SciCombinator

Discover the most talked about and latest scientific content & concepts.

Journal: Zeitschrift fur medizinische Physik

0

This paper tries to give a gentle introduction to deep learning in medical image processing, proceeding from theoretical foundations to applications. We first discuss general reasons for the popularity of deep learning, including several major breakthroughs in computer science. Next, we start reviewing the fundamental basics of the perceptron and neural networks, along with some fundamental theory that is often omitted. Doing so allows us to understand the reasons for the rise of deep learning in many application domains. Obviously medical image processing is one of these areas which has been largely affected by this rapid progress, in particular in image detection and recognition, image segmentation, image registration, and computer-aided diagnosis. There are also recent trends in physical simulation, modeling, and reconstruction that have led to astonishing results. Yet, some of these approaches neglect prior knowledge and hence bear the risk of producing implausible results. These apparent weaknesses highlight current limitations of deep ()learning. However, we also briefly discuss promising approaches that might be able to resolve these problems in the future.

0

The increasing frequency and complexity of medical radiation exposures to humans inevitably result in higher risks of harmful unintended or accidental radiation exposures. To ensure a high level of protection and its continuous improvement, the Directive 2013/59/Euratom thus requires to systematically record and analyze both events and near-miss events as well as, in the case of their significance, to disseminate information regarding lessons learned from these events promptly and nationwide to improve radiation protection in medicine. These requirements have been transposed into German legislation by the new radiation protection law and radiation protection ordinance that entered into force simultaneously on December 31th, 2018. The reporting and information system as provided by these regulations as well as the tasks, duties and powers of the parties involved are presented in the first part of this review article. In the second part, the established application-specified criteria for the significance - and thus the notification requirement - of (near-miss) events are itemized and explicated.

0

This addendum provides correction factors for the recombination and the polarity effect for the new ionization chamber PTW PinPoint (type 31023). The measurements were made in filtered (WFF) and unfiltered (FFF) high-energy photon beams. It could be confirmed that both the initial and the general recombination effect of the chamber mainly depends on dose per pulse at the point of measurement and is independent of the filtration of the photon beam.

0

Glioblastoma multiforme is the most frequent innate brain tumor and still yields an unfavorable prognosis of 15 months of survival after diagnosis. Animal models play an important role in the investigation of therapeutic strategies of brain tumors. Thereby, MRI is substantial to individual follow-up examination for therapeutic response. Contrast agent dosage at 1.5 and 3T MRI has been thoroughly tested, while there is hardly any data for 9.4T. Therefore, the aim of this study was to find the optimal contrast agent dosage at 9.4T for examination of T1 relaxation time and apparent tumor volume in an animal model.

0

Inter-fractional variations of breathing pattern and patient anatomy introduce dose uncertainties in proton therapy. One approach to monitor these variations is to utilize the cone-beam computed tomography (CT, CBCT) scans routinely taken for patient positioning, reconstruct them as 4DCBCTs, and generate ‘virtual CTs’ (vCTs), combining the accurate CT numbers of the diagnostic 4DCT and the geometry of the daily 4DCBCT by using deformable image registration (DIR). In this study different algorithms for 4DCBCT reconstruction and DIR were evaluated. For this purpose, CBCT scans of a moving ex vivo porcine lung phantom with 663 and 2350 projections respectively were acquired, accompanied by an additional 4DCT as reference. The CBCT projections were sorted in 10 phase bins with the Amsterdam-shroud method and reconstructed phase-by-phase using first a FDK reconstruction from the Reconstruction Toolkit (RTK) and again an iterative reconstruction algorithm implemented in the Gadgetron Toolkit. The resulting 4DCBCTs were corrected by DIR of the corresponding 4DCT phases, using both a morphons algorithm from REGGUI and a b-spline deformation from Plastimatch. The resulting 4DvCTs were compared to the 4DCT by visual inspection and by calculating water equivalent thickness (WET) maps from the phantom’s surface to the distal edge of a target from various angles. The optimized procedure was successfully repeated with mismatched input phases and on a clinical patient dataset. Proton treatment plans were simulated on the 4DvCTs and the dose distributions compared to the reference based on the 4DCT via gamma pass rate analysis. A combination of iterative reconstruction and morphons DIR yielded the most accurate 4DvCTs, with median WET differences under 2mm and 3%/3mm gamma pass rates per phase between 89% and 99%. These results suggest that image correction of iteratively reconstructed 4DCBCTs with a morphons DIR of the planning CT may yield sufficiently accurate 4DvCTs for daily time resolved proton dose calculations.

0

The current work investigates the performance of different multivariate supervised machine learning models to predict the presence or absence of multiple sclerosis (MS) based on features derived from quantitative MRI acquisitions. The performance of these models was evaluated for images which are significantly degraded due to subject motion, a problem which is often observed in clinical routine diagnostics. Finally, the difference between a true multivariate analysis and the corresponding univariate analysis based on single parameters alone was addressed.

