Abstract
Purpose: The development of accurate methods for retinal layer and fluid segmentation in optical coherence tomography images can help the ophthalmologists in the diagnosis and follow-up of retinal diseases. Recent works based on joint segmentation presented good results for the segmentation of most retinal layers, but the fluid segmentation results are still not satisfactory. We report a hierarchical framework that starts by distinguishing the retinal zone from the background, then separates the fluid-filled regions from the rest, and finally, discriminates the several retinal layers.Approach: Three fully convolutional networks were trained sequentially. The weighting scheme used for computing the loss function during training is derived from the outputs of the networks trained previously. To reinforce the relative position between retinal layers, the mutex Dice loss (included for optimizing the last network) was further modified so that errors between more distant layers are more penalized. The method's performance was evaluated using a public dataset.Results: The proposed hierarchical approach outperforms previous works in the segmentation of the inner segment ellipsoid layer and fluid (Dice coefficient = 0.95 and 0.82, respectively). The results achieved for the remaining layers are at a state-of-the-art level.Conclusions: The proposed framework led to significant improvements in fluid segmentation, without compromising the results in the retinal layers. Thus, its output can be used by ophthalmologists as a second opinion or as input for automatic extraction of relevant quantitative biomarkers.

Abstract
In many e-commerce platforms user communities share product information in the form of reviews and ratings to help other consumers to make their choices. This study develops a new theoretical framework generating a bipartite network of products sold by Amazon.com in the category musical instruments, by linking products through the reviews. We analyze product rating and perceived helpfulness of online customer reviews and the relationship between the centrality of reviews, product rating and the helpfulness of reviews using Clustering, regression trees, and random forests algorithms to, respectively, classify and find patterns in 2214 reviews. Results demonstrate: (1) that a high number of reviews do not imply a high product rating; (2) when reviews are helpful for consumer decision-making we observe an increase on the number of reviews; (3) a clear positive relationship between product rating and helpfulness of the reviews; and (4) a weak relationship between the centrality measures (betweenness and eigenvector) giving the importance of the product in the network, and the quality measures (product rating and helpfulness of reviews) regarding musical instruments. These results suggest that products may be central to the network, although with low ratings and with reviews providing little helpfulness to consumers. The findings in this study provide several important contributions for e-commerce businesses' improvement of the review service management to support customers' experiences and online customers' decision-making.

Abstract
Problem definition: Online retailers often receive customer orders comprising several products of differing origins. To fulfill these orders, retailers must ship multiple parcels from different locations and-unless they are grouped somewhere along the supply chain-these may reach the customer's doorstep one by one. Academic/practical relevance: We conjecture here that receiving products sequentially instead of all together affects a consumer's reaction to her purchases, possibly influencing-for good or ill-her decision to return products, as well as her overall service satisfaction. We use two-year granular data from an online fashion marketplace to test this hypothesis and characterize consumer behavioral responses to delivery consolidation and examine how it impacts supply chain stakeholders. Methodology: To achieve causal inference, we exploit the fact that the couriers used by the focal marketplace gather together certain parcels for reasons related more to the timing of their arrival than their actual customers, thereby exogenously consolidating the delivery of some orders. We construct a balanced sample of matched twin multiproduct orders that are alike in all respects except their delivery: consolidated (all parcels delivered jointly) versus otherwise (split). Results: We find that delivery consolidation benefits the marketplace and all its suppliers. By eliminating the stress associated with split deliveries, delivery consolidation pleases consumers as it leads to fewer returns and higher overall satisfaction. Managerial implications: Delivering all products in an order together, even if later, reduces the probability of a return, which improves the financial performance of the marketplace and its suppliers and reduces reverse logistics. Our results suggest that in our context, delivery speed matters less than the convenience of receiving all ordered goods in a single delivery, and we provide directions for adapting logistics strategies accordingly. Our empirical findings also imply that the return decisions of multiple products purchased at once should not be considered to be independent. Finding tractable ways of modeling this feature will be necessary in further driving retail practice through theoretical research that accounts for the behavioral implications of delivery consolidation when optimizing fulfillment decisions.

