-
Non-Redundant Combination of Hand-Crafted and Deep Learning Radiomics: Application to the Early Detection of Pancreatic Cancer
Authors:
Rebeca Vétil,
Clément Abi-Nader,
Alexandre Bône,
Marie-Pierre Vullierme,
Marc-Michel Rohé,
Pietro Gori,
Isabelle Bloch
Abstract:
We address the problem of learning Deep Learning Radiomics (DLR) that are not redundant with Hand-Crafted Radiomics (HCR). To do so, we extract DLR features using a VAE while enforcing their independence with HCR features by minimizing their mutual information. The resulting DLR features can be combined with hand-crafted ones and leveraged by a classifier to predict early markers of cancer. We ill…
▽ More
We address the problem of learning Deep Learning Radiomics (DLR) that are not redundant with Hand-Crafted Radiomics (HCR). To do so, we extract DLR features using a VAE while enforcing their independence with HCR features by minimizing their mutual information. The resulting DLR features can be combined with hand-crafted ones and leveraged by a classifier to predict early markers of cancer. We illustrate our method on four early markers of pancreatic cancer and validate it on a large independent test set. Our results highlight the value of combining non-redundant DLR and HCR features, as evidenced by an improvement in the Area Under the Curve compared to baseline methods that do not address redundancy or solely rely on HCR features.
△ Less
Submitted 22 August, 2023;
originally announced August 2023.
-
Weakly-supervised positional contrastive learning: application to cirrhosis classification
Authors:
Emma Sarfati,
Alexandre Bône,
Marc-Michel Rohé,
Pietro Gori,
Isabelle Bloch
Abstract:
Large medical imaging datasets can be cheaply and quickly annotated with low-confidence, weak labels (e.g., radiological scores). Access to high-confidence labels, such as histology-based diagnoses, is rare and costly. Pretraining strategies, like contrastive learning (CL) methods, can leverage unlabeled or weakly-annotated datasets. These methods typically require large batch sizes, which poses a…
▽ More
Large medical imaging datasets can be cheaply and quickly annotated with low-confidence, weak labels (e.g., radiological scores). Access to high-confidence labels, such as histology-based diagnoses, is rare and costly. Pretraining strategies, like contrastive learning (CL) methods, can leverage unlabeled or weakly-annotated datasets. These methods typically require large batch sizes, which poses a difficulty in the case of large 3D images at full resolution, due to limited GPU memory. Nevertheless, volumetric positional information about the spatial context of each 2D slice can be very important for some medical applications. In this work, we propose an efficient weakly-supervised positional (WSP) contrastive learning strategy where we integrate both the spatial context of each 2D slice and a weak label via a generic kernel-based loss function. We illustrate our method on cirrhosis prediction using a large volume of weakly-labeled images, namely radiological low-confidence annotations, and small strongly-labeled (i.e., high-confidence) datasets. The proposed model improves the classification AUC by 5% with respect to a baseline model on our internal dataset, and by 26% on the public LIHC dataset from the Cancer Genome Atlas. The code is available at: https://github.com/Guerbet-AI/wsp-contrastive.
△ Less
Submitted 19 September, 2023; v1 submitted 10 July, 2023;
originally announced July 2023.
-
Learning to diagnose cirrhosis from radiological and histological labels with joint self and weakly-supervised pretraining strategies
Authors:
Emma Sarfati,
Alexandre Bone,
Marc-Michel Rohe,
Pietro Gori,
Isabelle Bloch
Abstract:
Identifying cirrhosis is key to correctly assess the health of the liver. However, the gold standard diagnosis of the cirrhosis needs a medical intervention to obtain the histological confirmation, e.g. the METAVIR score, as the radiological presentation can be equivocal. In this work, we propose to leverage transfer learning from large datasets annotated by radiologists, which we consider as a we…
▽ More
Identifying cirrhosis is key to correctly assess the health of the liver. However, the gold standard diagnosis of the cirrhosis needs a medical intervention to obtain the histological confirmation, e.g. the METAVIR score, as the radiological presentation can be equivocal. In this work, we propose to leverage transfer learning from large datasets annotated by radiologists, which we consider as a weak annotation, to predict the histological score available on a small annex dataset. To this end, we propose to compare different pretraining methods, namely weakly-supervised and self-supervised ones, to improve the prediction of the cirrhosis. Finally, we introduce a loss function combining both supervised and self-supervised frameworks for pretraining. This method outperforms the baseline classification of the METAVIR score, reaching an AUC of 0.84 and a balanced accuracy of 0.75, compared to 0.77 and 0.72 for a baseline classifier.
