Publications

In this article a practical laboratory session for teaching operating systems in real time (RTOS) with 32-bit microcontrollers is described. This session is thought to be imparted in the Computer Engineering degree, through a methodology of project-based learning (PBL) [1]. Specifically, this practice session encompasses Hardware/Software RTOS communication between a microcontroller with Cortex- M4 core vector and a software application written in C # using Visual Studio 2010 via virtual serial ports (VCP). This practice is focused as a project that students must be making from scratch, by achieving several milestones, to obtain a final system. The system to develop is divided into two parts. First, we have a STM32 microcontroller family running a completely free RTOS, CoOS, which receives commands to move motors, and simultaneously sends information about distance sensors and using the serial port. On the other hand we have a PC with a Windows system operating family, where a visual application, which shows the value of the sensors and allows to send commands to the microcontroller motion and position is constructed.

Background: Monkeypox is similar to classical smallpox and, until now, isolated cases have been detected in Africa, but in 2022 it has spread around the world and was declared an international health emergency by the WHO at the end of July.Objective: In this work, we collect monkeypox images from official government websites and public datasets (as well as images of healthy people and people with other skin diseases) and create the MonkeypoxSkin dataset (publicly available), which contains close-up images of the skin where you can see the posthulas formed by the disease. With these images, we design, implement, and evaluate several diagnostic aid systems for monkeypox disease.Methods: The classifiers developed and evaluated are based on Convolutional Neural Networks models and some ensembles composed of a combination of those models, obtaining automatic classification results between healthy, monkeypox and other skin damages, given a close skin tissue image.Results: The results show a system accuracy greater than 93% when using a unique CNN model (VGG-19 and ResNet50), and greater than 98% when using a CNN ensemble formed by ResNet50, EfficientNet-B0 and MobileNet-V2.Conclusions: In contrast to the work found to date, this classifier focusses on disease classification based on a low-resolution image obtained from the skin surface, so it can be integrated into a mobile application for mass screening. To do this, it has been necessary to design a proprietary dataset that is publicly available for use. Furthermore, the results obtained improve the accuracy of the previous works.

Early and precise detection of brain tumors is critical for improving clinical outcomes and patient quality of life. This research focused on developing an image classifier using convolutional neural networks (CNN) to detect brain tumors in magnetic resonance imaging (MRI). Brain tumors are a significant cause of morbidity and mortality worldwide, with approximately 300,000 new cases diagnosed annually. Magnetic resonance imaging (MRI) offers excellent spatial resolution and soft tissue contrast, making it indispensable for identifying brain abnormalities. However, accurate interpretation of MRI scans remains challenging, due to human subjectivity and variability in tumor appearance. This study employed CNNs, which have demonstrated exceptional performance in medical image analysis, to address these challenges. Various CNN architectures were implemented and evaluated to optimize brain tumor detection. The best model achieved an accuracy of 97.5%, sensitivity of 99.2%, and binary accuracy of 98.2%, surpassing previous studies. These results underscore the potential of deep learning techniques in clinical applications, significantly enhancing diagnostic accuracy and reliability.

Neuromorphic engineering is a growing and promising discipline nowadays. Neuro-inspiration and brain understanding applied to solve engineering problems is boosting new architectures, solutions and products today. The biological brain and neural systems process information at relatively low speeds through small components, called neurons, and it is impressive how they connect each other to construct complex architectures to solve in a quasi-instantaneous way visual and audio processing tasks, object detection and tracking, target approximation, grasping..., etc., with very low power. Neuromorphs are beginning to be very promising for a new era in the development of new sensors, processors, robots and software systems that mimic these biological systems. The event-driven Scorbot (ED-Scorbot) is a robotic arm plus a set of FPGA / microcontroller's boards and a library of FPGA logic joined in a completely event-based framework (spike-based) from the sensors to the actuators. It is located in Seville (University of Seville) and can be used remotely. Spike-based commands, through neuro-inspired motor controllers, can be sent to the robot after visual processing object detection and tracking for grasping or manipulation, after complex visual and audio-visual sensory fusion, or after performing a learning task. Thanks to the cascade FPGA architecture through the Address-Event-Representation (AER) bus, supported by specialized boards, resources for algorithms implementation are not limited.

