RESUMO
Dementia arises from various brain-affecting diseases and injuries, with Alzheimer's disease being the most prevalent, impacting around 55 million people globally. Current clinical diagnosis often relies on biomarkers indicative of Alzheimer's distinctive features. Electroencephalography (EEG) serves as a cost-effective, user-friendly, and safe biomarker for early Alzheimer's detection. This study utilizes EEG signals processed with Short-Time Fourier Transform (STFT) to generate spectrograms, facilitating visualization of EEG signal properties. Leveraging the Brainlat database, we propose SpectroCVT-Net, a novel convolutional vision transformer architecture incorporating channel attention mechanisms. SpectroCVT-Net integrates convolutional and attention mechanisms to capture local and global dependencies within spectrograms. Comprising feature extraction and classification stages, the model enhances Alzheimer's disease classification accuracy compared to transfer learning methods, achieving 92.59 ± 2.3% accuracy across Alzheimer's, healthy controls, and behavioral variant frontotemporal dementia (bvFTD). This article introduces a new architecture and evaluates its efficacy with unconventional data for Alzheimer's diagnosis, contributing: SpectroCVT-Net, tailored for EEG spectrogram classification without reliance on transfer learning; a convolutional vision transformer (CVT) module in the classification stage, integrating local feature extraction with attention heads for global context analysis; Grad-CAM analysis for network decision insight, identifying critical layers, frequencies, and electrodes influencing classification; and enhanced interpretability through spectrograms, illuminating key brain wave contributions to Alzheimer's, frontotemporal dementia, and healthy control classifications, potentially aiding clinical diagnosis and management.
RESUMO
INTRODUCTION: Epilepsy is a disease characterized by an excessive discharge in neurons generally provoked without any external stimulus, known as convulsions. About 2 million people are diagnosed each year in the world. This process is carried out by a neurological doctor using an electroencephalogram (EEG), which is lengthy. METHOD: To optimize these processes and make them more efficient, we have resorted to innovative artificial intelligence methods essential in classifying EEG signals. For this, comparing traditional models, such as machine learning or deep learning, with cutting-edge models, in this case, using Capsule-Net architectures and Transformer Encoder, has a crucial role in finding the most accurate model and helping the doctor to have a faster diagnosis. RESULT: In this paper, a comparison was made between different models for binary and multiclass classification of the epileptic seizure detection database, achieving a binary accuracy of 99.92% with the Capsule-Net model and a multiclass accuracy with the Transformer Encoder model of 87.30%. CONCLUSION: Artificial intelligence is essential in diagnosing pathology. The comparison between models is helpful as it helps to discard those that are not efficient. State-of-the-art models overshadow conventional models, but data processing also plays an essential role in evaluating the higher accuracy of the models.
Assuntos
Inteligência Artificial , Epilepsia , Humanos , Epilepsia/diagnóstico , Convulsões/diagnóstico , Algoritmos , Aprendizado de Máquina , EletroencefalografiaRESUMO
We explore the data-parallel acceleration of physics-informed machine learning (PIML) schemes, with a focus on physics-informed neural networks (PINNs) for multiple graphics processing units (GPUs) architectures. In order to develop scale-robust and high-throughput PIML models for sophisticated applications which may require a large number of training points (e.g., involving complex and high-dimensional domains, non-linear operators or multi-physics), we detail a novel protocol based on h-analysis and data-parallel acceleration through the Horovod training framework. The protocol is backed by new convergence bounds for the generalization error and the train-test gap. We show that the acceleration is straightforward to implement, does not compromise training, and proves to be highly efficient and controllable, paving the way towards generic scale-robust PIML. Extensive numerical experiments with increasing complexity illustrate its robustness and consistency, offering a wide range of possibilities for real-world simulations.