△ Less
Submitted 16 February, 2023;
originally announced February 2023.
-
Potential for definitive discovery of a 70 GeV dark matter WIMP with only second-order gauge couplings
Authors:
Bailey Tallman,
Alexandra Boone,
Adhithya Vijayakumar,
Fiona Lopez,
Samuel Apata,
Jehu Martinez,
Roland Allen
Abstract:
As astronomical observations and their interpretation improve, the case for cold dark matter (CDM) becomes increasingly persuasive. A particularly appealing version of CDM is a weakly interacting massive particle (WIMP) with a mass near the electroweak scale, which can naturally have the observed relic abundance after annihilation in the early universe. But in order for a WIMP to be consistent wit…
▽ More
As astronomical observations and their interpretation improve, the case for cold dark matter (CDM) becomes increasingly persuasive. A particularly appealing version of CDM is a weakly interacting massive particle (WIMP) with a mass near the electroweak scale, which can naturally have the observed relic abundance after annihilation in the early universe. But in order for a WIMP to be consistent with the currently stringent experimental constraints it must have relatively small cross-sections for indirect, direct, and collider detection. Using our calculations and estimates of these cross-sections, we discuss the potential for discovery of a recently proposed dark matter WIMP which has a mass of about 70 GeV/c$^2$ and only second-order couplings to W and Z bosons. There is evidence that indirect detection may already have been achieved, since analyses of the gamma rays detected by Fermi-LAT and the antiprotons observed by AMS-02 are consistent with 70 GeV dark matter having our calculated $\langle σ_{ann} v \rangle \approx 1.2 \times 10^{-26} $ cm$^3$/s. The estimated sensitivities for LZ and XENONnT indicate that these experiments may achieve direct detection within the next few years, since we estimate the relevant cross-section to be slightly above $10^{-48}$ cm$^2$. Other experiments such as PandaX, SuperCDMS, and especially DARWIN should be able to confirm on a longer time scale. The high-luminosity LHC might achieve collider detection within about 15 years, since we estimate a collider cross-section slightly below 1 femtobarn. Definitive confirmation should come from still more powerful planned collider experiments (such as a future circular collider) within 15-35 years.
△ Less
Submitted 24 October, 2022;
originally announced October 2022.
-
Learning shape distributions from large databases of healthy organs: applications to zero-shot and few-shot abnormal pancreas detection
Authors:
Rebeca Vétil,
Clément Abi Nader,
Alexandre Bône,
Marie-Pierre Vullierme,
Marc-Michel Roheé,
Pietro Gori,
Isabelle Bloch
Abstract:
We propose a scalable and data-driven approach to learn shape distributions from large databases of healthy organs. To do so, volumetric segmentation masks are embedded into a common probabilistic shape space that is learned with a variational auto-encoding network. The resulting latent shape representations are leveraged to derive zeroshot and few-shot methods for abnormal shape detection. The pr…
▽ More
We propose a scalable and data-driven approach to learn shape distributions from large databases of healthy organs. To do so, volumetric segmentation masks are embedded into a common probabilistic shape space that is learned with a variational auto-encoding network. The resulting latent shape representations are leveraged to derive zeroshot and few-shot methods for abnormal shape detection. The proposed distribution learning approach is illustrated on a large database of 1200 healthy pancreas shapes. Downstream qualitative and quantitative experiments are conducted on a separate test set of 224 pancreas from patients with mixed conditions. The abnormal pancreas detection AUC reached up to 65.41% in the zero-shot configuration, and 78.97% in the few-shot configuration with as few as 15 abnormal examples, outperforming a baseline approach based on the sole volume.
△ Less
Submitted 21 October, 2022;
originally announced October 2022.
