-
Interactive Image Selection and Training for Brain Tumor Segmentation Network
Authors:
Matheus A. Cerqueira,
Flávia Sprenger,
Bernardo C. A. Teixeira,
Alexandre X. Falcão
Abstract:
Medical image segmentation is a relevant problem, with deep learning being an exponent. However, the necessity of a high volume of fully annotated images for training massive models can be a problem, especially for applications whose images present a great diversity, such as brain tumors, which can occur in different sizes and shapes. In contrast, a recent methodology, Feature Learning from Image…
▽ More
Medical image segmentation is a relevant problem, with deep learning being an exponent. However, the necessity of a high volume of fully annotated images for training massive models can be a problem, especially for applications whose images present a great diversity, such as brain tumors, which can occur in different sizes and shapes. In contrast, a recent methodology, Feature Learning from Image Markers (FLIM), has involved an expert in the learning loop, producing small networks that require few images to train the convolutional layers. In this work, We employ an interactive method for image selection and training based on FLIM, exploring the user's knowledge. The results demonstrated that with our methodology, we could choose a small set of images to train the encoder of a U-shaped network, obtaining performance equal to manual selection and even surpassing the same U-shaped network trained with backpropagation and all training images.
△ Less
Submitted 5 June, 2024;
originally announced June 2024.
-
Building Brain Tumor Segmentation Networks with User-Assisted Filter Estimation and Selection
Authors:
Matheus A. Cerqueira,
Flávia Sprenger,
Bernardo C. A. Teixeira,
Alexandre X. Falcão
Abstract:
Brain tumor image segmentation is a challenging research topic in which deep-learning models have presented the best results. However, the traditional way of training those models from many pre-annotated images leaves several unanswered questions. Hence methodologies, such as Feature Learning from Image Markers (FLIM), have involved an expert in the learning loop to reduce human effort in data ann…
▽ More
Brain tumor image segmentation is a challenging research topic in which deep-learning models have presented the best results. However, the traditional way of training those models from many pre-annotated images leaves several unanswered questions. Hence methodologies, such as Feature Learning from Image Markers (FLIM), have involved an expert in the learning loop to reduce human effort in data annotation and build models sufficiently deep for a given problem. FLIM has been successfully used to create encoders, estimating the filters of all convolutional layers from patches centered at marker voxels. In this work, we present Multi-Step (MS) FLIM - a user-assisted approach to estimating and selecting the most relevant filters from multiple FLIM executions. MS-FLIM is used only for the first convolutional layer, and the results already indicate improvement over FLIM. For evaluation, we build a simple U-shaped encoder-decoder network, named sU-Net, for glioblastoma segmentation using T1Gd and FLAIR MRI scans, varying the encoder's training method, using FLIM, MS-FLIM, and backpropagation algorithm. Also, we compared these sU-Nets with two State-Of-The-Art (SOTA) deep-learning models using two datasets. The results show that the sU-Net based on MS-FLIM outperforms the other training methods and achieves effectiveness within the standard deviations of the SOTA models.
△ Less
Submitted 19 March, 2024;
originally announced March 2024.