La sociedad avanza, y este avance favorece la aparicion de nuevas necesidades, las cuales son cubiertas por profesionales especializados en campos especificos, como medicos, ingenieros, profesores, etc. Pero cada vez mas, estos nuevos “problemas” requieren soluciones complejas y multidisciplinares, haciendo uso de varios campos de conocimiento. Es por esto que, en los ultimos anos, se hayan creado nuevas titulaciones en el ambito universitario para formar a profesionales que posean los conocimientos necesarios para afrontar estos nuevos retos. Este es el caso del grado en Ingenieria de la Salud, el cual tiene un caracter multidisciplinar, combinando conocimientos biomedicos aplicados a la ingenieria informatica e ingenieria en general. Los alumnos que eligen esta titulacion se pueden catalogar en dos perfiles diferenciados: los procedentes de la rama cientifica-tecnologica y los de ciencias de la salud del bachillerato. Dado que las asignaturas de la titulacion en los primeros anos son de formacion basica y se dividen practicamente en 50% de una rama y 50% de la otra, segun la procedencia del alumno puede que encuentre mayor dificultad en las asignaturas que no son de la suya en concreto. En el curso academico 14/15, propusimos la realizacion de un proyecto en grupo cuyos integrantes estaban combinados entre alumnos de la rama cientifica-tecnologica y de la salud, y cuyo tema requeria conocimientos de ambos campos, de forma que hubiera una colaboracion entre ambas partes. La finalidad de este proyecto era que los alumnos asimilaran los conceptos teoricos que eran nuevos para ellos aplicandolos a una situacion real apoyandose en el resto de los miembros del grupo. El proyecto se desarrollo siguiendo la metodologia de aprendizaje ABP aplicada en las sesiones practicas de la asignatura, marcandose en cada una de estas sesiones una serie de hitos que los alumnos deberan ir completando. Se midio el grado de satisfaccion de los alumnos mediante un cuestionario, obteniendo como resultado un alto grado de satisfaccion.

Falls are one of the most frequent causes of injuries in elderly people.Wearable Fall Detection Systems provided a ubiquitous tool for monitoring and alert when these events happen.Recurrent Neural Networks (RNN) are algorithms that demonstrates a great accuracy in some problems analyzing sequential inputs, such as temporal signal values.However, their computational complexity are an obstacle for the implementation in IoT devices.This work shows a performance analysis of a set of RNN architectures when trained with data obtained using different sampling frequencies.These architectures were trained to detect both fall and fall hazards by using accelerometers and were tested with 10-fold cross validation, using the F1-score metric.The results obtained show that sampling with a frequency of 25Hz does not affect the effectiveness, based on the F1-score, which implies a substantial increase in the performance in terms of computational cost.The architectures with two RNN layers and without a first dense layer had slightly better results than the smallest architectures.In future works, the best architectures obtained will be integrated in an IoT solution to determine the effectiveness empirically.

This work presents a disease detection classifier based on symptoms encoded by their severity. This model is presented as part of the solution to the saturation of the healthcare system, aiding in the initial screening stage. An open-source dataset is used, which undergoes pre-processing and serves as the data source to train and test various machine learning models, including SVM, RFs, KNN, and ANNs. A three-phase optimization process is developed to obtain the best classifier: first, the dataset is pre-processed; secondly, a grid search is performed with several hyperparameter variations to each classifier; and, finally, the best models obtained are subjected to additional filtering processes. The best-results model, selected based on the performance and the execution time, is a KNN with 2 neighbors, which achieves an accuracy and F1 score of over 98%. These results demonstrate the effectiveness and improvement of the evaluated models compared to previous studies, particularly in terms of accuracy. Although the ANN model has a longer execution time compared to KNN, it is retained in this work due to its potential to handle more complex datasets in a real clinical context.

Falls are one of the leading causes of permanent injury and/or disability among the elderly. When these people live alone, it is convenient that a caregiver or family member visits them periodically. However, these visits do not prevent falls when the elderly person is alone. Furthermore, in exceptional circumstances, such as a pandemic, we must avoid unnecessary mobility. This is why remote monitoring systems are currently on the rise, and several commercial solutions can be found. However, current solutions use devices attached to the waist or wrist, causing discomfort in the people who wear them. The users also tend to forget to wear the devices carried in these positions. Therefore, in order to prevent these problems, the main objective of this work is designing and recollecting a new dataset about falls, falling risks and activities of daily living using an ankle-placed device obtaining a good balance between the different activity types. This dataset will be a useful tool for researchers who want to integrate the fall detector in the footwear. Thus, in this work we design the fall-detection device, study the suitable activities to be collected, collect the dataset from 21 users performing the studied activities and evaluate the quality of the collected dataset. As an additional and secondary study, we implement a simple Deep Learning classifier based on this data to prove the system’s feasibility.

Neural networks algorithms are commonly used to recognize patterns from different data sources such as audio or vision. In image recognition, Convolutional Neural Networks are one of the most effective techniques due to the high accuracy they achieve. This kind of algorithms require billions of addition and multiplication operations over all pixels of an image. However, it is possible to reduce the number of operations using other computer vision techniques rather than frame-based ones, e.g. neuromorphic frame-free techniques. There exists many neuromorphic vision sensors that detect pixels that have changed their luminosity. In this study, an event-based convolution engine for FPGA is presented. This engine models an array of leaky integrate and fire neurons. It is able to apply different kernel sizes, from 1x1 to 7x7, which are computed row by row, with a maximum number of 64 different convolution kernels. The design presented is able to process 64 feature maps of 7x7 with a latency of 8.98μs.

