Search

Keratitis is an inflammatory corneal condition responsible for 10% of visual impairment in low- and middle-income countries (LMICs), with bacteria, fungi, or amoeba as the most common infection etiologies. While an accurate and timely diagnosis is crucial for the selected treatment and the patients' sight outcomes, due to the high cost and limited availability of laboratory diagnostics in LMICs, diagnosis is often made by clinical observation alone, despite its lower accuracy. In this study, we investigate and compare different deep learning approaches to diagnose the source of infection: 1) three separate binary models for infection type predictions; 2) a multitask model with a shared backbone and three parallel classification layers (Multitask V1); and, 3) a multitask model with a shared backbone and a multi-head classification layer (Multitask V2). We used a private Brazilian cornea dataset to conduct the empirical evaluation. We achieved the best results with Multitask V2, with an area under the receiver operating characteristic curve (AUROC) confidence intervals of 0.7413-0.7740 (bacteria), 0.8395-0.8725 (fungi), and 0.9448-0.9616 (amoeba). A statistical analysis of the impact of patient features on models' performance revealed that sex significantly affects amoeba infection prediction, and age seems to affect fungi and bacteria predictions., Comment: 6 pages; Accepted at IEEE's International Conference on Bioinformatics and Biomedicine (2024)

Algorithmic bias in healthcare mirrors existing data biases. However, the factors driving unfairness are not always known. Medical devices capture significant amounts of data but are prone to errors; for instance, pulse oximeters overestimate the arterial oxygen saturation of darker-skinned individuals, leading to worse outcomes. The impact of this bias in machine learning (ML) models remains unclear. This study addresses the technical challenges of quantifying the impact of medical device bias in downstream ML. Our experiments compare a "perfect world", without pulse oximetry bias, using SaO2 (blood-gas), to the "actual world", with biased measurements, using SpO2 (pulse oximetry). Under this counterfactual design, two models are trained with identical data, features, and settings, except for the method of measuring oxygen saturation: models using SaO2 are a "control" and models using SpO2 a "treatment". The blood-gas oximetry linked dataset was a suitable test-bed, containing 163,396 nearly-simultaneous SpO2 - SaO2 paired measurements, aligned with a wide array of clinical features and outcomes. We studied three classification tasks: in-hospital mortality, respiratory SOFA score in the next 24 hours, and SOFA score increase by two points. Models using SaO2 instead of SpO2 generally showed better performance. Patients with overestimation of O2 by pulse oximetry of > 3% had significant decreases in mortality prediction recall, from 0.63 to 0.59, P < 0.001. This mirrors clinical processes where biased pulse oximetry readings provide clinicians with false reassurance of patients' oxygen levels. A similar degradation happened in ML models, with pulse oximetry biases leading to more false negatives in predicting adverse outcomes., Comment: 10 pages; accepted at MICCAI's Third Workshop on Applications of Medical AI (2024)

The interactions between tumor cells and the tumor microenvironment (TME) dictate therapeutic efficacy of radiation and many systemic therapies in breast cancer. However, to date, there is not a widely available method to reproducibly measure tumor and immune phenotypes for each patient's tumor. Given this unmet clinical need, we applied multiple instance learning (MIL) algorithms to assess activity of ten biologically relevant pathways from the hematoxylin and eosin (H&E) slide of primary breast tumors. We employed different feature extraction approaches and state-of-the-art model architectures. Using binary classification, our models attained area under the receiver operating characteristic (AUROC) scores above 0.70 for nearly all gene expression pathways and on some cases, exceeded 0.80. Attention maps suggest that our trained models recognize biologically relevant spatial patterns of cell sub-populations from H&E. These efforts represent a first step towards developing computational H&E biomarkers that reflect facets of the TME and hold promise for augmenting precision oncology., Comment: Paper accepted at the First Workshop on Imageomics (Imageomics-AAAI-24) - Discovering Biological Knowledge from Images using AI (https://sites.google.com/vt.edu/imageomics-aaai-24/home), held as part of the 38th Annual AAAI Conference on Artificial Intelligence (https://aaai.org/aaai-conference/)

