Abstract
The grape variety plays an important role in the wine production chain, thus identifying it is crucial for production control. Ampelographers, professionals who identify grape varieties through plant visual analysis, are scarce, and molecular markers are expansive to identify grape varieties on a large scale. In this context, Deep Learning models become an effective way to handle ampelographers scarcity. In this work, we explore the benefit of using deep learning vision transformers architecture relative to conventional CNN to identify 12 grapevine varieties using leaf-centred RGB images acquired in the field. We train an Xception model as a baseline and four different configurations of the ViT_B model. The best model achieved 0.96 of F1-score, outperforming the state-of-the-art convolutional-based model in the used dataset.

Abstract
Multi-temporal InSAR (MT- InSAR) observations, which enable deformation monitoring at an unprecedented scale, are usually affected by decorrelation and other noise inducing factors. Such observations (PS - Persistent scatterers), are usually in the order of several thousand, making their respective evaluation frequently computationally expensive. In the present study, we propose an approach for the detection of MT-InSAR outlying observations through the implementation of Convolutional Neural Networks (CNN) classification models. For each PS, the corresponding MT-InSAR parameters and the respective parameters of the neighboring scatterers and its relative position are considered. Tests in two independent datasets, covering the regions of Bratislava city and the suburbs of Prievidza, Slovakia, were performed. The results showed that such models offer a robust and reduced computation time method for the evaluation of MT-InSAR outlying observations. However, the applicability of these models is limited by the deformation pattern in which such models were trained.

Abstract
Insect pests are the main cause of loss of productivity and quality in crops worldwide. Insect monitoring becomes necessary for the early detection of pests and thus avoiding the excessive use of pesticides. Automatic detection of insects attracted by traps is a form of monitoring. Modern data-driven methods present great results for object detection when representative datasets are available, but public datasets for insect detection are few and small. Pest24 public dataset is extensive, but noisy resulting in a poor detection rate. In this work, we aim to improve insect detection in the Pest24 dataset. We propose the creation of three sub-datasets selecting the highest represented classes, the highest colour discrepancy, and the one with the highest relative scale, respectively. Several Faster R-CNN and YOLOv5 architectures are explored, and the best results are achieved with the YOLOv5 with an mAP of 95.5%.

Abstract
Cancers related to the gastrointestinal tract have a high incidence rate in the population, with a high mortality rate. Videos obtained through endoscopic capsules are essential for evaluating anomalies that can progress to cancer. However, due to their duration, which can reach 10 hours, they demand great attention from the medical specialist in their analysis. Machine learning techniques have been successfully applied in developing computer-aided diagnostic systems since the 1990s, where Convolutional Neural Networks (CNNs) have become very successful for pattern recognition in images. CNNs use convolutions to extract features from the analyzed data, operating in a fixed-size window and thus having problems capturing pixel-level relationships considering the spatial and temporal domains. Otherwise, transformers use attention mechanisms, where data is structured in a vector space that can aggregate information from adjacent data to determine meaning in a given context. This work proposes a computational method for analyzing images extracted from videos obtained by endoscopic capsules, using a transformer-based model that helps diagnose of gastrointestinal tract abnormalities. Preliminary results are promising. The classification task of 11 classes evaluated on the publicly available Kvasir-Capsule dataset yielded an average value of 99.70% of accuracy, 99.64% of precision, 99.86% of sensitivity, and 99.54% of f1-score.

Abstract
In the healthcare domain, datasets are often private and lack large amounts of samples, making it difficult to cope with the inherent patient data heterogeneity. As an attempt to mitigate data scarcity, generative models are being used due to their ability to produce new data, using a dataset as a reference. However, synthesis studies often rely on a 2D representation of data, a seriously limited form of information when it comes to lung computed tomography scans where, for example, pathologies like nodules can manifest anywhere in the organ. Here, we develop a 3D Progressive Growing Generative Adversarial Network capable of generating thoracic CT volumes at a resolution of 1283, and analyze the model outputs through a quantitative metric (3D Muli-Scale Structural Similarity) and a Visual Turing Test. Clinical relevance - This paper is a novel application of the 3D PGGAN model to synthesize CT lung scans. This preliminary study focuses on synthesizing the entire volume of the lung rather than just the lung nodules. The synthesized data represent an attempt to mitigate data scarcity which is one of the major limitations to create learning models with good generalization in healthcare.

Abstract
Lung cancer is the leading cause of cancer death worldwide. Early low-dose computed tomography (CT) screening can decrease its mortality rate and computer-aided diagnoses systems may make these screenings more accessible. Radiomic features and supervised machine learning have traditionally been employed in these systems. Contrary to supervised methods, unsupervised learning techniques do not require large amounts of annotated data which are labor-intensive to gather and long training times. Therefore, recent approaches have used unsupervised methods, such as clustering, to improve the performance of supervised models. However, an analysis of purely unsupervised methods for malignancy prediction of lung nodules from CT images has not been performed. This work studies nodule malignancy in the LIDC-IDRI image collection of chest CT scans using established radiomic features and unsupervised learning methods based on k-Means, Spectral Clustering, and Gaussian Mixture clustering. All tested methods resulted in clusters of high homogeneity malignancy. Results suggest convex feature distributions and well-separated feature subspaces associated with different diagnoses. Furthermore, diagnosis uncertainty may be explained by common characteristics captured by radiomic features. The k-Means and Gaussian Mixture models are able to generalize to unseen data, achieving a balanced accuracy of 87.23% and 86.96% when inference was tested. These results motivate the usage of unsupervised approaches for malignancy prediction of lung nodules, such as cluster-then-label models. Clinical Relevance - Unsupervised clustering of radiomic features of lung nodules in chest CT scans can differentiate between malignant and benign cases and reflects experts' diagnosis uncertainty