Abstract
A escola desempenha um papel essencial na promoção de estilos de vida e hábitos alimentares saudáveis. Com o encerramento destes estabelecimentos, devido ao 2.º confinamento imposto pela pandemia COVID-19, tornou-se pertinente avaliar o seu impacto nos estilos de vida das crianças/adolescentes e estudar os respetivos fatores associados. Aplicou-se um questionário online (aos encarregados de educação dos alunos da Educação Pré-Escolar/1.º Ciclo e aos alunos do 2.º/3.º Ciclo) recolhendo informação sociodemográfica, antropométrica, de estado da saúde, estilos de vida e respetivas alterações provocadas pelo 2.º confinamento. Recorreu-se à UniANCOVA para identificar os fatores associados às alterações encontradas. Este estudo realizou-se num Agrupamento de Escolas do Norte de Portugal em 494 alunos (idade média de 11 anos, variando entre os 3 e os 16 anos, 51,0% sexo feminino, 49,2% 3.º Ciclo, 35,2% excesso de peso). Do total, 57,0% cumpre as recomendações das horas de sono, 24,0% do tempo de atividade física e 31,4% de tempo de ecrã. Mais de 70,0% dos alunos incluiu alimentos não recomendáveis aos lanches e ao jantar. Devido ao confinamento, 93,0% relatou pelo menos uma alteração negativa, sendo a mais frequente o aumento do tempo de ecrã diário (63,8%). A alteração positiva mais relatada foi a realização de mais refeições em família (66,5%). Os inquiridos que reportaram uma melhor autoperceção da sua alimentação foram os que tiveram mais alterações positivas dos estilos de vida devido à pandemia. Por outro lado, estar entre o percentil 15 e 85 de Índice de Massa Corporal associou-se a mais alterações negativas.

Abstract
Machine learning has been facing significant challenges over the last years, much of which stem from the new characteristics of machine learning problems, such as learning from streaming data or incorporating human feedback into existing datasets and models. In these dynamic scenarios, data change over time and models must adapt. However, new data do not necessarily mean new patterns. The main goal of this paper is to devise a method to predict a model's performance metrics before it is trained, in order to decide whether it is worth it to train it or not. That is, will the model hold significantly better results than the current one? To address this issue, we propose the use of meta-learning. Specifically, we evaluate two different meta-models, one built for a specific machine learning problem, and another built based on many different problems, meant to be a generic meta-model, applicable to virtually any problem. In this paper, we focus only on the prediction of the root mean square error (RMSE). Results show that it is possible to accurately predict the RMSE of future models, event in streaming scenarios. Moreover, results also show that it is possible to reduce the need for re-training models between 60% and 98%, depending on the problem and on the threshold used.

Abstract
Microsatellite Instability (MSI) is an important biomarker in cancer patients, showing a defective DNA mismatch repair system. Its detection allows the use of immunotherapy to treat cancer, an approach that is revolutionizing cancer treatment. MSI is especially relevant for three types of cancer: Colon Adenocarcinoma (COAD), Stomach Adenocarcinoma (STAD), and Uterus corpus endometrial cancer (UCEC). In this work, learning algorithms were employed to predict MSI using RNA-seq data from The Cancer Genome Atlas (TCGA) database, with a focus on the selection of the most informative genomic features. The Multi-Layer Perceptron (MLP) obtained the best score (AUC = 98.44%), showing that it is possible to exploit information from RNA-seq data to find relevant relationships with the instability levels of microsatellites (MS). The accurate prediction of MSI with transcription data from cancer patients will help with the correct determination of MSI status and adequate prescription of immunotherapy, creating more precise and personalized patient care. At the genetic level, the study revealed a high expression of genes related to cell regulation functions, and a low expression of genes responsible for Mismatch Repair functions, in patients with high instability.