(Fonte: Lattes)
Índice h a partir de 2011
Projetos de Pesquisa
Unidades Organizacionais
Instituto do Coração, Hospital das Clínicas, Faculdade de Medicina
LIM/65, Hospital das Clínicas, Faculdade de Medicina - Líder

Resultados de Busca

Agora exibindo 1 - 10 de 73
  • article 5 Citação(ões) na Scopus
    Directional analysis of cardiac motion field from gated fluorodeoxyglucose PET images using the Discrete Helmholtz Hodge Decomposition
    Objectives: Extract directional information related to left ventricular (LV) rotation and torsion from a 4D PET motion field using the Discrete Helmholtz Hodge Decomposition (DHHD). Materials and methods: Synthetic motion fields were created using superposition of rotational and radial field components and cardiac fields produced using optical flow from a control and patient image. These were decomposed into curl-free (CF) and divergence-free (DF) components using the DHHD. Results: Synthetic radial components were present in the CF field and synthetic rotational components in the DF field, with each retaining its center position, direction of motion and diameter after decomposition. Direction of rotation at apex and base for the control field were in opposite directions during systole, reversing during diastole. The patient DF field had little overall rotation with several small rotators. Conclusions: The decomposition of the LV motion field into directional components could assist quantification of LV torsion, but further processing stages seem necessary.
  • conferenceObject
    Fully Automated Quantification of Cardiac Indices from Cine MRI Using a Combination of Convolution Neural Networks
    (2020) PEREIRA, Renato F.; REBELO, Marina S.; MORENO, Ramon A.; MARCO, Anderson G.; LIMA, Daniel M.; ARRUDA, Marcelo A. F.; KRIEGER, Jose E.; GUTIERREZ, Marco A.
    Cardiovascular magnetic resonance imaging (CMRI) is one of the most accurate non-invasive modalities for evaluation of cardiac function, especially the left ventricle (LV). In this modality, the manual or semi-automatic delineation of LV by experts is currently the standard clinical practice for chambers segmentation. Despite these efforts, global quantification of LV remains a challenge. In this work, a combination of two convolutional neural network (CNN) architectures for quantitative evaluation of the LV is described, which estimates the cavity and the myocardium areas, endocardial cavity dimensions in three directions, and the myocardium regional wall thickness in six radial directions. The method was validated in CMRI exams of 56 patients (LVQuan19 dataset) and evaluated by metrics Dice Index, Mean Absolute Error, and Correlation with superior performance compared to the state-of-the-art methods. The combination of the CNN architectures provided a simpler yet fully automated approach, requiring no specialist interaction.
  • conferenceObject 2 Citação(ões) na Scopus
    Evaluation of the impact of physical adversarial attacks on deep learning models for classifying covid cases
    (2022) AGUIAR, Erikson J. de; MARCOMINI, Karem D.; QUIRINO, Felipe A.; GUTIERREZ, Marco A.; TRAINA JR., Caetano; TRAINA, Agma J. M.
    The SARS-CoV-2 (COVID-19) disease rapidly spread worldwide, thus increasing the need to create new strategies to fight it. Several researchers in different fields have attempted to develop methods to early identifying it and mitigating its effects. The Deep Learning (DL) approach, such as the Convolutional Neural Networks (CNNs), has been increasingly used in COVID-19 diagnoses. These models intend to support decision-making and are doing well to detecting patient status early. Although DL models have good accuracy to support diagnosis, they are vulnerable to Adversarial Attacks. These attacks are new methods to make DL models biased by adding small perturbations on the original image. This paper investigates the impact of Adversarial Attacks on DL models for classifying X-ray images of COVID-19 cases. We focused on the attack Fast Gradient Sign Method (FGSM), which aims to add perturbations to the testing images by combining a perturbation matrix, producing a crafted image. We conduct the experiments analyzing the model's performance attack-free and adding attacks. The following CNNs models were selected: DenseNet201, ResNet-50V2, MobileNetV2, NasNet and VGG16. In the attack-free environment, we reach precision around 99%. When it adds the attack, our results revealed that all models suffer from performance reduction, and the most affected was MobileNet that reduced its ability from 98.61% to 67.73%. However, the VGG16 network showed to be the least affected by the attacks. Our finds describe that DL models for COVID-19 are vulnerable to Adversarial Examples. The FGSM was capable of fooling the model, resulting in a significant reduction in the DL performance.