-
Federated Learning Enables Big Data for Rare Cancer Boundary Detection
Authors:
Sarthak Pati,
Ujjwal Baid,
Brandon Edwards,
Micah Sheller,
Shih-Han Wang,
G Anthony Reina,
Patrick Foley,
Alexey Gruzdev,
Deepthi Karkada,
Christos Davatzikos,
Chiharu Sako,
Satyam Ghodasara,
Michel Bilello,
Suyash Mohan,
Philipp Vollmuth,
Gianluca Brugnara,
Chandrakanth J Preetha,
Felix Sahm,
Klaus Maier-Hein,
Maximilian Zenk,
Martin Bendszus,
Wolfgang Wick,
Evan Calabrese,
Jeffrey Rudie,
Javier Villanueva-Meyer
, et al. (254 additional authors not shown)
Abstract:
Although machine learning (ML) has shown promise in numerous domains, there are concerns about generalizability to out-of-sample data. This is currently addressed by centrally sharing ample, and importantly diverse, data from multiple sites. However, such centralization is challenging to scale (or even not feasible) due to various limitations. Federated ML (FL) provides an alternative to train acc…
▽ More
Although machine learning (ML) has shown promise in numerous domains, there are concerns about generalizability to out-of-sample data. This is currently addressed by centrally sharing ample, and importantly diverse, data from multiple sites. However, such centralization is challenging to scale (or even not feasible) due to various limitations. Federated ML (FL) provides an alternative to train accurate and generalizable ML models, by only sharing numerical model updates. Here we present findings from the largest FL study to-date, involving data from 71 healthcare institutions across 6 continents, to generate an automatic tumor boundary detector for the rare disease of glioblastoma, utilizing the largest dataset of such patients ever used in the literature (25,256 MRI scans from 6,314 patients). We demonstrate a 33% improvement over a publicly trained model to delineate the surgically targetable tumor, and 23% improvement over the tumor's entire extent. We anticipate our study to: 1) enable more studies in healthcare informed by large and diverse data, ensuring meaningful results for rare diseases and underrepresented populations, 2) facilitate further quantitative analyses for glioblastoma via performance optimization of our consensus model for eventual public release, and 3) demonstrate the effectiveness of FL at such scale and task complexity as a paradigm shift for multi-site collaborations, alleviating the need for data sharing.
△ Less
Submitted 25 April, 2022; v1 submitted 22 April, 2022;
originally announced April 2022.
-
Assignment of resonances in dissociative recombination of HD+ ions: high-resolution measurements compared with accurate computations
Authors:
F. O. Waffeu Tamo,
H. Buhr,
O. Motapon,
S. Altevogt,
V. M. Andrianarijaona,
M. Grieser,
L. Lammich,
M. Lestinsky,
M. Motsch,
I. Nevo,
S. Novotny,
D. A. Orlov,
H. B. Pedersen,
D. Schwalm,
F. Sprenger,
X. Urbain,
U. Weigel,
A. Wolf,
I. F. Schneider
Abstract:
The collision-energy resolved rate coefficient for dissociative recombination of HD+ ions in the vibrational ground state is measured using the photocathode electron target at the heavy-ion storage ring TSR. Rydberg resonances associated with ro-vibrational excitation of the HD+ core are scanned as a function of the electron collision energy with an instrumental broadening below 1 meV in the low-e…
▽ More
The collision-energy resolved rate coefficient for dissociative recombination of HD+ ions in the vibrational ground state is measured using the photocathode electron target at the heavy-ion storage ring TSR. Rydberg resonances associated with ro-vibrational excitation of the HD+ core are scanned as a function of the electron collision energy with an instrumental broadening below 1 meV in the low-energy limit. The measurement is compared to calculations using multichannel quantum defect theory, accounting for rotational structure and interactions and considering the six lowest rotational energy levels as initial ionic states. Using thermal equilibrium level populations at 300 K to approximate the experimental conditions, close correspondence between calculated and measured structures is found up to the first vibrational excitation threshold of the cations near 0.24 eV. Detailed assignments, including naturally broadened and overlap** Rydberg resonances, are performed for all structures up to 0.024 eV. Resonances from purely rotational excitation of the ion core are found to have similar strengths as those involving vibrational excitation. A dominant low-energy resonance is assigned to contributions from excited rotational states only. The results indicate strong modifications in the energy dependence of the dissociative recombination rate coefficient through the rotational excitation of the parent ions, and underline the need for studies with rotationally cold species to obtain results reflecting low-temperature ionized media.
△ Less
Submitted 26 July, 2011;
originally announced July 2011.