Face recognition applications have grown in parallel with the size of datasets, complexity of deep learning models and computational power. However, while deep learning models evolve to become more capable and computational power keeps increasing, the datasets available are being retracted and removed from public access. Privacy and ethical concerns are relevant topics within these domains. Through generative artificial intelligence, researchers have put efforts into the development of completely synthetic datasets that can be used to train face recognition systems. Nonetheless, the recent advances have not been sufficient to achieve performance comparable to the state-of-the-art models trained on real data. To study the drift between the performance of models trained on real and synthetic datasets, we leverage a massive attribute classifier (MAC) to create annotations for four datasets: two real and two synthetic. From these annotations, we conduct studies on the distribution of each attribute within all four datasets. Additionally, we further inspect the differences between real and synthetic datasets on the attribute set. When comparing through the Kullback-Leibler divergence we have found differences between real and synthetic samples. Interestingly enough, we have verified that while real samples suffice to explain the synthetic distribution, the opposite could not be further from being true., Comment: Accepted at FG 2024

We study a symmetric teleparallel gravity with a Lagrangian of logarithmic form. The full model leads to an accelerated universe and for specific values of the free parameters the Hubble rate reduces to the well-known Dvali-Gabadadze-Porrati model, though the evolution of the gravitational potentials are different. We consider different branches of the logarithmic model, among which are self-accelerated branch and normal branch. The phenomenology of both the background and linear perturbations is discussed, including all the relevant effects on cosmic microwave background radiation (CMB) angular power spectrum, lensing and matter power spectra. To this purpose, we modified the Einstein-Boltzmann code mgcamb. Finally, we derive bounds on the free parameters which are in agreement with early dark energy constraint from CMB and big bang nucleosynthesis constraint on the helium abundance., Comment: 24 pages, 14 figures

Morphing attacks keep threatening biometric systems, especially face recognition systems. Over time they have become simpler to perform and more realistic, as such, the usage of deep learning systems to detect these attacks has grown. At the same time, there is a constant concern regarding the lack of interpretability of deep learning models. Balancing performance and interpretability has been a difficult task for scientists. However, by leveraging domain information and proving some constraints, we have been able to develop IDistill, an interpretable method with state-of-the-art performance that provides information on both the identity separation on morph samples and their contribution to the final prediction. The domain information is learnt by an autoencoder and distilled to a classifier system in order to teach it to separate identity information. When compared to other methods in the literature it outperforms them in three out of five databases and is competitive in the remaining., Comment: Accepted at EUSIPCO 2023

In this work, we use the dynamical system approach to explore the cosmological background evolution of the scalar-tensor representation of $f(R,T)$ gravity, where $R$ is the Ricci scalar and $T$ is the trace of the stress-energy tensor. The motivation for this work resides in finding dynamical cosmological behaviors comparable with the $\Lambda$CDM model without the necessity of recurring to a dark energy component. We introduce a set of dynamical variables that allow for a direct comparison with the cosmological standard model and the current experimental measurements, and develop a dynamical system framework to analyze the cosmological evolution of Friedmann-Lema\^itre-Robertson-Walker (FLRW) universes within this theory. In this framework, we obtain the critical points in the cosmological phase space and perform fully numerical integrations of the dynamical system to extract the cosmological behavior, subjected to initial conditions compatible with the measurements by the Planck satellite. The phase space of the theory is proven to feature fixed points associated with cosmological behaviors analogous to those of GR, whereas variations in the scalar field associated to the dependency in $T$ affect the phase space structure only quantitatively. Our results indicate that cosmological solutions featuring a radiation dominated epoch, followed by a transition into a matter dominated epoch, and finally a transition into an exponentially accelerated epoch, are allowed by the theory, while maintaining a present state compatible with the current measurements from the Planck satellite and solar system dynamics, and preserving the regularity of the scalar fields and their interaction potential., Comment: 18 pages, 8 figures, 6 tables. Substantial changes made to first version

