light ions, although heavy ions cannot be tracked down to very low energies (E<0.5 MeV/nucleon). Our model is based on the combination of an atomic resolution DNA geometrical model and a Monte Carlo simulation toolkit for tracking particles. The atomic coordinates of the DNA double helix are extracted from the Protein Data Bank. Since secondary particle spectra are used instead of simulating the interaction of neutrons explicitly, this method reduces the computation times dramatically. Double-strand break induction is used as the end point for the estimation of the RBE of fast neutrons. ^{60}Co Î³ rays are used as the reference radiation quality. Both strategies succeed in reproducing the behavior of the RBE_{max} as a function of the incident neutron energy ranging from 0.1 to 14 MeV, including the position of its peak. A comparison of the behavior of the two strategies shows that for neutrons with energies less than 0.7 MeV, the effect of heavy ions would not be very significant, but above 0.7 MeV, heavy ions have an important role in neutron RBE.

Evaluation of early radiation DNA damage in a fractal cell nucleus model using Geant4-DNA.

The advancement of multidisciplinary research fields dealing with ionising radiation induced biological damage - radiobiology, radiation physics, radiation protection and, in particular, medical physics - requires a clear mechanistic understanding of how cellular damage is induced by ionising radiation. Monte Carlo (MC) simulations provide a promising approach for the mechanistic simulation of radiation transport and radiation chemistry, towards the in silico simulation of early biological damage. We have recently developed a fully integrated MC simulation that calculates early single strand breaks (SSBs) and double strand breaks (DSBs) in a fractal chromatin based human cell nucleus model. The results of this simulation are almost equivalent to past MC simulations when considering direct/indirect strand break fraction, DSB yields and fragment distribution. The simulation results agree with experimental data on DSB yields within 13.6% on average and fragment distributions agree within an average of 34.8%.

Accounting for radiation-induced indirect damage on DNA with the Geant 4-DNA code.

The use of Monte Carlo (MC) simulations remains a powerful tool to study the biological effects induced by ionizing radiation on living beings. Several MC codes are commonly used in research fields such as nanodosimetry, radiotherapy, radiation protection, and space radiation. This work presents an enhancement of an existing model [1] for radiobiological purposes, to account for the indirect DNA damage induced by ionizing particles. The Geant4-DNA simulation toolkit was used to simulate the physical, pre-chemical, and chemical stages of early DNA damage induced by protons and α-particles. Liquid water was used as the medium for simulations. Two phase-space files were generated, one containing the energy deposition events and another with the position of chemical species produced by water radiolysis from 0.1 ps up to 1 ns. These files were used as input in the radiobiological code that contains the genetic material model with atomic resolution, consisting of several copies of 30 nm chromatin fibers. The B-DNA configuration was used. This work focused on the indirect damage produced by the hydroxyl radical (OH) attack on the sugar-phosphate group. The approach followed to account for the indirect DNA damage was the same as those used by other radiobiological codes [2,3]. The critical parameter considered here was the reaction radius, which was calculated from the Smoluchowski's diffusion equation. Single, double, and total strand break yields produced by direct, indirect, and mixed mechanisms are reported. The obtained results are consistent with experimental and calculation data sets published in the literature.

Numerical insight into the Dual Radiation Action Theory.

This work studies the first and second order mechanisms for the induction of lethal lesions in DNA after irradiation with protons and α-particles. The purpose is to numerically study the mechanisms behind the Dual Radiation Action Theory (DRAT) for these heavy particles. A genetic material geometrical model with atomic resolution is used. It accounts for the explicit position of 5.47 × 109 base pairs, organized up to the chromatin level. The GEANT4-DNA Monte Carlo code was employed to simulate the interaction of these ions with the genetic material model. The number of lethal lesions induced by one- and two-track mechanisms was determined as a function of dose. Values of the α/ß ratio were estimated as well as corresponding relative biological effectiveness (RBE). The number of lethal lesions produced by one-track and two-track mechanisms depends on the dose and squared dose, respectively, as predicted by the DRAT. RBE values consistent with experimental results were found, at least for LET below ∼100 keV/µm. Double strand break spatial distributions are qualitatively analyzed. According to this work, the α parameter determined from cellular surviving curves depends on both the physical α and ß parameters introduced here, and on the specific energy deposited by a single track into the region of interest. We found an increment of the ß parameter with LET, yet at a slower rate than α so that the α/ß ratio increases with LET. In addition, we observed and explained the saturation of the α parameter as the dose increases above ∼6 Gy.