-
Dielectronic Recombination of Fe XV forming Fe XIV: Laboratory Measurements and Theoretical Calculations
Authors:
D. V. Lukić,
M. Schnell,
D. W. Savin,
C. Brandau,
E. W. Schmidt,
S. Böhm,
A. Müller,
S. Schippers,
M. Lestinsky,
F. Sprenger,
A. Wolf,
Z. Altun,
N. R. Badnell
Abstract:
We have measured resonance strengths and energies for dielectronic recombination (DR) of Mg-like Fe XV forming Al-like Fe XIV via N=3 -> N' = 3 core excitations in the electron-ion collision energy range 0-45 eV. All measurements were carried out using the heavy-ion Test Storage Ring at the Max Planck Institute for Nuclear Physics in Heidelberg, Germany. We have also carried out new multiconfigu…
▽ More
We have measured resonance strengths and energies for dielectronic recombination (DR) of Mg-like Fe XV forming Al-like Fe XIV via N=3 -> N' = 3 core excitations in the electron-ion collision energy range 0-45 eV. All measurements were carried out using the heavy-ion Test Storage Ring at the Max Planck Institute for Nuclear Physics in Heidelberg, Germany. We have also carried out new multiconfiguration Breit-Pauli (MCBP) calculations using the AUTOSTRUCTURE code. For electron-ion collision energies < 25 eV we find poor agreement between our experimental and theoretical resonance energies and strengths. From 25 to 42 eV we find good agreement between the two for resonance energies. But in this energy range the theoretical resonance strengths are ~ 31% larger than the experimental results. This is larger than our estimated total experimental uncertainty in this energy range of +/- 26% (at a 90% confidence level). Above 42 eV the difference in the shape between the calculated and measured 3s3p(^1P_1)nl DR series limit we attribute partly to the nl dependence of the detection probabilities of high Rydberg states in the experiment. We have used our measurements, supplemented by our AUTOSTRUCTURE calculations, to produce a Maxwellian-averaged 3 -> 3 DR rate coefficient for Fe XV forming Fe XIV. The resulting rate coefficient is estimated to be accurate to better than +/- 29% (at a 90% confidence level) for k_BT_e > 1 eV. At temperatures of k_BT_e ~ 2.5-15 eV, where Fe XV is predicted to form in photoionized plasmas, significant discrepancies are found between our experimentally-derived rate coefficient and previously published theoretical results. Our new MCBP plasma rate coefficient is 19-28% smaller than our experimental results over this temperature range.
△ Less
Submitted 6 April, 2007;
originally announced April 2007.
-
Electron-ion recombination measurements motivated by AGN X-ray absorption features: Fe XIV forming Fe XIII
Authors:
E. W. Schmidt,
S. Schippers,
A. Mueller,
M. Lestinsky,
F. Sprenger,
M. Grieser,
R. Repnow,
A. Wolf,
C. Brandau,
D. Lukic,
M. Schnell,
D. W. Savin
Abstract:
Recent spectroscopic models of active galactic nuclei (AGN) have indicated that the recommended electron-ion recombination rate coefficients for iron ions with partially filled M-shells are incorrect in the temperature range where these ions form in photoionized plasmas. We have investigated this experimentally for Fe XIV forming Fe XIII. The recombination rate coefficient was measured employing…
▽ More
Recent spectroscopic models of active galactic nuclei (AGN) have indicated that the recommended electron-ion recombination rate coefficients for iron ions with partially filled M-shells are incorrect in the temperature range where these ions form in photoionized plasmas. We have investigated this experimentally for Fe XIV forming Fe XIII. The recombination rate coefficient was measured employing the electron-ion merged beams method at the Heidelberg heavy-ion storage-ring TSR. The measured energy range of 0-260 eV encompassed all dielectronic recombination (DR) 1s2 2s2 2p6 3l 3l' 3l'' nl''' resonances associated with the 3p1/2 -> 3p3/2, 3s -> 3p, 3p -> 3d and 3s -> 3d core excitations within the M-shell of the Fe XIV 1s2 2s2 2p6 3s2 3p parent ion. This range also includes the 1s2 2s2 2p6 3l 3l' 4l'' nl''' resonances associated with 3s -> 4l'' and 3p -> 4l'' core excitations. We find that in the temperature range 2--14 eV, where Fe XIV is expected to form in a photoionized plasma, the Fe XIV recombination rate coefficient is orders of magnitude larger than previously calculated values.
△ Less
Submitted 14 March, 2006;
originally announced March 2006.
