A complete list can be found here:


                       

Last 10 publications

2022

  • [65]
    Title:Infrared thermographic inspection of 3D hybrid aluminium-CFRP composite using different spectral bands and new unsupervised probabilistic low-rank component factorization model In: NDT & E International

    Abstract: In this work, infrared thermography is used to detect defects on a 3D hybrid aluminium-CFRP composite structure. First, radiometric calibration and geometric distortion correction are performed for 3D inspection. Second, we propose a new unsupervised probabilistic low-rank component factorization thermographic de-noising model to improve image performance and defect visualization. Signal profiles and standard deviation analysis is used to assess the results, and x-ray CT inspections are compared to the infrared inspection results. Finally, we can conclude that the proposed algorithm can detect voids and resin rich areas presenting a better image performance if compared to direct infrared inspection results. 10.1016/j.ndteint.2021.102561

  • [64]
    Title:CNN architecture optimization using bio-inspired algorithms for breast cancer detection in infrared images In: Computers in Biology and Medicine

    Abstract: The early detection of breast cancer is a vital factor when it comes to improving cure and recovery rates in patients. Among such early detection factors, one finds thermography, an imaging technique that demonstrates good potential as an early detection method. Convolutional neural networks (CNNs) are widely used in image classification tasks, but finding good hyperparameters and architectures for these is not a simple task. In this study, we use two bio-inspired optimization techniques, genetic algorithm and particle swarm optimization to find good hyperparameters and architectures for the fully connected layers of three state of the art CNNs: VGG-16, ResNet-50 and DenseNet-201. Through use of optimization techniques, we obtained F1-score results above 0.90 for all three networks, an improvement from 0.66 of the F1-score to 0.92 of the F1-score for the VGG-16. Moreover, we were also able to improve the ResNet-50 from 0.83 of the F1-score to 0.90 of the F1-score for the test data, when compared to previously published studies. 10.1016/j.compbiomed.2021.105205

2021

  • [63]
    Title:Numerical and Non-Destructive Analysis of an Aluminum-CFRP Hybrid 3D Structure In: Metals

    Abstract: Advanced materials are widely used in many industries. They play an important role especially in the aeronautic and automotive sectors where weight reduction is required in order to reduce fuel consumption. Composite materials have a high strength to weight ratio and are applied in airplane construction. Nevertheless, sometimes it is not viable to replace all metal parts by composite ones due to the cost factor. In this sense, hybrid structures are highly welcome. In order to ensure the safety of these hybrid components during their entire life cycle, non-destructive testing evaluation (NDT&E) methods are used and sometimes they are the only option. In this study, we use infrared thermography (IRT) to inspect an aluminum-composite hybrid structure with a 3D shape. The sample has a composite part with a small metal inlay (EN AW-6082) overmolded with a thermoplastic layer. The inlay is bended to reach the desired 3D geometry. This sample was design to be used for the connection between an A- or B-pillar and a car roof made of carbon fiber reinforced polymer (CFRP). A dual-band infrared camera is used in order to capture images in two different spectral ranges. In addition, two data processing techniques for infrared images are applied to enhance the images: principal component thermography (PCT) and partial least squares thermography (PLST). Then, a signal-to-noise ratio analysis is performed with three randomly chosen previous known defects to assess the quality of the images and detected defects. Results showed that principal component thermography has a slight advantage over partial least squares thermography in our specific experiments. Specifically, for the long-wave infrared band, PCT presented, among the defects analyzed, PCT presented a mean value 12.5% higher while the standard deviation was almost three times lower than PLST. In parallel to the non-detructive analysis, a numerical finite element model was formulated in ANSYS® to analyze the total deformations to which the metal-composite-hybrid structure is subjected during a possible use. Results obtained with the numerical model indicate that the interface region between composite and metal parts is where the highest degree of deformation occur, which indicates possible regions where defects and failures may occur in real use cases. 10.3390/met11121938