-
CoRe: An Automated Pipeline for The Prediction of Liver Resection Complexity from Preoperative CT Scans
Authors:
Omar Ali,
Alexandre Bone,
Caterina Accardo,
Omar Belkouchi,
Marc-Michel Rohe,
Eric Vibert,
Irene Vignon-Clementel
Abstract:
Surgical resections are the most prevalent curative treatment for primary liver cancer. Tumors located in critical positions are known to complexify liver resections (LR). While experienced surgeons in specialized medical centers may have the necessary expertise to accurately anticipate LR complexity, and prepare accordingly, an objective method able to reproduce this behavior would have the poten…
▽ More
Surgical resections are the most prevalent curative treatment for primary liver cancer. Tumors located in critical positions are known to complexify liver resections (LR). While experienced surgeons in specialized medical centers may have the necessary expertise to accurately anticipate LR complexity, and prepare accordingly, an objective method able to reproduce this behavior would have the potential to improve the standard routine of care, and avoid intra- and postoperative complications. In this article, we propose CoRe, an automated medical image processing pipeline for the prediction of postoperative LR complexity from preoperative CT scans, using imaging biomarkers. The CoRe pipeline first segments the liver, lesions, and vessels with two deep learning networks. The liver vasculature is then pruned based on a topological criterion to define the hepatic central zone (HCZ), a convex volume circumscribing the major liver vessels, from which a new imaging biomarker, BHCZ is derived. Additional biomarkers are extracted and leveraged to train and evaluate a LR complexity prediction model. An ablation study shows the HCZ-based biomarker as the central feature in predicting LR complexity. The best predictive model reaches an accuracy, F1, and AUC of 77.3, 75.4, and 84.1% respectively.
△ Less
Submitted 15 October, 2022;
originally announced October 2022.
-
Indirect detection, direct detection, and collider detection cross-sections for a 70 GeV dark matter WIMP
Authors:
Bailey Tallman,
Alexandra Boone,
Caden LaFontaine,
Trevor Croteau,
Quinn Ballard,
Sabrina Hernandez,
Spencer Ellis,
Adhithya Vijayakumarm,
Fiona Lopez,
Samuel Apata,
Jehu Martinez,
Roland Allen
Abstract:
Assuming a dark matter fraction $Ω_{DM} = 0.27$ and a reduced Hubble constant $h = 0.73$, we obtain a value of 70 GeV/c$^2$ for the mass of the dark matter WIMP we have previously proposed. We also obtain a value for the annihilation cross section given by $\langle σ_{ann} v \rangle = 1.19 \times 10^{-26} $ cm$^3$/s in the present universe, consistent with the current limits for dwarf spheroidal g…
▽ More
Assuming a dark matter fraction $Ω_{DM} = 0.27$ and a reduced Hubble constant $h = 0.73$, we obtain a value of 70 GeV/c$^2$ for the mass of the dark matter WIMP we have previously proposed. We also obtain a value for the annihilation cross section given by $\langle σ_{ann} v \rangle = 1.19 \times 10^{-26} $ cm$^3$/s in the present universe, consistent with the current limits for dwarf spheroidal galaxies. Both the mass and cross-section are consistent with analyses of the Galactic-center gamma rays observed by Fermi-LAT and the antiprotons observed by AMS-02 if these data are interpreted as resulting from dark matter annihilation. The spin-independent cross-section for direct detection in Xe-based experiments is estimated to be slightly above $10^{-48}$ cm$^2$, presumably just within reach of the LZ and XENONnT experiments with $\gtrsim 1000$ days of data taking. The cross-section for production in high-energy proton collisions via vector boson fusion is estimated to be $\sim 1$ femtobarn, possibly within reach of the high-luminosity LHC, with $\ge 140$ GeV of missing energy accompanied by two jets.
△ Less
Submitted 8 October, 2022;
originally announced October 2022.
