-
Realistic 3D printed imaging tumor phantoms for validation of image processing algorithms
Authors:
Sepideh Hatamikia,
Ingo Gulyas,
Wolfgang Birkfellner,
Gernot Kronreif,
Alexander Unger,
Gunpreet Oberoi,
Andrea Lorenz,
Ewald Unger,
Joachim Kettenbach,
Michael Figl,
Janina Patsch,
Andreas Strassl,
Dietmar Georg,
Andreas Renner
Abstract:
Medical imaging phantoms are widely used for validation and verification of imaging systems and algorithms in surgical guidance and radiation oncology procedures. Especially, for the performance evaluation of new algorithms in the field of medical imaging, manufactured phantoms need to replicate specific properties of the human body, e.g., tissue morphology and radiological properties. Additive ma…
▽ More
Medical imaging phantoms are widely used for validation and verification of imaging systems and algorithms in surgical guidance and radiation oncology procedures. Especially, for the performance evaluation of new algorithms in the field of medical imaging, manufactured phantoms need to replicate specific properties of the human body, e.g., tissue morphology and radiological properties. Additive manufacturing (AM) technology provides an inexpensive opportunity for accurate anatomical replication with customization capabilities. In this study, we proposed a simple and cheap protocol to manufacture realistic tumor phantoms based on the filament 3D printing technology. Tumor phantoms with both homogenous and heterogenous radiodensity were fabricated. The radiodensity similarity between the printed tumor models and real tumor data from CT images of lung cancer patients was evaluated. Additionally, it was investigated whether a heterogeneity in the 3D printed tumor phantoms as observed in the tumor patient data had an influence on the validation of image registration algorithms. A density range between -217 to 226 HUs was achieved for 3D printed phantoms; this range of radiation attenuation is also observed in the human lung tumor tissue. The resulted HU range could serve as a lookup-table for researchers and phantom manufactures to create realistic CT tumor phantoms with the desired range of radiodensities. The 3D printed tumor phantoms also precisely replicated real lung tumor patient data regarding morphology and could also include life-like heterogeneity of the radiodensity inside the tumor models. An influence of the heterogeneity on accuracy and robustness of the image registration algorithms was not found.
△ Less
Submitted 27 November, 2022;
originally announced November 2022.
-
clDice -- A Novel Topology-Preserving Loss Function for Tubular Structure Segmentation
Authors:
Suprosanna Shit,
Johannes C. Paetzold,
Anjany Sekuboyina,
Ivan Ezhov,
Alexander Unger,
Andrey Zhylka,
Josien P. W. Pluim,
Ulrich Bauer,
Bjoern H. Menze
Abstract:
Accurate segmentation of tubular, network-like structures, such as vessels, neurons, or roads, is relevant to many fields of research. For such structures, the topology is their most important characteristic; particularly preserving connectedness: in the case of vascular networks, missing a connected vessel entirely alters the blood-flow dynamics. We introduce a novel similarity measure termed cen…
▽ More
Accurate segmentation of tubular, network-like structures, such as vessels, neurons, or roads, is relevant to many fields of research. For such structures, the topology is their most important characteristic; particularly preserving connectedness: in the case of vascular networks, missing a connected vessel entirely alters the blood-flow dynamics. We introduce a novel similarity measure termed centerlineDice (short clDice), which is calculated on the intersection of the segmentation masks and their (morphological) skeleta. We theoretically prove that clDice guarantees topology preservation up to homotopy equivalence for binary 2D and 3D segmentation. Extending this, we propose a computationally efficient, differentiable loss function (soft-clDice) for training arbitrary neural segmentation networks. We benchmark the soft-clDice loss on five public datasets, including vessels, roads and neurons (2D and 3D). Training on soft-clDice leads to segmentation with more accurate connectivity information, higher graph similarity, and better volumetric scores.
△ Less
Submitted 15 July, 2022; v1 submitted 16 March, 2020;
originally announced March 2020.