  • conferenceObject
    Spatial-Frequency Approach to Fibrous Tissue Classification in Intracoronary Optical Images
    (2016) MACEDO, Maysa M. G.; NICZ, Pedro F. G.; CAMPOS, Carlos M.; LEMOS, Pedro A.; GUTIERREZ, Marco A.
    Increased understanding about the mechanisms of coronary thrombosis in humans has been limited by the lack of imaging modalities with resolution sufficient to characterize fibrous cap tissue and determine its thickness in vivo. Intravascular optical coherence tomography (IOCT) provides images with micrometer axial (10-15 mu m) and lateral resolution (40 mu m), enabling detailed visualization of micro-structural changes of the arterial wall. This article describes a fully automated method for identification and quantification of fibrous tissue in IOCT human coronary images based on spatial-frequency analysis by means Short-Time Fourier transform. Forty IOCT frames from nine IOCT in-vivo datasets were annotated by an expert and used to evaluate the proposed fibrous tissue characterization method.
  • conferenceObject
    Directional Analysis of Cardiac Motion Field based on the Discrete Helmholtz Hodge Decomposition
    (2016) SIMS, John A.; MACEDO, Maysa M. G.; GUTIERREZ, Marco A.
    The analysis of LV rotational motion could provide insights into myocardial dysfunction and predict the outcome of interventions, and this analysis could be performed more simply in separate rotational and radial components. In this study we present an automatic method for decomposing the cardiac motion field into radial and rotational components using the Discrete Helmholtz Hodge Decomposition (DHHD). The DHHD was applied to the following 3D motion fields (i) Synthetic complex motion fields, created by applying curl and gradient operators to Gaussian potentials, to determine numerical accuracy; (ii) Synthetic motion field from the 4D Extended Cardiac-Torso (XCAT) phantom (v2.0), to validate the use of the DHHD in decomposing cardiac motion fields. Decomposition error was found to decrease with increasing smoothness of the fields, while keeping motion field components small at the boundary of the motion field domain. The DHHD was seen to separate radial and rotational cardiac motion, allowing possible simplification of motion analysis.
  • conferenceObject
    Description of patellar movement by 3D parameters obtained from dynamic CT acquisition
    (2014) REBELO, Marina de Sa; MORENO, Ramon Alfredo; GOBBI, Riccardo Gomes; CAMANHO, Gilberto Luis; AVILA, Luiz Francisco Rodrigues de; DEMANGE, Marco Kawamura; PECORA, Jose Ricardo; GUTIERREZ, Marco Antonio
    The patellofemoral joint is critical in the biomechanics of the knee. The patellofemoral instability is one condition that generates pain, functional impairment and often requires surgery as part of orthopedic treatment. The analysis of the patellofemoral dynamics has been performed by several medical image modalities. The clinical parameters assessed are mainly based on 2D measurements, such as the patellar tilt angle and the lateral shift among others. Besides, the acquisition protocols are mostly performed with the leg laid static at fixed angles. The use of helical multi slice CT scanner can allow the capture and display of the joint's movement performed actively by the patient. However, the orthopedic applications of this scanner have not yet been standardized or widespread. In this work we present a method to evaluate the biomechanics of the patellofemoral joint during active contraction using multi slice CT images. This approach can greatly improve the analysis of patellar instability by displaying the physiology during muscle contraction. The movement was evaluated by computing its 3D displacements and rotations from different knee angles. The first processing step registered the images in both angles based on the femur's position. The transformation matrix of the patella from the images was then calculated, which provided the rotations and translations performed by the patella from its position in the first image to its position in the second image. Analysis of these parameters for all frames provided real 3D information about the patellar displacement.
  • article 17 Citação(ões) na Scopus
    Detecting Vascular Bifurcation in IVOCT Images Using Convolutional Neural Networks With Transfer Learning
    (2019) MIYAGAWA, Makoto; COSTA, Marly Guimaraes Fernandes; GUTIERREZ, Marco Antonio; COSTA, Joao Pedro Guimaraes Fernandes; COSTA FILHO, Cicero Ferreira Fernandes
    Optical coherence tomography (OCT) technology enables experts to analyze coronary lesions from high-resolution intravascular images. Studies have shown a relationship between vascular bifurcation and a higher occurrence of wall thickening and lesions in these areas. Some level of automation could benefit experts since the visual analysis of pullback frames is a laborious and time-consuming task. Although convolutional neural networks (CNNs) have shown promising results in classifying medical images, in this paper, we found no studies using CNNs in IVOCT images to classify the vascular bifurcation. In this paper, we evaluated four different CNN architectures in the bifurcation classification task trained with the IVOCT images from nine pullbacks from nine different patients. We used data augmentation to balance the dataset, due to the small number of bifurcation-labeled frames, and also applied transfer learning methods to incorporate the knowledge from a lumen segmentation task into some of the evaluated networks. Our classification outperforms other works in this literature, presenting AUC = 99.72%, obtained by a CNN with transferred knowledge.