RESUMO
Alzheimer's disease (AD) is a progressive type of dementia characterized by loss of memory and other cognitive abilities, including speech. Since AD is a progressive disease, detection in the early stages is essential for the appropriate care of the patient throughout its development, going from asymptomatic to a stage known as mild cognitive impairment (MCI), and then progressing to dementia and severe dementia; is worth mentioning that everyone suffers from cognitive impairment to some degree as we age, but the relevant task here is to identify which people are most likely to develop AD. Along with cognitive tests, evaluation of the brain morphology is the primary tool for AD diagnosis, where atrophy and loss of volume of the frontotemporal lobe are common features in patients who suffer from the disease. Regarding medical imaging techniques, magnetic resonance imaging (MRI) scans are one of the methods used by specialists to assess brain morphology. Recently, with the rise of deep learning (DL) and its successful implementation in medical imaging applications, it is of growing interest in the research community to develop computer-aided diagnosis systems that can help physicians to detect this disease, especially in the early stages where macroscopic changes are not so easily identified. This article presents a DL-based approach to classifying MRI scans in the different stages of AD, using a curated set of images from Alzheimer's Disease Neuroimaging Initiative and Open Access Series of Imaging Studies databases. Our methodology involves image pre-processing using FreeSurfer, spatial data-augmentation operations, such as rotation, flip, and random zoom during training, and state-of-the-art 3D convolutional neural networks such as EfficientNet, DenseNet, and a custom siamese network, as well as the relatively new approach of vision transformer architecture. With this approach, the best detection percentage among all four architectures was around 89% for AD vs. Control, 80% for Late MCI vs. Control, 66% for MCI vs. Control, and 67% for Early MCI vs. Control.
RESUMO
An evolutionary computation framework to learn binary threshold networks is presented. Inspired by the recent trend of binary neural networks, where weights and activation thresholds are represented using 1 and -1 such that they can be stored in 1-bit instead of full precision, we explore this approach for gene regulatory network modeling. We test our method by inferring binary threshold networks of two regulatory network models: Quorum sensing systems in bacterium Paraburkholderia phytofirmans PsJN and the fission yeast cell-cycle. We considered differential evolution and particle swarm optimization for the simulations. Results for weights having only 1 and -1 values, and different activation thresholds are presented. Full binary threshold networks were found with minimum error (2 bits), whereas when the binary restriction is relaxed for the activation thresholds, networks with 0 bit error were found.
Assuntos
Redes Reguladoras de Genes , Redes Neurais de Computação , Redes Reguladoras de Genes/genética , Percepção de Quorum/genética , AlgoritmosRESUMO
Several artificial intelligence algorithms have been developed for COVID-19-related topics. One that has been common is the COVID-19 diagnosis using chest X-rays, where the eagerness to obtain early results has triggered the construction of a series of datasets where bias management has not been thorough from the point of view of patient information, capture conditions, class imbalance, and careless mixtures of multiple datasets. This paper analyses 19 datasets of COVID-19 chest X-ray images, identifying potential biases. Moreover, computational experiments were conducted using one of the most popular datasets in this domain, which obtains a 96.19% of classification accuracy on the complete dataset. Nevertheless, when evaluated with the ethical tool Aequitas, it fails on all the metrics. Ethical tools enhanced with some distribution and image quality considerations are the keys to developing or choosing a dataset with fewer bias issues. We aim to provide broad research on dataset problems, tools, and suggestions for future dataset developments and COVID-19 applications using chest X-ray images.
Assuntos
Inteligência Artificial , COVID-19 , Humanos , Teste para COVID-19 , Raios X , ViésRESUMO
BACKGROUND: When designing a treatment in orthodontics, especially for children and teenagers, it is crucial to be aware of the changes that occur throughout facial growth because the rate and direction of growth can greatly affect the necessity of using different treatment mechanics. This paper presents a Bayesian network approach for facial biotype classification to classify patients' biotypes into Dolichofacial (long and narrow face), Brachyfacial (short and wide face), and an intermediate kind called Mesofacial, we develop a novel learning technique for tree augmented Naive Bayes (TAN) for this purpose. RESULTS: The proposed method, on average, outperformed all the other models based on accuracy, precision, recall, [Formula: see text], and kappa, for the particular dataset analyzed. Moreover, the proposed method presented the lowest dispersion, making this model more stable and robust against different runs. CONCLUSIONS: The proposed method obtained high accuracy values compared to other competitive classifiers. When analyzing a resulting Bayesian network, many of the interactions shown in the network had an orthodontic interpretation. For orthodontists, the Bayesian network classifier can be a helpful decision-making tool.