-
Generalizing Emergent Communication
Authors:
Thomas A. Unger,
Elia Bruni
Abstract:
We converted the recently developed BabyAI grid world platform to a sender/receiver setup in order to test the hypothesis that established deep reinforcement learning techniques are sufficient to incentivize the emergence of a grounded discrete communication protocol between generalized agents. This is in contrast to previous experiments that employed straight-through estimation or specialized ind…
▽ More
We converted the recently developed BabyAI grid world platform to a sender/receiver setup in order to test the hypothesis that established deep reinforcement learning techniques are sufficient to incentivize the emergence of a grounded discrete communication protocol between generalized agents. This is in contrast to previous experiments that employed straight-through estimation or specialized inductive biases. Our results show that these can indeed be avoided, by instead providing proper environmental incentives. Moreover, they show that a longer interval between communications incentivized more abstract semantics. In some cases, the communicating agents adapted to new environments more quickly than a monolithic agent, showcasing the potential of emergent communication for transfer learning and generalization in general.
△ Less
Submitted 14 December, 2020; v1 submitted 6 January, 2020;
originally announced January 2020.
-
Enhancing Flood Impact Analysis using Interactive Retrieval of Social Media Images
Authors:
Björn Barz,
Kai Schröter,
Moritz Münch,
Bin Yang,
Andrea Unger,
Doris Dransch,
Joachim Denzler
Abstract:
The analysis of natural disasters such as floods in a timely manner often suffers from limited data due to a coarse distribution of sensors or sensor failures. This limitation could be alleviated by leveraging information contained in images of the event posted on social media platforms, so-called "Volunteered Geographic Information (VGI)". To save the analyst from the need to inspect all images p…
▽ More
The analysis of natural disasters such as floods in a timely manner often suffers from limited data due to a coarse distribution of sensors or sensor failures. This limitation could be alleviated by leveraging information contained in images of the event posted on social media platforms, so-called "Volunteered Geographic Information (VGI)". To save the analyst from the need to inspect all images posted online manually, we propose to use content-based image retrieval with the possibility of relevance feedback for retrieving only relevant images of the event to be analyzed. To evaluate this approach, we introduce a new dataset of 3,710 flood images, annotated by domain experts regarding their relevance with respect to three tasks (determining the flooded area, inundation depth, water pollution). We compare several image features and relevance feedback methods on that dataset, mixed with 97,085 distractor images, and are able to improve the precision among the top 100 retrieval results from 55% with the baseline retrieval to 87% after 5 rounds of feedback.
△ Less
Submitted 9 August, 2019;
originally announced August 2019.
-
$J$-factors for self-interacting dark matter in 20 dwarf spheroidal galaxies
Authors:
Sebastian Bergström,
Riccardo Catena,
Andrea Chiappo,
Jan Conrad,
Björn Eurenius,
Magdalena Eriksson,
Michael Högberg,
Susanna Larsson,
Emelie Olsson,
Andreas Unger,
Rikard Wadman
Abstract:
Dwarf spheroidal galaxies are among the most promising targets for indirect dark matter (DM) searches in $γ$-rays. The $γ$-ray flux from DM annihilation in a dwarf spheroidal galaxy is proportional to the $J$-factor of the source. The $J$-factor of a dwarf spheroidal galaxy is the line-of-sight integral of the DM mass density squared times…
▽ More
Dwarf spheroidal galaxies are among the most promising targets for indirect dark matter (DM) searches in $γ$-rays. The $γ$-ray flux from DM annihilation in a dwarf spheroidal galaxy is proportional to the $J$-factor of the source. The $J$-factor of a dwarf spheroidal galaxy is the line-of-sight integral of the DM mass density squared times $\langle σ_{\rm ann} v_{\rm rel} \rangle/(σ_{\rm ann} v_{\rm rel})_0$, where $σ_{\rm ann} v_{\rm rel}$ is the DM annihilation cross-section times relative velocity $v_{\rm rel}=|{\bf v}_{\rm rel}|$, angle brackets denote average over ${\bf v}_{\rm rel}$, and $(σ_{\rm ann} v_{\rm rel})_0$ is the $v_{\rm rel}$-independent part of $σ_{\rm ann} v_{\rm rel}$. If $σ_{\rm ann} v_{\rm rel}$ is constant in $v_{\rm rel}$, $J$-factors only depend on the DM space distribution in the source. However, if $σ_{\rm ann} v_{\rm rel}$ varies with $v_{\rm rel}$, as in the presence of DM self-interactions, $J$-factors also depend on the DM velocity distribution, and on the strength and range of the DM self-interaction. Models for self-interacting DM are increasingly important in the study of the small scale clustering of DM, and are compatible with current cosmological observations. Here we derive the $J$-factor of 20 dwarf spheroidal galaxies from stellar kinematic data under the assumption of Yukawa DM self-interactions. $J$-factors are derived through a profile Likelihood approach, assuming either NFW or cored DM profiles. We also compare our results with $J$-factors derived assuming the same velocity for all DM particles in the target galaxy. We find that this common approximation overestimates the $J$-factors by up to one order of magnitude. $J$-factors for a sample of DM particle masses, self-interaction coupling constants and density profiles are provided electronically, ready to be used in other projects.