  • conferenceObject
    IoT Medical Device Architecture to Estimate Non-invasive Arterial Blood Pressure
    (2022) MORENO, Ramon; DIAS, Felipe; ARRUDA, Marcelo; OLIVEIRA, Filipe; BULHOES, Thiago; KRIEGER, Jose; GUTIERREZ, Marco
    High blood pressure (BP) is the leading cause of death worldwide. Besides being a treatable condition, alongside medication and a healthy diet, it requires regular BP measurements to assess whether a patient is properly responding to treatment. There have been many attempts to use the photoplethysmography (PPG) signal to estimate BP continuously, but there has yet to be an effective solution. This work presents our efforts to develop a new method for estimating BP from PPG and infrastructure to collect, process, and store this information. PPG signal is measured from a smartband; our App reads the data from the smartband to a smartphone, processes them using a machine learning method, and estimates BP, which is sent to a server that stores and displays the data
  • conferenceObject
    CardioBERTpt: Transformer-based Models for Cardiology Language Representation in Portuguese
    (2023) SCHNEIDER, Elisa Terumi Rubel; GUMIEL, Yohan Bonescki; SOUZA, Joao Vitor Andrioli de; MUKAI, Lilian Mie; OLIVEIRA, Lucas Emanuel Silva e; REBELO, Marina de Sa; GUTIERREZ, Marco Antonio; KRIEGER, Jose Eduardo; TEODORO, Douglas; MORO, Claudia; PARAISO, Emerson Cabrera
    Contextual word embeddings and the Transformers architecture have reached state-of-the-art results in many natural language processing (NLP) tasks and improved the adaptation of models for multiple domains. Despite the improvement in the reuse and construction of models, few resources are still developed for the Portuguese language, especially in the health domain. Furthermore, the clinical models available for the language are not representative enough for all medical specialties. This work explores deep contextual embedding models for the Portuguese language to support clinical NLP tasks. We transferred learned information from electronic health records of a Brazilian tertiary hospital specialized in cardiology diseases and pre-trained multiple clinical BERT-based models. We evaluated the performance of these models in named entity recognition experiments, fine-tuning them in two annotated corpora containing clinical narratives. Our pre-trained models outperformed previous multilingual and Portuguese BERT-based models for cardiology and multi-specialty environments, reaching the state-of-the-art for analyzed corpora, with 5.5% F1 score improvement in TempClinBr (all entities) and 1.7% in SemClinBr (Disorder entity) corpora. Hence, we demonstrate that data representativeness and a high volume of training data can improve the results for clinical tasks, aligned with results for other languages.
  • conferenceObject
    Automated radiographic bone suppression with deep convolutional neural networks
    (2021) CARDENAS, Diego Armando Cardona; FERREIRA JUNIOR, Jose Raniery; MORENO, Ramon Alfredo; REBELO, Marina de Fatima de Sa; KRIEGER, Jose Eduardo; GUTIERREZ, Marco Antonio
    Dual-energy subtraction (DES) is a technique that separates soft tissue from bones in a chest radiograph (CR). As DES requires specialized equipment, we propose an automatic method based on convolutional neural networks (CNNs) to generate virtual soft tissue images. A dataset comprising 35 pairs of CR and its soft-tissue version split in training (28 image pairs) and testing (7 image pairs) sets were used with data augmentation. We tested two types of images: the lung region's cropped image and the segmented lung image. The ribs suppression was treated as a local problem, so each image was divided into 784 patches. The U-Net architecture was used to perform bone suppression. We tested two types of loss functions: mean squared error (L-mse) and L-sm, which combines L-mse with the structural similarity index measure (SSIM). Due to the patches overlapping, it was necessary to interpolate the gray levels on the reconstructed image from the predicted patches. Evaluations were based on SSIM and root mean square error (RMSE) over the reconstructed lung area. The combination that presented the best results used the loss L-sm and the segmented lung image as input to the U-Net (SSIM of 0.858 and RMSE of 0.033). We observed that the U-Net has poor performance when trained with cropped images containing all information from the chest cavity and how the loss using local information can improve CR rib bone suppression. Our results suggest that it is possible removing the rib bones accurately in CR using CNN and a patch-based approach.y