Assuntos
Algoritmos , Conscientização , Criança , Adolescente , Humanos , Teorema de BayesRESUMO
The prognostics and health management disciplines provide an efficient solution to improve a system's durability, taking advantage of its lifespan in functionality before a failure appears. Prognostics are performed to estimate the system or subsystem's remaining useful life (RUL). This estimation can be used as a supply in decision-making within maintenance plans and procedures. This work focuses on prognostics by developing a recurrent neural network and a forecasting method called Prophet to measure the performance quality in RUL estimation. We apply this approach to degradation signals, which do not need to be monotonical. Finally, we test our system using data from new generation telescopes in real-world applications.
Assuntos
Análise de Falha de Equipamento , Redes Neurais de Computação , Análise de Falha de Equipamento/métodosRESUMO
BACKGROUND: An important process for plant survival is the immune system. The induced systemic resistance (ISR) triggered by beneficial microbes is an important cost-effective defense mechanism by which plants are primed to an eventual pathogen attack. Defense mechanisms such as ISR depend on an accurate and context-specific regulation of gene expression. Interactions between genes and their products give rise to complex circuits known as gene regulatory networks (GRNs). Here, we explore the regulatory mechanism of the ISR defense response triggered by the beneficial bacterium Paraburkholderia phytofirmans PsJN in Arabidopsis thaliana plants infected with Pseudomonas syringae DC3000. To achieve this, a GRN underlying the ISR response was inferred using gene expression time-series data of certain defense-related genes, differential evolution, and threshold Boolean networks. RESULTS: One thousand threshold Boolean networks were inferred that met the restriction of the desired dynamics. From these networks, a consensus network was obtained that helped to find plausible interactions between the genes. A representative network was selected from the consensus network and biological restrictions were applied to it. The dynamics of the selected network showed that the largest attractor, a limit cycle of length 3, represents the final stage of the defense response (12, 18, and 24 h). Also, the structural robustness of the GRN was studied through the networks' attractors. CONCLUSIONS: A computational intelligence approach was designed to reconstruct a GRN underlying the ISR defense response in plants using gene expression time-series data of A. thaliana colonized by P. phytofirmans PsJN and subsequently infected with P. syringae DC3000. Using differential evolution, 1000 GRNs from time-series data were successfully inferred. Through the study of the network dynamics of the selected GRN, it can be concluded that it is structurally robust since three mutations were necessary to completely disarm the Boolean trajectory that represents the biological data. The proposed method to reconstruct GRNs is general and can be used to infer other biologically relevant networks to formulate new biological hypotheses.
Assuntos
Arabidopsis/genética , Arabidopsis/microbiologia , Resistência à Doença/genética , Redes Reguladoras de Genes , Burkholderiaceae/fisiologia , Pseudomonas syringaeRESUMO
Plant defense responses to biotic stresses are complex biological processes, all governed by sophisticated molecular regulations. Induced systemic resistance (ISR) is one of these defense mechanisms where beneficial bacteria or fungi prime plants to resist pathogens or pest attacks. In ISR, the defense arsenal in plants remains dormant and it is only triggered by an infection, allowing a better allocation of plant resources. Our group recently described that the well-known beneficial bacterium Paraburkholderia phytofirmans PsJN is able to induce Arabidopsis thaliana resistance to Pseudomonas syringae pv. tomato (Pst) DC3000 through ISR, and that ethylene, jasmonate and salicylic acid are involved in this protection. Nevertheless, the molecular networks governing this beneficial interaction remain unknown. To tackle this issue, we analyzed the temporal changes in the transcriptome of PsJN-inoculated plants before and after being infected with Pst DC3000. These data were used to perform a gene network analysis to identify highly connected transcription factors. Before the pathogen challenge, the strain PsJN regulated 405 genes (corresponding to 1.8% of the analyzed genome). PsJN-inoculated plants presented a faster and stronger transcriptional response at 1-hour post infection (hpi) compared with the non-inoculated plants, which presented the highest transcriptional changes at 24 hpi. A principal component analysis showed that PsJN-induced plant responses to the pathogen could be differentiated from those induced by the pathogen itself. Forty-eight transcription factors were regulated by PsJN at 1 hpi, and a system biology analysis revealed a network with four clusters. Within these clusters LHY, WRKY28, MYB31 and RRTF1 are highly connected transcription factors, which could act as hub regulators in this interaction. Concordantly with our previous results, these clusters are related to jasmonate, ethylene, salicylic, acid and ROS pathways. These results indicate that a rapid and specific response of PsJN-inoculated plants to the virulent DC3000 strain could be the pivotal element in the protection mechanism.