-
Understanding the Digital News Consumption Experience During the COVID Pandemic
Authors:
Mingrui Ray Zhang,
Ashley Boone,
Sara M Behbakht,
Alexis Hiniker
Abstract:
During the COVID-19 pandemic, people sought information through digital news platforms. To investigate how to design these platforms to support users' needs in a crisis, we conducted a two-week diary study with 22 participants across the United States. Participants' news-consumption experience followed two stages: in the \textbf{seeking} stage, participants increased their general consumption, mot…
▽ More
During the COVID-19 pandemic, people sought information through digital news platforms. To investigate how to design these platforms to support users' needs in a crisis, we conducted a two-week diary study with 22 participants across the United States. Participants' news-consumption experience followed two stages: in the \textbf{seeking} stage, participants increased their general consumption, motivated by three common informational needs -- specifically, to find, understand and verify relevant news pieces. Participants then moved to the \textbf{sustaining} stage, and coping with the news emotionally became as important as their informational needs. We elicited design ideas from participants and used these to distill six themes for creating digital news platforms that provide better informational and emotional support during a crisis. Thus, we contribute, first, a model of users' needs over time with respect to engaging with crisis news, and second, example design concepts for supporting users' needs in each of these stages.
△ Less
Submitted 10 February, 2022;
originally announced February 2022.
-
Stochastic paths controlling speed and dissipation
Authors:
Rebecca A. Bone,
Daniel J. Sharpe,
David J. Wales,
Jason R. Green
Abstract:
Near equilibrium, thermodynamic intuition suggests that fast, irreversible processes will dissipate more energy and entropy than slow, quasistatic processes connecting the same initial and final states. Here, we test the hypothesis that this relationship between speed and dissipation holds for stochastic processes far from equilibrium. To analyze these processes on finite timescales, we derive an…
▽ More
Near equilibrium, thermodynamic intuition suggests that fast, irreversible processes will dissipate more energy and entropy than slow, quasistatic processes connecting the same initial and final states. Here, we test the hypothesis that this relationship between speed and dissipation holds for stochastic processes far from equilibrium. To analyze these processes on finite timescales, we derive an exact expression for the path probabilities of continuous-time Markov chains from the path summation solution of the master equation. Applying this formula to a model for nonequilibrium self-assembly, we show that more speed can lead to less dissipation when there are strong nonequilibrium currents. In the model, the relative energies of the initial and target states control the speed, and the nonequilibrium currents of a cycle situated between these states control the path-level dissipation. This model serves as a minimal prototype for designing kinetics to sculpt the nonequilibrium path space, so that faster structure-forming paths dissipate less.
△ Less
Submitted 15 December, 2021; v1 submitted 24 September, 2021;
originally announced September 2021.
-
Benchmarking off-the-shelf statistical shape modeling tools in clinical applications
Authors:
Anupama Goparaju,
Alexandre Bone,
Nan Hu,
Heath B. Henninger,
Andrew E. Anderson,
Stanley Durrleman,
Matthijs Jacxsens,
Alan Morris,
Ibolya Csecs,
Nassir Marrouche,
Shireen Y. Elhabian
Abstract:
Statistical shape modeling (SSM) is widely used in biology and medicine as a new generation of morphometric approaches for the quantitative analysis of anatomical shapes. Technological advancements of in vivo imaging have led to the development of open-source computational tools that automate the modeling of anatomical shapes and their population-level variability. However, little work has been do…
▽ More
Statistical shape modeling (SSM) is widely used in biology and medicine as a new generation of morphometric approaches for the quantitative analysis of anatomical shapes. Technological advancements of in vivo imaging have led to the development of open-source computational tools that automate the modeling of anatomical shapes and their population-level variability. However, little work has been done on the evaluation and validation of such tools in clinical applications that rely on morphometric quantifications (e.g., implant design and lesion screening). Here, we systematically assess the outcome of widely used, state-of-the-art SSM tools, namely ShapeWorks, Deformetrica, and SPHARM-PDM. We use both quantitative and qualitative metrics to evaluate shape models from different tools. We propose validation frameworks for anatomical landmark/measurement inference and lesion screening. We also present a lesion screening method to objectively characterize subtle abnormal shape changes with respect to learned population-level statistics of controls. Results demonstrate that SSM tools display different levels of consistencies, where ShapeWorks and Deformetrica models are more consistent compared to models from SPHARM-PDM due to the groupwise approach of estimating surface correspondences. Furthermore, ShapeWorks and Deformetrica shape models are found to capture clinically relevant population-level variability compared to SPHARM-PDM models.
△ Less
Submitted 6 September, 2020;
originally announced September 2020.
