Abstract

Abstract
Precancerous conditions such as intestinal metaplasia (IM) have a key role in gastric cancer development and can be detected during endoscopy. During upper gastrointestinal endoscopy (UGIE), misdiagnosis can occur due to technical and human factors or by the nature of the lesions, leading to a wrong diagnosis which can result in no surveillance/treatment and impairing the prevention of gastric cancer. Deep learning systems show great potential in detecting precancerous gastric conditions and lesions by using endoscopic images and thus improving and aiding physicians in this task, resulting in higher detection rates and fewer operation errors. This study aims to develop deep learning algorithms capable of detecting IM in UGIE images with a focus on model explainability and interpretability. In this work, white light and narrow-band imaging UGIE images collected in the Portuguese Institute of Oncology of Porto were used to train deep learning models for IM classification. Standard models such as ResNet50, VGG16 and InceptionV3 were compared to more recent algorithms that rely on attention mechanisms, namely the Vision Transformer (ViT), trained in 818 UGIE images (409 normal and 409 IM). All the models were trained using a 5-fold cross-validation technique and for validation, an external dataset will be tested with 100 UGIE images (50 normal and 50 IM). In the end, explainability methods (Grad-CAM and attention rollout) were used for more clear and more interpretable results. The model which performed better was ResNet50 with a sensitivity of 0.75 (±0.05), an accuracy of 0.79 (±0.01), and a specificity of 0.82 (±0.04). This model obtained an AUC of 0.83 (±0.01), where the standard deviation was 0.01, which means that all iterations of the 5-fold cross-validation have a more significant agreement in classifying the samples than the other models. The ViT model showed promising performance, reaching similar results compared to the remaining models.

Abstract
Underwater acoustic localization is a challenging task. Most techniques rely on a network of acoustic sensors and beacons to estimate relative position, therefore localization uncertainty becomes highly dependent on the selected sensor configuration. Although several works in literature exploit optimal sensor placement to improve localization over large regions, the conditions contemplated in these are not applicable for the optimization of the acoustic sensors on constrained 3D shapes, such as the body of small underwater vehicles or structures. Additionally, most commercial systems used for localization with ultra-short baseline (USBL) configurations have compact acoustic sensors that cannot be spatially positioned independently. This work tackles the optimization of acoustic sensor placement in a limited 3D shape, in order to improve the localization accuracy for USBL applications. The implemented multi-objective memetic algorithm combines the Cramer-Rao Lower Bound (CRLB) configuration evaluation with incidence angle considerations for the sensor placement.

Abstract
Councils are a common organisational structure of Portuguese Universities and Polytechnic Institutes. They make the key decisions, in these organisations, by nominal voting at assembly meetings. The COVID pandemic forced the remote work upon most organisations, including universities and polytechnic institutes. Assuming that a remote assembly requires additional efforts in order to guarantee the integrity of the majority decisions taken by votes expressed by its members, opportunity arises for the use of a blockchain-assisted voting system. Benefits of blockchain, such as verifiability, immutability, tamper resistant, and its distributed nature appear to be a good fit. We propose a novel blockchain-assisted system to support the decision making of academic councils that operate by nominal voting in assemblies, gathering remotely and online.

Abstract
<p>The management of natural hazards is a vital concern for the sustainable development of any country and information is the single most important factor to tackle the risks from natural hazards within the risk reduction phase, and to manage response during a crisis. To cope with these challenges it is required, on one hand, to collect baseline information on the natural systems to understand their current state, to identify changes and predict or forecast their future behaviour and, on the other hand, to update information during crisis to review and determine management strategies.</p><p>One major difficulty to this approach is the economic weight of the classic monitoring systems, requiring heavy investments, costly maintenance, and substantial human resources. To overcome these obstacles, an alternative concept was developed based on low-cost and fast deployable wireless sensors networks made by autonomous devices, each capable to communicate to a cloud computing service that compiles and processes data, producing information readily accessible via web.</p><p>The 2021 eruption of the Cumbre Vieja volcano presented an excellent opportunity for a proof of concept of this idea. A trial run was set up on this challenging environment, focusing mainly on the detection and measurement of eruptive products, targeting the measurement of eruptive plume components, such as carbon dioxide (CO<sub>2</sub>), sulphur dioxide (SO<sub>2</sub>) and ash (particle matter, PM), and the monitoring of lava flows entering the sea. Besides the sensor’s setups, also the automatic data processing and different communications were tested.</p><p>The experiment consisted of a proximal network of different stations measuring CO<sub>2</sub>, SO<sub>2</sub>, PM<sub>10</sub>, PM<sub>2.5</sub>, temperature, and humidity; a set of trials to intercept the eruptive plume with weather balloons to measure in-situ the same parameters; a distal aethalometer to detect particles from the distal plume; and a set of buoys to monitor hydroacoustic and environmental parameters in the proximity of the lava deltas. The proximal network allowed for a continuous monitoring with information immediately available via web, with good spatial and temporal correlations between different parameters. The atmospheric soundings allowed to measure particle mass concentrations and sulphur dioxide along a profile of the eruptive plume and characterize its vertical profile, with in situ measurements, while back trajectory of air parcel analyses and aethalometer measurements carried out at Izaña Atmospheric Observatory (2367 m.a.s.l.) showed attenuation variability that could be associated with small volcanic particles transported to at least 140 km from the source. The buoys trial allowed to record the acoustic environment near the lava deltas and to test the design and configurations of the device regarding sensors integration and communications.</p><p>The Cumbre Vieja eruption experiment allowed to try-out a fast deployment low-cost multi-sensor system with good results on volcanic plume characterization and real-time data production that proved to be useful for managing volcanic crisis and demonstrated the relevance of this alternative monitoring concept.</p>

Abstract
Ideally, to carry out screening for eye diseases, it is expected to use specialized medical equipment to capture retinal fundus images. However, since this kind of equipment is generally expensive and has low portability, and with the development of technology and the emergence of smartphones, new portable and cheaper screening options have emerged, one of them being the D-Eye device. When compared to specialized equipment, this equipment and other similar devices associated with a smartphone present lower quality and less field-of-view in the retinal video captured, yet with sufficient quality to perform a medical pre-screening. Individuals can be referred for specialized screening to obtain a medical diagnosis if necessary. Two methods were proposed to extract the relevant regions from these lower-quality videos (the retinal zone). The first one is based on classical image processing approaches such as thresholds and Hough Circle transform. The other performs the extraction of the retinal location by applying a neural network, which is one of the methods reported in the literature with good performance for object detection, the YOLO v4, which was demonstrated to be the preferred method to apply. A mosaicing technique was implemented from the relevant retina regions to obtain a more informative single image with a higher field of view. It was divided into two stages: the GLAMpoints neural network was applied to extract relevant points in the first stage. Some homography transformations are carried out to have in the same referential the overlap of common regions of the images. In the second stage, a smoothing process was performed in the transition between images.