<title>Abstract</title> Diabetic retinopathy is a dangerous eye anomaly that may cause vision loss and blindness in people with diabetes (more than 500 million adults in 2021). This study presents a novel approach to diabetic retinopathy grading using explainable deep learning techniques to create a diagnosis aid tool that provides a fully detailed final report with alternative explanations for the result. This is clear novelty over existing solutions. The proposed approach is based on an ensemble of two very efficient deep learning networks: efficientNetV2 and ConvNeXt. This deep learning model is evaluated using a public available dataset and compared with the results obtained from previous works. Also, explainable deep learning techniques are applied to present the final report. The architecture is capable of achieving state-of-the-art performance for both the two-class problem and the five-class international clinical diabetes retinopathy (ICDR) classification problem. This work achieves a 96.7% accuracy and an AUC over 96% for all classes. The ensemble provides explainability records which are significantly improved when compared with those obtained from a single network. The ensemble architecture provides good quality explanations to the ophthalmologist with several configurable superimposed heat maps and two probabilityordered diagnostic suggestions, including a quality factor indicating the estimation of the probability ratio between the alternatives. This makes it a valuable diagnostic assistance tool for ophthalmologists and other healthcare professionals. This study introduces a novel approach to diabetic retinopathy (DR) grading, utilizing explainable deep learning techniquesto create a diagnostic aid tool. The system provides a detailed final report with alternative explanations, which is a clear advancement over existing solutions. Our method is based on an ensemble of two highly efficient deep learning networks: EfficientNetV2 and ConvNeXt. This ensemble approach achieves a 96.7% accuracy and an AUC over 96% for the ICDR classification problem, outperforming single-network models in both accuracy and explainability. The system offers ophthalmologists configurable heat maps and probability-ordered diagnostic suggestions, making it a valuable diagnostic tool.

Several studies have focused on classifying behavioral patterns in wildlife and captive species to monitor their activities and so to understanding the interactions of animals and control their welfare, for biological research or commercial purposes. The use of pattern recognition techniques, statistical methods and Overall Dynamic Body Acceleration (ODBA) are well known for animal behavior recognition tasks. The reconfigurability and scalability of these methods are not trivial, since a new study has to be done when changing any of the configuration parameters. In recent years, the use of Artificial Neural Networks (ANN) has increased for this purpose due to the fact that they can be easily adapted when new animals or patterns are required. In this context, a comparative study between a theoretical research is presented, where statistical and spectral analyses were performed and an embedded implementation of an ANN on a smart collar device was placed on semi-wild animals. This system is part of a project whose main aim is to monitor wildlife in real time using a wireless sensor network infrastructure. Different classifiers were tested and compared for three different horse gaits. Experimental results in a real time scenario achieved an accuracy of up to 90.7%, proving the efficiency of the embedded ANN implementation.

The emotional or affective state has a direct impact not only on personal life, but also in the field of work, sports, rehabilitation processes, among other fields. In the evolving understanding of emotional theory, it has been theorized that an emotion can be classified according to a two-dimensional model composed of an Arousal value and a Valence value, as well as empirically demonstrating the impact of emotions on physiological variables. This work presents the development of a wearable device for capturing physiological signals, the collection of a dataset (after approval by the ethics committee) in which participants' emotional states are induced, and the development of an automatic classifier of the emotional state based on neural networks. According to this last point, a 4-phase optimization process is presented in which the physiological sensors are evaluated independently and with multiple variations of the hyperparameters of the neural networks, keeping those that provide the most information, combinations are made between them and the robustness of the final system obtained is evaluated. The results exceed 92% accuracy in all cases, which, compared with previous work, significantly improves the classifiers developed in recent years. The key contributions of this study are detailed as follows: (a) a wearable device designed to collect physiological signals from the user in a non-invasive way is presented, proving that it works properly in a controlled environment; (b) a data-collection protocol is designed to induce emotional states in test subjects using small video clips, demonstrating that the user evokes the feelings that are induced; and (c) a machine learning-based system is developed and optimized to classify the emotional state based on the two-dimensional model of emotion, demonstrating its efficiency and accuracy.

Colorectal cancer is the second leading cause of cancer-related deaths worldwide. To prevent deaths, regular screenings with histopathological analysis of colorectal tissue should be performed. A diagnostic aid system could reduce the time required by medical professionals, and provide an initial approach to the final diagnosis. In this study, we analyze low computational custom architectures, based on Convolutional Neural Networks, which can serve as high-accuracy binary classifiers for colorectal cancer screening using histopathological images. For this purpose, we carry out an optimization process to obtain the best performance model in terms of effectiveness as a classifier and computational cost by reducing the number of parameters. Subsequently, we compare the results obtained with previous work in the same field. Cross-validation reveals a high robustness of the models as classifiers, yielding superior accuracy outcomes of 99.4 ± 0.58% and 93.2 ± 1.46% for the lighter model. The classifiers achieved an accuracy exceeding 99% on the test subset using low-resolution images and a significantly reduced layer count, with images sized at 11% of those used in previous studies. Consequently, we estimate a projected reduction of up to 50% in computational costs compared to the most lightweight model proposed in the existing literature.