A study on the physicochemical parameters for Penicillium expansum growth and patulin production: effect of temperature, pH, and water activity.

Penicillium expansum is among the most ubiquitous fungi disseminated worldwide, that could threaten the fruit sector by secreting patulin, a toxic secondary metabolite. Nevertheless, we lack sufficient data regarding the growth and the toxigenesis conditions of this species. This work enables a clear differentiation between the favorable conditions to the P. expansum growth and those promising for patulin production. A mathematical model allowing the estimation of the P. expansum growth rate according to temperature, a W, and pH, was also developed. An optimal growth rate of 0.92 cm/day was predicted at 24°C with pH level of 5.1 and high a W level of 0.99. The model's predictive capability was tested successfully on artificial contaminated apples. This model could be exploited by apple growers and the industrialists of fruit juices in order to predict the development of P. expansum during storage and apple processing.

Analysis of the EBT3 Gafchromic film irradiated with 6 MV photons and 6 MeV electrons using reflective mode scanners.

We explore in our study the effects of electrons and X-rays irradiations on the newest version of the Gafchromic EBT3 film. Experiments are performed using the Varian "TrueBeam 1.6" medical accelerator delivering 6 MV X-ray photons and 6 MeV electron beams as desired. The main interest is to compare the responses of EBT3 films exposed to two separate beams of electrons and photons, for radiation doses ranging up to 500 cGy. The analysis is done on a flatbed EPSON 10000 XL scanner and cross checked on a HP Scanjet 4850 scanner. Both scanners are used in reflection mode taking into account landscape and portrait scanning positions. After thorough verifications, the reflective scanning method can be used on EBT3 as an economic alternative to the transmission method which was also one of the goals of this study. A comparison is also done between single scan configuration including all samples in a single A4 (HP) or A3 (EPSON) format area and multiple scan procedure where each sample is scanned separately on its own. The images analyses are done using the ImageJ software. Results show significant influence of the scanning configuration but no significant differences between electron and photon irradiations for both single and multiple scan configurations. In conclusion, the film provides a reliable relative dose measurement method for electrons and photons irradiations in the medical field applications.

Simulation of DNA damage clustering after proton irradiation using an adapted DBSCAN algorithm.

In this work the "Density Based Spatial Clustering of Applications with Noise" (DBSCAN) algorithm was adapted to early stage DNA damage clustering calculations. The resulting algorithm takes into account the distribution of energy deposit induced by ionising particles and a damage probability function that depends on the total energy deposit amount. Proton track simulations were carried out in small micrometric volumes representing small DNA containments. The algorithm was used to determine the damage concentration clusters and thus to deduce the DSB/SSB ratios created by protons between 500keV and 50MeV. The obtained results are compared to other calculations and to available experimental data of fibroblast and plasmid cells irradiations, both extracted from literature.

Methylene blue dye, an accurate dye for sentinel lymph node identification in early breast cancer.

PURPOSE: The aim of this prospective study was to analyze the safety of methylene blue dye (MBD) and compare its efficacy with that of isotopic mapping for sentinel lymph node (SLN) identification in breast cancer. PATIENTS AND METHODS: The SLN procedure, involving isotopic mapping and MBD (subareolar intraparenchymal injections of 2 mL, 10 mg/mL), was performed on 100 patients with early breast cancer. RESULTS: The procedure was safe with a success rate of 99%; SLNs were, respectively, found in 65% by MBD, in 73% by lymphoscintigraphy and in 94% by gamma-probe. Out of 40 metastatic SLNs, 37 were "hot" and 32 stained. Digital examination allowed the detection of 2 additional metastatic LNs. CONCLUSION: MBD is safe and combination mapping associated with digital examination is the superior method. Modification of the procedure, favouring injections of dilute MBD (4 mL, 1.25 mg/mL) increases MBD efficiency (90%) and maintains low rates of complications.