Assuntos
Arabidopsis/genética , Burkholderiaceae/fisiologia , Regulação da Expressão Gênica de Plantas/imunologia , Doenças das Plantas/genética , Pseudomonas syringae/patogenicidade , Fatores de Transcrição/genética , Arabidopsis/imunologia , Arabidopsis/microbiologia , Ciclopentanos/imunologia , Ciclopentanos/metabolismo , Resistência à Doença/genética , Etilenos/imunologia , Etilenos/metabolismo , Perfilação da Expressão Gênica , Redes Reguladoras de Genes/imunologia , Oxilipinas/imunologia , Oxilipinas/metabolismo , Doenças das Plantas/imunologia , Doenças das Plantas/microbiologia , Reguladores de Crescimento de Plantas/imunologia , Reguladores de Crescimento de Plantas/metabolismo , Imunidade Vegetal/genética , Análise de Componente Principal , Pseudomonas syringae/crescimento & desenvolvimento , Ácido Salicílico/imunologia , Ácido Salicílico/metabolismo , Fatores de Transcrição/imunologia , Transcriptoma/imunologiaRESUMO
In intertidal marine crustaceans, phenotypic variation in physiological and life-history traits is pervasive along latitudinal clines. However, organisms have complex phenotypes, and their traits do not vary independently but rather interact differentially between them, effect that is caused by genetic and/or environmental forces. We evaluated the geographic variation in phenotypic integration of three marine crab species that inhabit different vertical thermal microhabitats of the intertidal zone. We studied seven populations of each species along a latitudinal gradient that spans more than 3000â¯km of the Chilean coast. Specifically we measured nine physiological traits that are highly related to thermal physiology. Of the nine traits, we selected four that contributed significantly to the observed geographical variation among populations; this variation was then evaluated using mixed linear models and an integrative approach employing machine learning. The results indicate that patterns of physiological variation depend on species vertical microhabitat, which may be subject to chronic or acute environmental variation. The species that inhabit the high- intertidal sites (i.e., exposed to chronic variation) better tolerated thermal stress compared with populations that inhabit the lower intertidal. While those in the low-intertidal only face conditions of acute thermal variation, using to a greater extent the plasticity to face these events. Our main results reflect that (1) species that inhabit the high-intertidal maintain a greater integration between their physiological traits and present lower plasticity than those that inhabit the low-intertidal. (2) Inverse relationship that exists between phenotypic plasticity and phenotypic integration of the physiological traits identified, which could help optimize energy resources. In general, the study of multiple physiological traits provides a more accurate picture of how the thermal traits of organisms vary along temperature gradients especially when exposed to conditions close to tolerance limits.