  • [62]
    Title:Concentrated Thermomics for Early Diagnosis of Breast Cancer In: 16th International Workshop on Advanced Infrared Technology & Applications 2021

    Abstract: Thermography has been employed broadly as a corresponding diagnostic instrument in breast cancer diagnosis. Among thermographic techniques, deep neural networks show an unequivocal potential to detect heterogeneous thermal patterns related to vasodilation in breast cancer cases. Such methods are used to extract high-dimensional thermal features, known as deep thermomics. In this study, we applied convex non-negative matrix factorization (convex NMF) to extract three predominant bases of thermal sequences. Then, the data were fed into a sparse autoencoder model, known as SPAER, to extract low-dimensional deep thermomics, which were then used to assist the clinical breast exam (CBE) in breast cancer screening. The application of convex NMF-SPAER, combining clinical and demographic covariates, yielded a result of 79.3% (73.5%, 86.9%); the highest result belonged to NMF-SPAER at 84.9% (79.3%, 88.7%). The proposed approach preserved thermal heterogeneity and led to early detection of breast cancer. It can be used as a noninvasive tool aiding CBE. 10.3390/engproc2021008030

  • [61]
    Title:Characterization of Ancient Marquetry Using Different Non-Destructive Testing Techniques In: Applied Sciences

    Abstract: Early diagnosis of breast cancer unequivocally improves the survival rate of patients and is crucial for disease treatment. With the current developments in infrared imaging, breast screening using dynamic thermography seems to be a great complementary method for clinical breast examination (CBE) prior to mammography. In this study, we propose a sparse deep convolutional autoencoder model named SPAER to extract low-dimensional deep thermomics to aid breast cancer diagnosis. The model receives multichannel, low-rank, approximated thermal bases as input images. SPAER provides a solution for high-dimensional deep learning features and selects the predominant basis matrix using matrix factorization techniques. The model has been evaluated using five state-of-the-art matrix factorization methods and 208 thermal breast cancer screening cases. The best accuracy was for non-negative matrix factorization (NMF)-SPAER + Clinical and NMF-SPAER for maintaining thermal heterogeneity, leading to finding symptomatic cases with accuracies of 78.2% (74.3–82.5%) and 77.7% (70.9–82.1%), respectively. SPAER showed significant robustness when tested for additive Gaussian noise cases (3–20% noise), evaluated by the signal-to-noise ratio (SNR). The results suggest high performance of SPAER for preserveing thermal heterogeneity, and it can be used as a noninvasive in vivo tool aiding CBE in the early detection of breast cancer. 10.3390/app11177979

  • [60]
    Title:Stacked denoising autoenocder for infrared thermography image enhancement In: 21st IEEE INDIN International Conference on Industrial Informatics, Palma de Mallorca, Spain (Virtual)

    Abstract: Pulsed thermography is one of the most popular thermography inspection methods. During an experiment of pulsed thermography, a specimen is quickly heated, and infrared images are captured to provide information about the specimen's surface and subsurface conditions. Adequate transformations are usually performed to enhance the contrast of the thermal images and to highlight the abnormal regions before these thermal images are visually inspected. Given that deep neural networks have been a success in computer vision in the past few years, a data contrast enhancement approach with stacked denoising autoencoder(DAE) is proposed in this paper to enhance the abnormal regions in the thermal frames gathered by pulsed thermography. Compared to the direct principal component thermography, the proposed method can enhance the abnormalities evidently without weakening important details. 10.1109/INDIN45523.2021.9557407

  • [59]
    Title:Classification of static infrared images using pre-trained CNN for breast cancer detection In: Proceedings of 34th IEEE CBMS International Symposium on Computer-Based Medical Systems, Aveiro, Portugal (Virtual)

