-
3D U-Net for segmentation of COVID-19 associated pulmonary infiltrates using transfer learning: State-of-the-art results on affordable hardware
Authors:
Keno K. Bressem,
Stefan M. Niehues,
Bernd Hamm,
Marcus R. Makowski,
Janis L. Vahldiek,
Lisa C. Adams
Abstract:
Segmentation of pulmonary infiltrates can help assess severity of COVID-19, but manual segmentation is labor and time-intensive. Using neural networks to segment pulmonary infiltrates would enable automation of this task. However, training a 3D U-Net from computed tomography (CT) data is time- and resource-intensive. In this work, we therefore developed and tested a solution on how transfer learni…
▽ More
Segmentation of pulmonary infiltrates can help assess severity of COVID-19, but manual segmentation is labor and time-intensive. Using neural networks to segment pulmonary infiltrates would enable automation of this task. However, training a 3D U-Net from computed tomography (CT) data is time- and resource-intensive. In this work, we therefore developed and tested a solution on how transfer learning can be used to train state-of-the-art segmentation models on limited hardware and in shorter time. We use the recently published RSNA International COVID-19 Open Radiology Database (RICORD) to train a fully three-dimensional U-Net architecture using an 18-layer 3D ResNet, pretrained on the Kinetics-400 dataset as encoder. The generalization of the model was then tested on two openly available datasets of patients with COVID-19, who received chest CTs (Corona Cases and MosMed datasets). Our model performed comparable to previously published 3D U-Net architectures, achieving a mean Dice score of 0.679 on the tuning dataset, 0.648 on the Coronacases dataset and 0.405 on the MosMed dataset. Notably, these results were achieved with shorter training time on a single GPU with less memory available than the GPUs used in previous studies.
△ Less
Submitted 25 January, 2021;
originally announced January 2021.
-
Scalar Field Wave (Fuzzy) Dark Matter and the Formation of Galaxies
Authors:
Benjamin Hamm
Abstract:
We investigate Scalar Field Wave Dark Matter in the context of galactic Dark Matter halos. In particular, we offer an analysis of the Baryonic Tully-Fisher Relation (BTFR). We detail a particular family of excited state solutions to the Einstein-Klein-Gordon equations, and use it to provide a novel theoretical model for producing the BTFR. We then solve this model computationally to simulate the B…
▽ More
We investigate Scalar Field Wave Dark Matter in the context of galactic Dark Matter halos. In particular, we offer an analysis of the Baryonic Tully-Fisher Relation (BTFR). We detail a particular family of excited state solutions to the Einstein-Klein-Gordon equations, and use it to provide a novel theoretical model for producing the BTFR. We then solve this model computationally to simulate the BTFR. Interpreting the Dark Matter scalar field as an ultralight boson, this returns a conservative mass constraint of $m\geq 10^{-23}eV$. Assuming slightly stronger conditions suggests $m\geq 10^{-22}eV$ to be more compatible with the BTFR. We provide a discussion of Scalar Field Dark Matter rotation curves and the structure of Scalar Field Dark Matter halos. Compatibility with the BTFR requires the excited state solutions to obey particular boundary conditions; this may have implications for the behavior of Dark Matter halos and the formation of galaxies.
△ Less
Submitted 16 April, 2020;
originally announced April 2020.
-
Comparing Different Deep Learning Architectures for Classification of Chest Radiographs
Authors:
Keno K. Bressem,
Lisa Adams,
Christoph Erxleben,
Bernd Hamm,
Stefan Niehues,
Janis Vahldiek
Abstract:
Chest radiographs are among the most frequently acquired images in radiology and are often the subject of computer vision research. However, most of the models used to classify chest radiographs are derived from openly available deep neural networks, trained on large image-datasets. These datasets routinely differ from chest radiographs in that they are mostly color images and contain several poss…
▽ More
Chest radiographs are among the most frequently acquired images in radiology and are often the subject of computer vision research. However, most of the models used to classify chest radiographs are derived from openly available deep neural networks, trained on large image-datasets. These datasets routinely differ from chest radiographs in that they are mostly color images and contain several possible image classes, while radiographs are greyscale images and often only contain fewer image classes. Therefore, very deep neural networks, which can represent more complex relationships in image-features, might not be required for the comparatively simpler task of classifying grayscale chest radiographs. We compared fifteen different architectures of artificial neural networks regarding training-time and performance on the openly available CheXpert dataset to identify the most suitable models for deep learning tasks on chest radiographs. We could show, that smaller networks such as ResNet-34, AlexNet or VGG-16 have the potential to classify chest radiographs as precisely as deeper neural networks such as DenseNet-201 or ResNet-151, while being less computationally demanding.
△ Less
Submitted 20 February, 2020;
originally announced February 2020.
-
Flatly Foliated Relativity
Authors:
Hubert Bray,
Benjamin Hamm,
Sven Hirsch,
James Wheeler,
Yiyue Zhang
Abstract:
Flatly Foliated Relativity (FFR) is a new theory which conceptually lies between Special Relativity (SR) and General Relativity (GR), in which spacetime is foliated by flat Euclidean spaces. While GR is based on the idea that "matter curves spacetime", FFR is based on the idea that "matter curves spacetime, but not space". This idea, inspired by the observed spatial flatness of our local universe,…
▽ More
Flatly Foliated Relativity (FFR) is a new theory which conceptually lies between Special Relativity (SR) and General Relativity (GR), in which spacetime is foliated by flat Euclidean spaces. While GR is based on the idea that "matter curves spacetime", FFR is based on the idea that "matter curves spacetime, but not space". This idea, inspired by the observed spatial flatness of our local universe, is realized by considering the same action as used in GR, but restricting it only to metrics which are foliated by flat spatial slices. FFR can be thought of as describing gravity without gravitational waves.
In FFR, a positive cosmological constant implies several interesting properties which do not follow in GR: the metric equations are elliptic on each euclidean slice, there exists a unique vacuum solution among those spherically symmetric at infinity, and there exists a geometric way to define the arrow of time. Furthermore, as gravitational waves do not exist in FFR, there are simple analogs to the positive mass theorem and Penrose-type inequalities.
Importantly, given that gravitational waves have a negligible effect on the curvature of spacetime, and that the universe appears to be locally flat, FFR may be a good approximation of GR. Moreover, FFR still admits many notable features of GR including the big bang, an accelerating expansion of the universe, and the Schwarzschild spacetime. Lastly, FFR is already known to have an existence theory for some simplified cases, which provokes an interesting discussion regarding the possibility of a more general existence theory, which may be relevant to understanding existence of solutions to GR.
△ Less
Submitted 13 November, 2019; v1 submitted 3 November, 2019;
originally announced November 2019.
