arXiv subjects

Florian Mentzel

Publications and source records attributed to Florian Mentzel.

6 recordsLinked to original sources

arXiv2023

k-strip: A novel segmentation algorithm in k-space for the application of skull stripping

Objectives: Present a novel deep learning-based skull stripping algorithm for magnetic resonance imaging (MRI) that works directly in the information rich k-space. Materials and Methods: Using two datasets from different institutions with a total of 36,900 MRI slices, we trained a deep learning-based model to work directly with the complex raw k-space data. Skull stripping performed by HD-BET (Brain Extraction Tool) in the image domain were used as the ground truth. Results: Both datasets were very similar to the ground truth (DICE scores of 92\%-98\% and Hausdorff distances of under 5.5 mm). Results on slices above the eye-region reach DICE scores of up to 99\%, while the accuracy drops in regions around the eyes and below, with partially blurred output. The output of k-strip often smoothed edges at the demarcation to the skull. Binary masks are created with an appropriate threshold. Conclusion: With this proof-of-concept study, we were able to show the feasibility of working in the k-space frequency domain, preserving phase information, with consistent results. Future research should be dedicated to discovering additional ways the k-space can be used for innovative image analysis and further workflows.

eess.IV↗

arXiv2023

Accurate and fast deep learning dose prediction for a preclinical microbeam radiation therapy study using low-statistics Monte Carlo simulations

Microbeam radiation therapy (MRT) utilizes coplanar synchrotron radiation beamlets and is a proposed treatment approach for several tumour diagnoses that currently have poor clinical treatment outcomes, such as gliosarcomas. Prescription dose estimations for treating preclinical gliosarcoma models in MRT studies at the Imaging and Medical Beamline at the Australian Synchrotron currently rely on Monte Carlo (MC) simulations. The steep dose gradients associated with the 50$\,μ$m wide coplanar beamlets present a significant challenge for precise MC simulation of the MRT irradiation treatment field in a short time frame. Much research has been conducted on fast dose estimation methods for clinically available treatments. However, such methods, including GPU Monte Carlo implementations and machine learning (ML) models, are unavailable for novel and emerging cancer radiation treatment options like MRT. In this work, the successful application of a fast and accurate machine learning dose prediction model in a retrospective preclinical MRT rodent study is presented for the first time. The ML model predicts the peak doses in the path of the microbeams and the valley doses between them, delivered to the gliosarcoma in rodent patients. The predictions of the ML model show excellent agreement with low-noise MC simulations, especially within the investigated tumour volume. This agreement is despite the ML model being deliberately trained with MC-calculated samples exhibiting significantly higher statistical uncertainties. The successful use of high-noise training set data samples, which are much faster to generate, encourages and accelerates the transfer of the ML model to different treatment modalities for other future applications in novel radiation cancer therapies.

physics.med-ph↗

arXiv2022

A step towards treatment planning for microbeam radiation therapy: fast peak and valley dose predictions with 3D U-Nets

Fast and accurate dose predictions are one of the bottlenecks in treatment planning for microbeam radiation therapy (MRT). In this paper, we propose a machine learning (ML) model based on a 3D U-Net. Our approach predicts separately the large doses of the narrow high intensity synchrotron microbeams and the lower valley doses between them. For this purpose, a concept of macro peak doses and macro valley doses is introduced, describing the respective doses not on a microscopic level but as macroscopic quantities in larger voxels. The ML model is trained to mimic full Monte Carlo (MC) data. Complex physical effects such as polarization are therefore automatically taking into account by the model. The macro dose distribution approach described in this study allows for superimposing single microbeam predictions to a beam array field making it an interesting candidate for treatment planning. It is shown that the proposed approach can overcome a main obstacle with microbeam dose predictions by predicting a full microbeam irradiation field in less than a minute while maintaining reasonable accuracy.

physics.med-ph↗

arXiv2022

Fast and accurate dose predictions for novel radiotherapy treatments in heterogeneous phantoms using conditional 3D-UNet generative adversarial networks

Novel radiotherapy techniques like synchrotron X-ray microbeam radiation therapy (MRT), require fast dose distribution predictions that are accurate at the sub-mm level, especially close to tissue/bone/air interfaces. Monte Carlo physics simulations are recognised to be one of the most accurate tools to predict the dose delivered in a target tissue but can be very time consuming and therefore prohibitive for treatment planning. Faster dose prediction algorithms are usually developed for clinically deployed treatments only. In this work, we explore a new approach for fast and accurate dose estimations suitable for novel treatments using digital phantoms used in pre-clinical development and modern machine learning techniques. We develop a generative adversarial network (GAN) model, which is able to emulate the equivalent Geant4 Monte Carlo simulation with adequate accuracy, and use it to predict the radiation dose delivered by a broad synchrotron beam to various phantoms. The energy depositions used for the training of the GAN are obtained using full Geant4 Monte Carlo simulations of a synchrotron radiation broad beam passing through the phantoms. The energy deposition is scored and predicted in voxel matrices of size 140x18x18 with a voxel edge length of 1 mm. The GAN model consists of two competing 3D convolutional neural networks, which are conditioned on the photon beam and phantom properties. The energy deposition predictions inside all phantom geometries under investigation show deviations of less than 3% of the maximum deposited energy from the simulation for roughly 99% of the voxels in the field of the beam. The computing time for a single prediction is reduced from several hundred hours using Geant4 simulation to less than a second using the GAN model.

physics.med-ph↗

arXiv2021

No more glowing in the dark: How deep learning improves exposure date estimation in thermoluminescence dosimetry

The time- or temperature-resolved detector signal from a thermoluminescence dosimeter can reveal additional information about circumstances of an exposure to ionizing irradiation. We present studies using deep neural networks to estimate the date of a single irradiation with 12 mSv within a monitoring interval of 42 days from glow curves of novel TL-DOS personal dosimeters developed by the Materialprüfungsamt NRW in cooperation with TU Dortmund University. Using a deep convolutional network, the irradiation date can be predicted from raw time-resolved glow curve data with an uncertainty of roughly 1-2 days on a 68% confidence level without the need for a prior transformation into temperature space and a subsequent glow curve deconvolution. This corresponds to a significant improvement in prediction accuracy compared to a prior publication, which yielded a prediction uncertainty of 2-4 days using features obtained from a glow curve deconvolution as input to a neural network.

physics.med-ph↗

arXiv2017

Objective Bayesian analysis of counting experiments with correlated sources of background

Searches for faint signals in counting experiments are often encountered in particle physics and astrophysics, as well as in other fields. Many problems can be reduced to the case of a model with independent and Poisson-distributed signal and background. Often several background contributions are present at the same time, possibly correlated. We provide the analytic solution of the statistical inference problem of estimating the signal in the presence of multiple backgrounds, in the framework of objective Bayes statistics. The model can be written in the form of a product of a single Poisson distribution with a multinomial distribution. The first is related to the total number of events, whereas the latter describes the fraction of events coming from each individual source. Correlations among different backgrounds can be included in the inference problem by a suitable choice of the priors.

physics.data-an↗