    Abstract: Breast cancer is a disease that affects many women throughout the world. It is the second most common type of cancer. The early diagnosis of the disease is relevant for increasing the chances of the patient recovering. Thermography is a promising technique that might be used to help the early diagnosis of breast cancer. In this work, we use three state of the art CNNs (VGG-16, Densenet201, and Resnet50) combined with transfer learning to classify static thermography images (sick and healthy). In our experiments, the best results have an F1-score of 0.92, 91.67% for accuracy, 100% for sensitivity, and 83.3% for specificity obtained with the Densenet using 38 static images for each class. 10.1109/CBMS52027.2021.00094

  • [58]
    Title:SPAER: Sparse Deep Convolutional Autoencoder Model to Extract Low Dimensional Imaging Biomarkers for Early Detection of Breast Cancer Using Dynamic Thermography In: Applied Sciences

    Abstract: Early diagnosis of breast cancer unequivocally improves the survival rate of patients and is crucial for disease treatment. With the current developments in infrared imaging, breast screening using dynamic thermography seems to be a great complementary method for clinical breast examination (CBE) prior to mammography. In this study, we propose a sparse deep convolutional autoencoder model named SPAER to extract low-dimensional deep thermomics to aid breast cancer diagnosis. The model receives multichannel, low-rank, approximated thermal bases as input images. SPAER provides a solution for high-dimensional deep learning features and selects the predominant basis matrix using matrix factorization techniques. The model has been evaluated using five state-of-the-art matrix factorization methods and 208 thermal breast cancer screening cases. The best accuracy was for non-negative matrix factorization (NMF)-SPAER + Clinical and NMF-SPAER for maintaining thermal heterogeneity, leading to finding symptomatic cases with accuracies of 78.2% (74.3–82.5%) and 77.7% (70.9–82.1%), respectively. SPAER showed significant robustness when tested for additive Gaussian noise cases (3–20% noise), evaluated by the signal-to-noise ratio (SNR). The results suggest high performance of SPAER for preserveing thermal heterogeneity, and it can be used as a noninvasive in vivo tool aiding CBE in the early detection of breast cancer. 10.3390/app11073248

  • [57]
    Title: Estimating Thermal Material Properties Using Solar Loading Lock-In Thermography In: Applied Sciences

    Abstract: This work investigates the application of lock-in thermography approach for solar loading thermography applications. In conventional lock-in thermography, a specimen is subjected to a periodically changing heat flux. This heat flux usually enters the specimen in one of three ways: by a point source, a line source or an extended source (area source). Calculations based on area sources are particularly well suited to adapt to solar loading thermography, because most natural heat sources and heat sinks can be approximated to be homogenously extended over a certain region of interest. This is of particular interest because natural heat phenomena cover a large area, which makes this method suitable for measuring large-scale samples. This work investigates how the extended source approximation formulas for determining thermally thick and thermally thin material properties can be used in a naturally excited setup, shows possible error sources, and gives quantitative results for estimating thermal effusivity of a retaining wall structure. It shows that this method can be used on large-scale structures that are subject to natural outside heating phenomena. 10.3390/app11073097

  • [55]
    Title: A Deep Learning Method for the Impact Damage Segmentation of Curve-Shaped CFRP Specimens Inspected by Infrared Thermography In: Sensors

    Abstract: Advanced materials such as continuous carbon fiber-reinforced thermoplastic (CFRP) laminates are commonly used in many industries, mainly because of their strength, stiffness to weight ratio, toughness, weldability, and repairability. Structural components working in harsh environments such as satellites are permanently exposed to some sort of damage during their lifetimes. To detect and characterize these damages, non-destructive testing and evaluation techniques are essential tools, especially for composite materials. In this study, artificial intelligence was applied in combination with infrared thermography to detected and segment impact damage on curved laminates that were previously submitted to a severe thermal stress cycles and subsequent ballistic impacts. Segmentation was performed on both mid-wave and long-wave infrared sequences obtained simultaneously during pulsed thermography experiments by means of a deep neural network. A deep neural network was trained for each wavelength. Both networks generated satisfactory results. The model trained with mid-wave images achieved an F1-score of 92.74% and the model trained with long-wave images achieved an F1-score of 87.39%. 10.3390/s21020395