Assuntos
Aclimatação , Temperatura Corporal , Crustáceos/fisiologia , Ecótipo , Animais , Crustáceos/genética , Aprendizado de MáquinaRESUMO
Background: Long-read sequencing technologies are the ultimate solution for genome repeats, allowing near reference-level reconstructions of large genomes. However, long-read de novo assembly pipelines are computationally intense and require a considerable amount of coverage, thereby hindering their broad application to the assembly of large genomes. Alternatively, hybrid assembly methods that combine short- and long-read sequencing technologies can reduce the time and cost required to produce de novo assemblies of large genomes. Results: Here, we propose a new method, called Fast-SG, that uses a new ultrafast alignment-free algorithm specifically designed for constructing a scaffolding graph using light-weight data structures. Fast-SG can construct the graph from either short or long reads. This allows the reuse of efficient algorithms designed for short-read data and permits the definition of novel modular hybrid assembly pipelines. Using comprehensive standard datasets and benchmarks, we show how Fast-SG outperforms the state-of-the-art short-read aligners when building the scaffoldinggraph and can be used to extract linking information from either raw or error-corrected long reads. We also show how a hybrid assembly approach using Fast-SG with shallow long-read coverage (5X) and moderate computational resources can produce long-range and accurate reconstructions of the genomes of Arabidopsis thaliana (Ler-0) and human (NA12878). Conclusions: Fast-SG opens a door to achieve accurate hybrid long-range reconstructions of large genomes with low effort, high portability, and low cost.
Assuntos
Algoritmos , Sequenciamento de Nucleotídeos em Larga Escala/métodos , Arabidopsis/genética , Escherichia coli K12/genética , Biblioteca Gênica , Genoma Bacteriano , Genoma Humano , HumanosRESUMO
In the first decades of the 20th century, political actors diagnosed the incubation of a crisis in the Chilean schooling process. Low rates of enrollment, literacy, and attendance, inefficiency in the use of resources, poverty, and a reduced number of schools were the main factors explaining the crisis. As a response, the Law on Compulsory Primary Education, considering mandatory for children between 6 and 14 years old to attend any school for at least four years, was passed in 1920. Using data from Censuses of the Republic of Chile from 1920 and 1930, reports of the Ministry of Justice, the Ministry of Education, and the Statistical Yearbooks between 1895 and 1930, we apply machine learning techniques (clustering and decision trees) to assess the impact of this law on the Chilean schooling process between 1920 and 1930. We conclude that the law had a positive impact on the schooling indicators in this period. Even though it did not overcome the differences between urban and rural zones, it brought about a general improvement of the schooling process and a more efficient use of resources and infrastructure in both big urban centers and small-urban and rural zones, thereby managing the so-called crisis of the Republic.
Assuntos
Educação/história , Aprendizado de Máquina , Algoritmos , Chile , Análise por Conglomerados , Bases de Dados como Assunto , Árvores de Decisões , Geografia , História do Século XX , Análise de Componente Principal , Reprodutibilidade dos TestesRESUMO
Quorum-sensing systems play important roles in host colonization and host establishment of Burkholderiales species. Beneficial Paraburkholderia species share a conserved quorum-sensing (QS) system, designated BraI/R, that controls different phenotypes. In this context, the plant growth-promoting bacterium Paraburkholderia phytofirmans PsJN possesses two different homoserine lactone QS systems BpI.1/R.1 and BpI.2/R.2 (BraI/R-like QS system). The BpI.1/R.1 QS system was previously reported to be important to colonize and produce beneficial effects in Arabidopsis thaliana plants. Here, we analyzed the temporal variations of the QS gene transcript levels in the wild-type strain colonizing plant roots. The gene expression patterns showed relevant differences in both QS systems compared with the wild-type strain in the unplanted control treatment. The gene expression data were used to reconstruct a regulatory network model of QS systems in P. phytofirmans PsJN, using a Boolean network model. Also, we examined the phenotypic traits and transcript levels of genes involved in QS systems, using P. phytofirmans mutants in homoserine lactone synthases genes. We observed that the BpI.1/R.1 QS system regulates biofilm formation production in strain PsJN and this phenotype was associated with the lower expression of a specific extracytoplasmic function sigma factor ecf26.1 gene (implicated in biofilm formation) in the bpI.1 mutant strain.