In this paper we explore the effects of quasiperiodicity in paradigmatic models of Chern insulators. We identify a plethora of topological phase transitions and characterize them based on spectral and localization properties. Contrary to uncorrelated disorder, gap closing and reopening topological transitions can be induced by quasiperiodicity. These can separate widely different phases, including (i) trivial and Chern insulators, both with ballistic states near the gap edges; (ii) Chern insulators with critical states around the gap edges or (iii) Chern and trivial insulators respectively with ballistic and localized gap-edge states. Transition (i) is similar to clean-limit topological transitions due to the ballistic character of the gap-edge states, but at the same time resembles (quasi)disorder driven topological Anderson insulator phenomena. On the other hand, transitions (ii) and (iii) have no clean-limit counterpart. Additionally, quasiperiodicity can also induce topological transitions into a trivial state for which the gap closes and does not reopen, a scenario that resembles more what is observed with uncorrelated disorder. However, we found that such transitions can also be non-conventional in that they can be accompanied by the emergence of intermediate metallic and critical phases where the Chern number is not quantized. Our results show that a rich variety of topological phase transitions, not previously realized experimentally nor predicted theoretically can be attained when applying quasiperiodic modulations to simple Chern insulators. Such models have previously been realized experimentally in widely different platforms, including in optical lattices and photonic or acoustic media, where quasiperiodicity effects can be incorporated. The unveiled topological phase transitions can in principle be observed experimentally with state-of-the-art techniques., Comment: Preliminary work 10 pages, 7 figures

As two sides of the same coin, causality and explainable artificial intelligence (xAI) were initially proposed and developed with different goals. However, the latter can only be complete when seen through the lens of the causality framework. As such, we propose a novel causality-inspired framework for xAI that creates an environment for the development of xAI approaches. To show its applicability, biometrics was used as case study. For this, we have analysed 81 research papers on a myriad of biometric modalities and different tasks. We have categorised each of these methods according to our novel xAI Ladder and discussed the future directions of the field.

Morphing attacks are one of the many threats that are constantly affecting deep face recognition systems. It consists of selecting two faces from different individuals and fusing them into a final image that contains the identity information of both. In this work, we propose a novel regularisation term that takes into account the existent identity information in both and promotes the creation of two orthogonal latent vectors. We evaluate our proposed method (OrthoMAD) in five different types of morphing in the FRLL dataset and evaluate the performance of our model when trained on five distinct datasets. With a small ResNet-18 as the backbone, we achieve state-of-the-art results in the majority of the experiments, and competitive results in the others. The code of this paper will be publicly available., Comment: Accepted at BIOSIG 2022

This paper presents a summary of the Competition on Face Morphing Attack Detection Based on Privacy-aware Synthetic Training Data (SYN-MAD) held at the 2022 International Joint Conference on Biometrics (IJCB 2022). The competition attracted a total of 12 participating teams, both from academia and industry and present in 11 different countries. In the end, seven valid submissions were submitted by the participating teams and evaluated by the organizers. The competition was held to present and attract solutions that deal with detecting face morphing attacks while protecting people's privacy for ethical and legal reasons. To ensure this, the training data was limited to synthetic data provided by the organizers. The submitted solutions presented innovations that led to outperforming the considered baseline in many experimental settings. The evaluation benchmark is now available at: https://github.com/marcohuber/SYN-MAD-2022., Comment: Accepted at International Joint Conference on Biometrics (IJCB) 2022

The increasing popularity of attention mechanisms in deep learning algorithms for computer vision and natural language processing made these models attractive to other research domains. In healthcare, there is a strong need for tools that may improve the routines of the clinicians and the patients. Naturally, the use of attention-based algorithms for medical applications occurred smoothly. However, being healthcare a domain that depends on high-stake decisions, the scientific community must ponder if these high-performing algorithms fit the needs of medical applications. With this motto, this paper extensively reviews the use of attention mechanisms in machine learning (including Transformers) for several medical applications. This work distinguishes itself from its predecessors by proposing a critical analysis of the claims and potentialities of attention mechanisms presented in the literature through an experimental case study on medical image classification with three different use cases. These experiments focus on the integrating process of attention mechanisms into established deep learning architectures, the analysis of their predictive power, and a visual assessment of their saliency maps generated by post-hoc explanation methods. This paper concludes with a critical analysis of the claims and potentialities presented in the literature about attention mechanisms and proposes future research lines in medical applications that may benefit from these frameworks., Comment: Pre-print submitted to Nature Scientific Reports