0

Proton radiotherapy (PT) requires accurate target alignment before each treatment fraction, ideally utilizing 3D in-room X-ray computed tomography (CT) imaging. Typically, the optimal patient position is determined based on anatomical landmarks or implanted markers. In the presence of non-rigid anatomical changes, however, the planning scenario cannot be exactly reproduced and positioning should rather aim at finding the optimal position in terms of the actually applied dose. In this work, dose-guided patient alignment, implemented as multicriterial optimization (MCO) problem, was investigated in the scope of intensity-modulated and double-scattered PT (IMPT and DSPT) for the first time. A method for automatically determining the optimal patient position with respect to pre-defined clinical goals was implemented. Linear dose interpolation was used to access a continuous space of potential patient shifts. Fourteen head and neck (H&N) and eight prostate cancer patients with up to five repeated CTs were included. Dose interpolation accuracy was evaluated and the potential dosimetric advantages of dose-guided over bony-anatomy-based patient alignment investigated by comparison of clinically relevant target and organ-at-risk (OAR) dose-volume histogram (DVH) parameters. Dose interpolation was found sufficiently accurate with average pass-rates of 90% and 99% for an exemplary H&N and prostate patient, respectively, using a 2% dose-difference criterion. Compared to bony-anatomy-based alignment, the main impact of automated MCO-based dose-guided positioning was a reduced dose to the serial OARs (spinal cord and brain stem) for the H&N cohort. For the prostate cohort, under-dosage of the target structures could be efficiently diminished. Limitations of dose-guided positioning were mainly found in reducing target over-dosage due to weight loss for H&N patients, which might require adaptation of the treatment plan. Since labor-intense online quality-assurance is not required for dose-guided patient positioning, it might, nevertheless, be considered an interesting alternative to full online re-planning for initially mitigating the effects of anatomical changes.

0

The purpose of this study was to investigate the suitability of the microDiamond detector (mDD) type 60019 (PTW-Freiburg, Germany) for radial dose function measurements with High Dose Rate (HDR) 192Ir brachytherapy sources. An HDR 192Ir source model mHDR v2r (Nucletron BV, an Elekta company, The Netherlands) was placed at the centre of a MP3 water phantom (PTW-Freiburg, Germany) within a 4F needle. Three mDDs were employed to measure the radial dose function of the source by acquiring profiles along the source transverse axis. Meanwhile, the experimental setup was simulated using the Monte Carlo (MC) code MCNP6.1™ (Los Alamos National Laboratory, USA) to calculate phantom-size, absorbed-dose energy dependence and volume averaging correction factors. After applying the correction factors, the radial dose function gL® for the line source approximation was calculated as defined in the TG-43 formalism at radial distances from 0.5cm to 10cm and compared to the consensus gL® (ESTRO and AAPM). The percentage differences to the consensus gL® for all the three mDDs were from -2.3% to +1.4% for distances r≤5cm and -6.2% to +2.6% for larger distances. These results indicate the suitability of the mDD for HDR brachytherapy measurements when all required corrections are applied.

0

Quantitative susceptibility mapping provides a measure for the local susceptibility within a voxel in magnetic resonance imaging (MRI). So far, theoretical and numerical studies focus on the assumption of a constant susceptibility inside each MR voxel. For blood vessel networks, however, susceptibility differences between blood and surrounding tissue occur on a much smaller length scale than the typical voxel size in routine MRI. In this work, the dependency of the quantitative susceptibility value on vessel size and voxel size is analyzed.

0

The University Hospital of Düsseldorf, Germany (UKD) recently installed the Respiratory Gating for Scanners module (RGSC) (Varian Medical Systems, Palo Alto, USA). The aim of this article is to report on the commissioning and clinical implementation of the RGSC system. The steps encompassed the validation of the manufacturer’s specifications including functionality tests using a commercial and in-house developed breathing phantom, to establish calibration procedures, and clinical workflow analysis involving breath acquisition and patient data evaluation. In this context also the RGSC signal without motion was performed to assess the calibration procedure. Reproducibility test were conducted as well with breathing phantoms. Fifteen clinical breath curves were examined in order to assess the impact of treatment related uncertainties such as noises of the CT, patient positioning, movement of the CT table, unintended patient motion. Finally, different binning approaches were assessed and the effect on the CT reconstructions and methodic advantages were investigated. All technical specifications of the manufacturer were confirmed. A baseline drift of 1.83mm of the measured breath curve occurred during longitudinal movement of the CT table. This drift is smaller if the direction of table motion coincides precisely with the level of calibration. If the calibration is carried out on extensions for patient positioning we measured a baseline drift up to 6mm. It was found that especially for a combination of a ceiling mounted IR-camera and amplitude based 4D-CT reconstructions, precise calibration is prerequisite. The evaluations of patient breath curves and corresponding CT reconstructions revealed patient specific aspects and variations, respectively. Consequently patient selection criteria need to be established in parallel with the technical implementation and validation phase of respiratory gating.