Martin German Gonzalez

Biofuel characterization using electrical measurements is a promising alternative to conventional techniques, particularly given the increasing use of radio frequency and microwave processing in this field. Accurate measurement of electrical properties, such as permittivity and conductivity, is critical for process control and quality assurance. However, commercially available sensors are often expensive and not fully suited for biofuel applications. This paper presents a novel, low-cost sensor for measuring the electrical properties of liquid biofuels, which is robust and highly accurate for industrial use. The sensor operates at frequencies of up to 60 MHz, including the 13.56, 27.12, and 40.68 MHz industrial, scientific, and medical bands, enabling the characterization of a wide range of biofuels, from low-loss dielectrics to highly polar and conductive liquids. The sensor performance is demonstrated by its high accuracy: a relative uncertainty of 2% for permittivity and 1% for conductivity. We demonstrate its use in two key applications: the accurate verification of commercial gasoline/bioethanol and diesel/biodiesel blends, and the real-time monitoring of electrical properties during biodiesel production via transesterification.

In this paper we consider the problem of acoustic inversion in the context of the optoacoustic tomography image reconstruction problem. By leveraging the ability of the recently proposed diffusion models for image generative tasks among others, we devise an image reconstruction architecture based on a conditional diffusion process. The scheme makes use of an initial image reconstruction, which is preprocessed by an autoencoder to generate an adequate representation. This representation is used as conditional information in a generative diffusion process. Although the computational requirements for training and implementing the architecture are not low, several design choices discussed in the work were made to keep them manageable. Numerical results show that the conditional information allows to properly bias the parameters of the diffusion model to improve the quality of the initial reconstructed image, eliminating artifacts or even reconstructing finer details of the ground-truth image that are not recoverable by the initial image reconstruction method. We also tested the proposal under experimental conditions and the obtained results were in line with those corresponding to the numerical simulations. Improvements in image quality up to 17 % in terms of peak signal-to-noise ratio were observed.

En este trabajo se propone el uso de una red generativa de confrontación (GAN) para efectuar un aumento de datos con el objetivo de mejorar la reconstrucción de imágenes en sistemas para tomografía optoacústica (TOA). Se utilizó el modelo denominado FastGAN que es una red compacta, capaz de generar imágenes de alta resolución a partir de un conjunto de datos reducidos. La calidad de los datos generados se evaluó a través de dos métodos. Por un lado, se usó la distancia de inicio de Fréchet (FID), observándose una tendencia decreciente a largo de todo el entrenamiento de la GAN. En el segundo método se entrenó una red neuronal U-Net diseñada para un sistema de TOA con y sin datos aumentados. En este caso, el modelo entrenado con los datos extras aportados por la GAN logró una mejora apreciable en las figuras de mérito asociadas a la reconstrucción.

El objetivo de este trabajo es el estudio de un método de pre-procesamiento de los datos medidos por un tomógrafo optoacústico bidimensional para reducir o eliminar los artefactos introducidos por la escasa cantidad de detectores en el sistema experimental y el acotado ancho de banda de estos. Para esta tarea, se utilizó una red neuronal profunda generativa adversaria y se comparó su rendimiento con una red neuronal de referencia U-Net. En la mayoría de los casos de testeo realizados, se encontró una leve mejora aplicando la red propuesta al medir la correlación de Pearson y la relación señal a ruido piso entre la imagen reconstruida producto de los datos procesados por el modelo y la imagen de alta resolución de referencia.

Image reconstruction in optoacoustic tomography (OAT) is a trending learning task highly dependent on measured physical magnitudes present at sensing time. A large number of different settings and also the presence of uncertainties or partial knowledge of parameters can lead to reconstruction algorithms that are specifically tailored and designed to a particular configuration, which could not be the one that will ultimately be faced in a final practical situation. Being able to learn reconstruction algorithms that are robust to different environments (e.g., the different OAT image reconstruction settings) or invariant to such environments is highly valuable because it allows us to focus on what truly matters for the application at hand and discard what are considered spurious features. In this work, we explore the use of deep learning algorithms based on learning invariant and robust representations for the OAT inverse problem. In particular, we consider the application of the ANDMask scheme due to its easy adaptation to the OAT problem. Numerical experiments are conducted showing that when out-of-distribution generalization (against variations in parameters such as the location of the sensors) is imposed, there is no degradation of the performance and, in some cases, it is even possible to achieve improvements with respect to standard deep learning approaches where invariance robustness is not explicitly considered.

In this paper we consider the problem of image reconstruction in optoacoustic tomography. In particular, we devise a deep neural architecture that can explicitly take into account the band-frequency information contained in the sinogram. This is accomplished by two means. First, we jointly use a linear filtered back-projection method and a fully dense UNet for the generation of the images corresponding to each one of the frequency bands considered in the separation. Secondly, in order to train the model, we introduce a special loss function consisting of three terms: (i) a separating frequency bands term; (ii) a sinogram-based consistency term and (iii) a term that directly measures the quality of image reconstruction and which takes advantage of the presence of ground-truth images present in training dataset. Numerical experiments show that the proposed model, which can be easily trainable by standard optimization methods, presents an excellent generalization performance quantified by a number of metrics commonly used in practice. Also, in the testing phase, our solution has a comparable (in some cases lower) computational complexity, which is a desirable feature for real-time implementation of optoacoustic imaging.

Speeding up data acquisition and reducing the complexity of the detection system are among the central goals for advancing optoacoustic imaging. In this way, new reconstruction algorithms using the compressed sensing (CS) formalism have received considerable interest in recent years. This work presents a comparative study of reconstruction algorithms using time-domain CS schemes for optoacoustic tomography. This is motivated by the well-known capabilities of the CS paradigm in achieving good reconstruction performance even with limited sensing capabilities. Formulation of the mathematical problem is provided along with simulation results, where the performance of different representation basis used in time-domain CS strategies are analyzed in a qualitative and quantitative fashion. Comparison with a well-established optoacoustic tomography reconstruction technique as backprojection is also provided. These experiments show the suitability of time-domain CS techniques for this application and point out that the use of the canonical basis for the representation of the optoacoustic measurements is well suited in comparison with other more sophisticated basis.