Assuntos
Arabidopsis/crescimento & desenvolvimento , Biofilmes , Burkholderia/genética , Raízes de Plantas/crescimento & desenvolvimento , Percepção de Quorum/genética , 4-Butirolactona/análogos & derivados , 4-Butirolactona/metabolismo , Arabidopsis/microbiologia , Burkholderia/metabolismo , Burkholderia/fisiologia , Perfilação da Expressão Gênica/métodos , Regulação Bacteriana da Expressão Gênica , Redes Reguladoras de Genes , Interações Hospedeiro-Patógeno , Modelos Genéticos , Mutação , Raízes de Plantas/microbiologiaRESUMO
In this paper we study the dynamical behavior of neural networks such that their interconnections are the incidence matrix of an undirected finite graph G=(V,E) (i.e., the weights belong to {0,1}). The network may be updated synchronously (every node is updated at the same time), sequentially (nodes are updated one by one in a prescribed order) or in a block-sequential way (a mixture of the previous schemes). We characterize completely the attractors (fixed points or cycles). More precisely, we establish the convergence to fixed points related to a parameter α(G), taking into account the number of loops, edges, vertices as well as the minimum number of edges to remove from E in order to obtain a maximum bipartite graph. Roughly, α(G')<0 for any G' subgraph of G implies the convergence to fixed points. Otherwise, cycles appear. Actually, for very simple networks (majority functions updated in a block-sequential scheme such that each block is of minimum cardinality two) we exhibit cycles with non-polynomial periods.
Assuntos
Algoritmos , Redes Neurais de ComputaçãoRESUMO
A common gene regulatory network model is the threshold Boolean network, used for example to model the Arabidopsis thaliana floral morphogenesis network or the fission yeast cell cycle network. In this paper, we analyze a logical model of the mammalian cell cycle network and its threshold Boolean network equivalent. Firstly, the robustness of the network was explored with respect to update perturbations, in particular, what happened to the attractors for all the deterministic updating schemes. Results on the number of different limit cycles, limit cycle lengths, basin of attraction size, for all the deterministic updating schemes were obtained through mathematical and computational tools. Secondly, we analyzed the topology robustness of the network, by reconstructing synthetic networks that contained exactly the same attractors as the original model by means of a swarm intelligence approach. Our results indicate that networks may not be very robust given the great variety of limit cycles that a network can obtain depending on the updating scheme. In addition, we identified an omnipresent network with interactions that match with the original model as well as the discovery of new interactions. The techniques presented in this paper are general, and can be used to analyze other logical or threshold Boolean network models of gene regulatory networks.
Assuntos
Ciclo Celular/fisiologia , Redes Reguladoras de Genes/fisiologia , Mamíferos/fisiologia , Modelos Biológicos , Algoritmos , Animais , Biologia Computacional/métodos , Simulação por Computador , Biologia de Sistemas/métodosRESUMO
BACKGROUND: Interactions between genes and their products give rise to complex circuits known as gene regulatory networks (GRN) that enable cells to process information and respond to external stimuli. Several important processes for life, depend of an accurate and context-specific regulation of gene expression, such as the cell cycle, which can be analyzed through its GRN, where deregulation can lead to cancer in animals or a directed regulation could be applied for biotechnological processes using yeast. An approach to study the robustness of GRN is through the neutral space. In this paper, we explore the neutral space of a Schizosaccharomyces pombe (fission yeast) cell cycle network through an evolution strategy to generate a neutral graph, composed of Boolean regulatory networks that share the same state sequences of the fission yeast cell cycle. RESULTS: Through simulations it was found that in the generated neutral graph, the functional networks that are not in the wildtype connected component have in general a Hamming distance more than 3 with the wildtype, and more than 10 between the other disconnected functional networks. Significant differences were found between the functional networks in the connected component of the wildtype network and the rest of the network, not only at a topological level, but also at the state space level, where significant differences in the distribution of the basin of attraction for the G1 fixed point was found for deterministic updating schemes. CONCLUSIONS: In general, functional networks in the wildtype network connected component, can mutate up to no more than 3 times, then they reach a point of no return where the networks leave the connected component of the wildtype. The proposed method to construct a neutral graph is general and can be used to explore the neutral space of other biologically interesting networks, and also formulate new biological hypotheses studying the functional networks in the wildtype network connected component.