Abstract
The segmentation of retinal vessels in fundus images has been heavily focused in the past years, given their relevance in the diagnosis of several health conditions. Even though the recent advent of deep learning allowed to foster the performance of computer-based algorithms in this task, further improvement concerning the detection of vessels while suppressing background noise has clinical significance. Moreover, the best performing state-of-the-art methodologies conduct patch-based predictions. This, put together with the preprocessing techniques used in those methodologies, may hinder their use in screening scenarios. Thus, in this paper, we explore a fully convolutional setting that takes raw fundus images and allows to combine patch-based training with global image prediction. Our experiments on the DRIVE, STARE and CHASEDB1 databases show that the proposed methodology achieves state-of-the-art performance in the first and the last, allowing at the same time much faster segmentation of new images.

Abstract
When learning from images, it is desirable to augment the dataset with plausible transformations of its images. Unfortunately, it is not always intuitive for the user how much shear or translation to apply. For this reason, training multiple models through hyperparameter search is required to find the best augmentation policies. But these methods are computationally expensive. Furthermore, since they generate static policies, they do not take advantage of smoothly introducing more aggressive augmentation transformations. In this work, we propose repeating each epoch twice with a small difference in data augmentation intensity, walking towards the best policy. This process doubles the number of epochs, but avoids having to train multiple models. The method is compared against random and Bayesian search for classification and segmentation tasks. The proposal improved twice over random search and was on par with Bayesian search for 4% of the training epochs.

Abstract
The enhancement of tubular structures such as vessels in medical images has been addressed in the past, aiming for easier extraction and or visualization of such structures by professionals. Some literature methodologies propose vesselness measures whose design is motivated by local properties of vascular networks and how these influence the eigenvalues of the Hessian matrix. However, past work fails to combine properly the scale-space and neighborhood information, thus leading to the proposal of suboptimal vesselness measures. In this paper, we show that a shallow convolutional neural network is able to learn more optimal embedding spaces from the eigenvalue analysis at different scales, thus leading to a stronger vessel enhancement. Additionally, we also show that such a system maintains one of the biggest advantages of Hessian-based vesselness measures, which is the robustness to data with varying statistics. © 2019, Springer Nature Switzerland AG.

Abstract
The performance of biometric systems is known to decay over time, eventually rendering them ineffective. Focused on ECG-based biometrics, this work aims to study the permanence of these signals for biometric identification in state-of-the-art methods, and measure the effect of template update on their long-term performance. Ensuring realistic testing settings, four literature methods based on autocorrelation, autoencoders, and discrete wavelet and cosine transforms, were evaluated with and without template update, using Holter signals from THEW’s E-HOL 24 h database. The results reveal ECG signals are unreliable for long-term biometric applications, and template update techniques offer considerable improvements over the state-of-the-art results. Nevertheless, further efforts are required to ensure long-term effectiveness in real applications. © 2019, Springer Nature Switzerland AG.

Abstract
Transfer learning focuses on building better predictive models by exploiting knowledge gained in previous related tasks, being able to soften the traditional supervised learning assumption of having identical train-test distributions. Most efforts on transfer learning consider revisiting the data from the source tasks or rely on transferring knowledge for specific models. In this paper, a general framework is proposed for transferring knowledge by including a regularization factor based on the structural model similarity between related tasks. The proposed approach is instantiated to different models for regression, classification, ranking and recommender systems, obtaining competitive results in all of them. Also, we explore high-level concepts in transfer learning like sparse transfer, partially observable transfer and cross-model transfer.

Abstract
Breast cancer high survival rate led to an increased interest in the quality of life after treatment, particularly regarding the aesthetic outcome. Currently used aesthetic assessment methods are subjective, which make reproducibility and impartiality impossible. To create an objective method capable of being selected as the gold standard, it is fundamental to detect, in a completely automatic manner, keypoints in photographs of women's torso after being subjected to breast cancer surgeries. This paper proposes a deep and a hybrid model to detect keypoints with high accuracy. Our methods are tested on two datasets, one composed of images with a clean and consistent background and a second one that contains photographs taken under poor lighting and background conditions. The proposed methods represent an improvement in the detection of endpoints, nipples and breast contour for both datasets in terms of average error distance when compared with the current state-of-the-art.