In this work, we study the possibility of finite-time future cosmological singularities appearing in $f(R,T)$ gravity, where $R$ is the Ricci scalar and $T$ is the trace of the stress-energy tensor. We present the theory in both the geometrical and the dynamically equivalent scalar-tensor representation and obtain the respective equations of motion. In a background Friedmann-Lema\^{i}tre-Robertson-Walker (FLRW) universe with an arbitrary curvature and for a generic $C^\infty$ function $f(R,T)$, we prove that the conservation of the stress-energy tensor prevents the appearance of sudden singularities in the cosmological context at any order in the time-derivatives of the scale factor. However, if this assumption is dropped, the theory allows for sudden singularities to appear at the level of the third time-derivative of the scale factor $a(t)$, which are compensated by divergences in either the first time-derivatives of the energy density $\rho(t)$ or the isotropic pressure $p(t)$. For these cases, we introduce a cosmological model featuring a sudden singularity that is consistent with the current measurements for the cosmological parameters, namely, the Hubble constant, deceleration parameter, and age of the universe, and provide predictions for the still unmeasured jerk and snap parameters. Finally, we analyse the constraints on a particular model of the function $f(R,T)$ that guarantees that the system evolves in a direction favorable to the energy conditions at the divergence time., Comment: 14 pages, 4 figures; v2: Matches published version in EPJC

We apply cosmological reconstruction methods to the $f(R,T)$ modified gravity, in its recently developed scalar-tensor representation. We do this analysis assuming a perfect fluid in a Friedmann-Lema\^{i}tre-Robsertson-Walker (FLRW) universe. In this contribution we show the equations of motion obtained and we present the solutions found for one of the particular cases we analysed: an exponential evolution of the cosmological scale factor., Comment: 11 pages, 3 figures; Contribution to the Sixteenth Marcel Grossman Conference (MG16, Rome, July, 2021), to be published by World Scientific; based on arXiv:2112.02541 [gr-qc]

In this work, we use reconstruction methods to obtain cosmological solutions in the recently developed scalar-tensor representation of $f(R,T)$ gravity. Assuming that matter is described by an isotropic perfect fluid and the spacetime is homogeneous and isotropic, i.e., the Friedmann-Lema\^itre-Robsertson-Walker (FLRW) universe, the energy density, the pressure, and the scalar field associated with the arbitrary dependency of the action in $T$ can be written generally as functions of the scale factor. We then select three particular forms of the scale factor: an exponential expansion with ${a(t)\propto e^t}$ (motivated by the de Sitter solution); and two types of power-law expansion with ${a(t)\propto t^{1/2}}$ and ${a(t)\propto t^{2/3}}$ (motivated by the behaviors of radiation- and matter-dominated universes in general relativity, respectively). A complete analysis for different curvature parameters ${k=\{-1,0,1\}}$ and equation of state parameters ${w=\{-1,0,1/3\}}$ is provided. Finally, the explicit forms of the functions $f\left(R,T\right)$ associated with the scalar-field potentials of the representation used are deduced., Comment: 14 pages, 18 figures

The imaging of active nanoparticles represents a milestone in decoding heterogeneous catalysts dynamics. We report the facet resolved, surface strain state of a single PtRh alloy nanoparticle on SrTiO3 determined by coherent x-ray diffraction imaging under catalytic reaction conditions. Density functional theory calculations allow us to correlate the facet surface strain state to its reaction environment dependent chemical composition. We find that the initially Pt terminated nanoparticle surface gets Rh enriched under CO oxidation reaction conditions. The local composition is facet orientation dependent and the Rh enrichment is non-reversible under subsequent CO reduction. Tracking facet resolved strain and composition under operando conditions is crucial for a rational design of more efficient heterogeneous catalysts with tailored activity, selectivity and lifetime., Comment: 15 pages, 4 figures, 32 references