-
Learning distributions of shape trajectories from longitudinal datasets: a hierarchical model on a manifold of diffeomorphisms
Authors:
Alexandre Bône,
Olivier Colliot,
Stanley Durrleman
Abstract:
We propose a method to learn a distribution of shape trajectories from longitudinal data, i.e. the collection of individual objects repeatedly observed at multiple time-points. The method allows to compute an average spatiotemporal trajectory of shape changes at the group level, and the individual variations of this trajectory both in terms of geometry and time dynamics. First, we formulate a non-…
▽ More
We propose a method to learn a distribution of shape trajectories from longitudinal data, i.e. the collection of individual objects repeatedly observed at multiple time-points. The method allows to compute an average spatiotemporal trajectory of shape changes at the group level, and the individual variations of this trajectory both in terms of geometry and time dynamics. First, we formulate a non-linear mixed-effects statistical model as the combination of a generic statistical model for manifold-valued longitudinal data, a deformation model defining shape trajectories via the action of a finite-dimensional set of diffeomorphisms with a manifold structure, and an efficient numerical scheme to compute parallel transport on this manifold. Second, we introduce a MCMC-SAEM algorithm with a specific approach to shape sampling, an adaptive scheme for proposal variances, and a log-likelihood tempering strategy to estimate our model. Third, we validate our algorithm on 2D simulated data, and then estimate a scenario of alteration of the shape of the hippocampus 3D brain structure during the course of Alzheimer's disease. The method shows for instance that hippocampal atrophy progresses more quickly in female subjects, and occurs earlier in APOE4 mutation carriers. We finally illustrate the potential of our method for classifying pathological trajectories versus normal ageing.
△ Less
Submitted 13 June, 2018; v1 submitted 27 March, 2018;
originally announced March 2018.
-
Parallel transport in shape analysis: a scalable numerical scheme
Authors:
Maxime Louis,
Alexandre Bône,
Benjamin Charlier,
Stanley Durrleman
Abstract:
The analysis of manifold-valued data requires efficient tools from Riemannian geometry to cope with the computational complexity at stake. This complexity arises from the always-increasing dimension of the data, and the absence of closed-form expressions to basic operations such as the Riemannian logarithm. In this paper, we adapt a generic numerical scheme recently introduced for computing parall…
▽ More
The analysis of manifold-valued data requires efficient tools from Riemannian geometry to cope with the computational complexity at stake. This complexity arises from the always-increasing dimension of the data, and the absence of closed-form expressions to basic operations such as the Riemannian logarithm. In this paper, we adapt a generic numerical scheme recently introduced for computing parallel transport along geodesics in a Riemannian manifold to finite-dimensional manifolds of diffeomorphisms. We provide a qualitative and quantitative analysis of its behavior on high-dimensional manifolds, and investigate an application with the prediction of brain structures progression.
△ Less
Submitted 23 November, 2017;
originally announced November 2017.
-
Prediction of the progression of subcortical brain structures in Alzheimer's disease from baseline
Authors:
Alexandre Bône,
Maxime Louis,
Alexandre Routier,
Jorge Samper,
Michael Bacci,
Benjamin Charlier,
Olivier Colliot,
Stanley Durrleman
Abstract:
We propose a method to predict the subject-specific longitudinal progression of brain structures extracted from baseline MRI, and evaluate its performance on Alzheimer's disease data. The disease progression is modeled as a trajectory on a group of diffeomorphisms in the context of large deformation diffeomorphic metric mapping (LDDMM). We first exhibit the limited predictive abilities of geodesic…
▽ More
We propose a method to predict the subject-specific longitudinal progression of brain structures extracted from baseline MRI, and evaluate its performance on Alzheimer's disease data. The disease progression is modeled as a trajectory on a group of diffeomorphisms in the context of large deformation diffeomorphic metric mapping (LDDMM). We first exhibit the limited predictive abilities of geodesic regression extrapolation on this group. Building on the recent concept of parallel curves in shape manifolds, we then introduce a second predictive protocol which personalizes previously learned trajectories to new subjects, and investigate the relative performances of two parallel shifting paradigms. This design only requires the baseline imaging data. Finally, coefficients encoding the disease dynamics are obtained from longitudinal cognitive measurements for each subject, and exploited to refine our methodology which is demonstrated to successfully predict the follow-up visits.
△ Less
Submitted 23 November, 2017;
originally announced November 2017.
