Abstract
Stomach cancer is the third deadliest type of cancer in the world (0.86 million deaths in 2017). In 2035, a 20% increase will be observed both in incidence and mortality due to demographic effects if no interventions are foreseen. Upper GI endoscopy (UGIE) plays a paramount role in early diagnosis and, therefore, improved survival rates. On the other hand, human and technical factors can contribute to misdiagnosis while performing UGIE. In this scenario, artificial intelligence (AI) has recently shown its potential in compensating for the pitfalls of UGIE, by leveraging deep learning architectures able to efficiently recognize endoscopic patterns from UGIE video data. This work presents a review of the current state-of-the-art algorithms in the application of AI to gastroscopy. It focuses specifically on the threefold tasks of assuring exam completeness (i.e., detecting the presence of blind spots) and assisting in the detection and characterization of clinical findings, both gastric precancerous conditions and neoplastic lesion changes. Early and promising results have already been obtained using well-known deep learning architectures for computer vision, but many algorithmic challenges remain in achieving the vision of AI-assisted UGIE. Future challenges in the roadmap for the effective integration of AI tools within the UGIE clinical practice are discussed, namely the adoption of more robust deep learning architectures and methods able to embed domain knowledge into image/video classifiers as well as the availability of large, annotated datasets.

Abstract
Crowdsourced social media data has become popular for assessing cultural ecosystem services (CES). Nevertheless, social media data analyses in the context of CES can be time consuming and costly, particularly when based on the manual classification of images or texts shared by people. The potential of deep learning for automating the analysis of crowdsourced social media content is still being explored in CES research. Here, we use freely available deep learning models, i.e., Convolutional Neural Networks, for automating the classification of natural and human (e.g., species and human structures) elements relevant to CES from Flickr and Wikiloc images. Our approach is developed for Peneda-Ger <^>es (Portugal) and then applied to Sierra Nevada (Spain). For Peneda-Ger <^>es, image classification showed promising results (F1-score ca. 80%), highlighting a preference for aesthetics appreciation by social media users. In Sierra Nevada, even though model performance decreased, it was still satisfactory (F1-score ca. 60%), indicating a predominance of people's pursuit for cultural heritage and spiritual enrichment. Our study shows great potential from deep learning to assist in the automated classification of human-nature interactions and elements from social media content and, by extension, for supporting researchers and stakeholders to decode CES distributions, benefits, and values.

Abstract
This work focuses on detection of upper gas-trointestinal (GI) landmarks, which are important anatomical areas of the upper GI tract digestive system that should be photodocumented during endoscopy to guarantee a complete examination. The aim of this work consisted in testing new automatic algorithms, specifically based on convolutional neural network (CNN) systems, able to detect upper GI landmarks, that can help to avoid the presence of blind spots during esophagogastroduodenoscopy. We tested pre-trained CNN architectures, such as the ResNet-50 and VGG-16, in conjunction with different training approaches, including the use of class weights, batch normalization, dropout, and data augmentation. The ResNet-50 model trained with class weights was the best performing CNN, achieving an accuracy of 71.79% and a Mathews Correlation Coefficient (MCC) of 65.06%. The combination of supervised and unsupervised learning was also explored to increase classification performance. In particular, convolutional autoencoder architectures trained with unlabeled GI images were used to extract representative features. Such features were then concatenated with those extracted by the pre-trained ResNet-50 architecture. This approach achieved a classification accuracy of 72.45% and an MCC of 65.08%. Clinical relevance - Esophagogastroduodenoscopy (EGD) photodocumentation is essential to guarantee that all areas of the upper GI system are examined avoiding blind spots. This work has the objective to help the EGD photodocumentation monitorization by testing new CNN-based systems able to detect EGD landmarks.

Abstract
Objective: Murmurs are abnormal heart sounds, identified by experts through cardiac auscultation. The murmur grade, a quantitative measure of the murmur intensity, is strongly correlated with the patient's clinical condition. This work aims to estimate each patient's murmur grade (i.e., absent, soft, loud) from multiple auscultation location phonocardiograms (PCGs) of a large population of pediatric patients from a low-resource rural area. Methods: The Mel spectrogram representation of each PCG recording is given to an ensemble of 15 convolutional residual neural networks with channel-wise attention mechanisms to classify each PCG recording. The final murmur grade for each patient is derived based on the proposed decision rule and considering all estimated labels for available recordings. The proposed method is cross-validated on a dataset consisting of 3456 PCG recordings from 1007 patients using a stratified ten-fold cross-validation. Additionally, the method was tested on a hidden test set comprised of 1538 PCG recordings from 442 patients. Results: The overall cross-validation performances for patient-level murmur gradings are 86.3% and 81.6% in terms of the unweighted average of sensitivities and F1-scores, respectively. The sensitivities (and F1-scores) for absent, soft, and loud murmurs are 90.7% (93.6%), 75.8% (66.8%), and 92.3% (84.2%), respectively. On the test set, the algorithm achieves an unweighted average of sensitivities of 80.4% and an F1-score of 75.8%. Conclusions: This study provides a potential approach for algorithmic pre-screening in low-resource settings with relatively high expert screening costs. Significance: The proposed method represents a significant step beyond detection of murmurs, providing characterization of intensity, which may provide an enhanced classification of clinical outcomes.

Abstract
The link between epicardial adipose tissue (EAT) and cardiovascular risk is well established, with EAT volume being strongly associated with inflammation, coronary artery disease (CAD) risk, and mortality. However, its EAT quantification is hindered by the time-consuming nature of manual EAT segmentation in cardiac computed tomography (CT). 300 non-contrast cardiac CT scans were collected and the pericardium was manually delineated. In a subset of this data (N = 30), manual delineation was repeated by the same operator and by a second operator. Two automatic methods were then used for pericardial segmentation: a commercially available tool, Siemens Cardiac Risk Assessment (CRA) software; and a deep learning solution based on a U-Net architecture trained exclusively with external public datasets (CardiacFat and OSIC). EAT segmentations were obtained through thresholding to [- 150,- 50] Hounsfield units. Pericardial and EAT segmentation performance was evaluated considering the segmentations by the first operator as reference. Statistical significance of differences for all metrics and segmentation methods was tested through Student t-tests. Pericardial segmentation intra-/interobserver variability was excellent, with the U-Net outperforming Siemens CRA (p < 0.0001). The intra- and interobserver agreement for EAT segmentation was lower with Dice Scores (DSC) of 0.862 and 0.775 respectively, while the U-Net and Siemens CRA obtained DSCs of 0.723 and 0.679 respectively. EAT volume quantification showed that the agreement between a human observer and the U-Net was better than that of two human observers (p = 0.0141), with a Pearson Correlation Coefficient (PCC) of 0.896 and a bias of - 2.83 cm(3) (below the interobserver bias of 9.05 cm3). The lower performances of EAT segmentation highlight the difficulty in segmenting this structure. For both pericardial and EAT segmentation, the deep learning method outperformed the commercial solution. While the segmentation performance of the U-Net solution was below interobserver variability, EAT volume quantification performance was competitive with human readers, motivating future use of these tools. Clinical trial number: NCT03280433, registered retrospectively on 2017-09-08.

Abstract