△ Less
Submitted 8 December, 2017;
originally announced December 2017.
-
The mineral clouds on HD 209458b and HD189733b
Authors:
Ch. Helling,
G. Lee,
I. Dobbs-Dixon,
N. Mayne,
D. S. Amundsen,
J. Khaimova,
A. A. Unger,
J. Manners,
D. Acreman,
C. Smith
Abstract:
3D atmosphere model results are used to comparatively study the kinetic, non-equilibrium cloud formation in the atmospheres of two example planets guided by the giant gas planets HD209458b and HD189733b. Rather independently of hydrodynamic model differences, our cloud modelling suggests that both planets are covered in mineral clouds throughout the entire modelling domain. Both planets harbour ch…
▽ More
3D atmosphere model results are used to comparatively study the kinetic, non-equilibrium cloud formation in the atmospheres of two example planets guided by the giant gas planets HD209458b and HD189733b. Rather independently of hydrodynamic model differences, our cloud modelling suggests that both planets are covered in mineral clouds throughout the entire modelling domain. Both planets harbour chemically complex clouds that are made of mineral particles that have a height-dependent material composition and size. The remaining gas-phase element abundances strongly effects the molecular abundances of the atmosphere in the cloud forming regions. Hydrocarbon and cyanopolyyne molecules can be rather abundant in the inner, dense part of the atmospheres of HD189733b and HD209458b. No one value for metallicity and the C/O ratio can be used to describe an extrasolar planet. Our results concerning the presence and location of water in relation to the clouds explain some of the observed differences between the two planets. In HD189733b, strong water features have been reported while such features appear less strong for HD209458b. By considering the location of the clouds in the two atmospheres, we see that obscuring clouds exist high in the atmosphere of HD209458b, but much deeper in HD189733b. We further conclude that the (self-imposed) degeneracy of cloud parameters in retrieval methods can only be lifted if the cloud formation processes are accurately modelled in contrast to prescribing them by independent parameters.
△ Less
Submitted 23 March, 2016; v1 submitted 13 March, 2016;
originally announced March 2016.
-
Detecting molecules with plasmonic resonators - Analytic expressions and bounds for the sensitivity and figure of merit
Authors:
A. Unger,
M. Kreiter
Abstract:
We derive analytic expressions for the sensitivity and figure of merit for refractive index sensing with plasmonic resonators in the quasistatic limit. Based on these expressions the limits of detection and their origins are investigated. The critical role of limited energy concentration within the resonators surrounding is shown. The classical figure of merit is found to be solely dependent on th…
▽ More
We derive analytic expressions for the sensitivity and figure of merit for refractive index sensing with plasmonic resonators in the quasistatic limit. Based on these expressions the limits of detection and their origins are investigated. The critical role of limited energy concentration within the resonators surrounding is shown. The classical figure of merit is found to be solely dependent on the ratio of real to imaginary part of the resonators permittivity . Based on these findings we discuss the optimum resonance wavelength and resonator size and shape for single molecule detection.
△ Less
Submitted 6 July, 2010;
originally announced July 2010.
