The unstructured nature of psychiatric clinical notes poses a significant challenge for automated information extraction and data structuring. In this study, we explore the use of transformer-based language models to perform Named Entity Recognition (NER) on de-identified Spanish electronic health records (EHRs) provided by the Psychiatry Service of Complejo Asistencial Universitario de León (CAULE). A manually annotated gold standard, consisting of 200 clinical notes, was developed by domain experts to evaluate the performance of five models: BETO (cased and uncased), ALBETO, ClinicalBERT, and Bio_ClinicalBERT. Each model was fine-tuned and assessed using a strict exact matching criterion across six clinically relevant label types. Results demonstrate that ClinicalBERT, despite being pre-trained on English medical corpora, achieved the highest macro-average F1-score on the test set (80 %). However, BETO-cased outperformed ClinicalBERT in four out of six label types, being better in categories with higher syntactic variability. Lower-performing models, such as ALBETO and Bio_ClinicalBERT, struggled to generalize to Spanish psychiatric language, likely due to domain and language mismatches. This work highlights the effectiveness of transformer-based architectures for structuring psychiatric narratives in Spanish and provides a robust foundation for future clinical NLP applications in non-English contexts.

Pancreatic cancer remains one of the most aggressive malignancies, with limited survival rates and significant variability in patient outcomes. This study evaluates the performance of three machine learning models (Random Forest, Decision Tree, and XGBoost) in predicting patient survival at 3, 12, and 18 months, using data from the Complejo Asistencial Universitario de León (CAULE) Radiology Department. To systematically analyze the impact of different features on survival prediction, the dataset was structured into seven variable groups (G1G7), incorporating demographic, clinical, and treatment-related information. To address the inherent class imbalance in survival prediction, an Autoencoder-based synthetic data generation approach was applied, ensuring a balanced distribution of survival and non-survival cases across all timeframes. Hyperparameter tuning was performed, and experimental results indicate that Random Forest and XGBoost achieved comparable performance, both obtaining an accuracy above 81 % at 3 months, 83 % at 12 months, and 88 % at 18 months when trained on Group G7. To enhance model interpretability, SHapley Additive exPlanations (SHAP) was applied to the best-performing model, identifying key factors influencing survival.

Abstract Parkinson’s disease is easy to diagnose when it is advanced, but it is very difficult to diagnose in its early stages. Early diagnosis is essential to be able to treat the symptoms. It impacts on daily activities and reduces the quality of life of both the patients and their families and it is also the second most prevalent neurodegenerative disorder after Alzheimer in people over the age of 60. Most current studies on the prediction of Parkinson’s severity are carried out in advanced stages of the disease. In this work, the study analyzes a set of variables that can be easily extracted from voice analysis, making it a very non-intrusive technique. In this paper, a method based on different deep learning techniques is proposed with two purposes. On the one hand, to find out if a person has severe or non-severe Parkinson’s disease, and on the other hand, to determine by means of regression techniques the degree of evolution of the disease in a given patient. The UPDRS (Unified Parkinson’s Disease Rating Scale) has been used by taking into account both the motor and total labels, and the best results have been obtained using a mixed multi-layer perceptron (MLP) that classifies and regresses at the same time and the most important features of the data obtained are taken as input, using an autoencoder. A success rate of 99.15% has been achieved in the problem of predicting whether a person suffers from severe Parkinson’s disease or non-severe Parkinson’s disease. In the degree of disease involvement prediction problem case, a MSE (Mean Squared Error) of 0.15 has been obtained. Using a full deep learning pipeline for data preprocessing and classification has proven to be very promising in the field Parkinson’s outperforming the state-of-the-art proposals.

Detecting individuals with autism spectrum disorder (ASD) remains a challenge due to the resources and specialized professionals needed for accurate diagnosis, particularly for children where time is a critical factor. Early diagnosis of ASD is crucial for improving the quality of life for affected individuals, as it allows for timely intervention and support. In this study, we underscore the importance of artificial intelligence (AI) in developing innovative diagnostic methods, with the primary objective of creating AI models that assist in identifying users who may have ASD. Although several studies have utilized traditional machine learning (ML) and deep learning (DL) techniques to detect various illnesses, few have focused on detecting ASD using text as input. We employ natural language processing (NLP) techniques combined with AI models, specifically decision trees, extreme gradient boosting (XGB), k-nearest neighbors algorithm (KNN) as ML models, and bidirectional encoder representations from transformers (BERT) as DL models. The core idea involves extracting tweets from Twitter users through the platform's API, classifying the texts as written by individuals who claim to have ASD (ASD users) or by those without ASD (non-ASD users). We generated a dataset of 404,627 tweets and used a subset of 90,000 tweets, comprising 45,000 from each classification group, for training and testing the models. The results demonstrate a predictive model with an accuracy of over 84% when classifying texts potentially originated from ASD users. This research paves the way for using DL models to enhance the accuracy of detecting and diagnosing ASD in individuals effectively, emphasizing the critical role of AI in advancing early diagnostic methods for better patient outcomes.

Abstract A new deep learning system is proposed for the rapid and accurate identification of anthropogenic elements of the Roman mining infrastructure in NW Iberia, providing a new approach for automatic recognition of different mining elements without the need for human intervention or implicit subjectivity. The recognition of archaeological and other abandoned mining elements provides an optimal test case for decision-making and management in a broad variety of research fields. A new image dataset was created by obtaining UAV images from different anthropic features. A convolutional neural network architecture was implemented, achieving recognition results of close to 95% accuracy. This methodological approach is suitable for the identification and accurate location of ancient mines and hydrologic infrastructure, providing new tools for accurate mapping of mining landforms. Additionally, this novel application of deep learning can be implemented to reduce potential risks caused by abandoned mines, which can cause significant annual human and economic losses worldwide.

This work addresses the performance comparison between four clustering techniques with the objective of achieving strong hybrid models in supervised learning tasks. A real dataset from a bio-climatic house named Sotavento placed on experimental wind farm and located in Xermade (Lugo) in Galicia (Spain) has been collected. Authors have chosen the thermal solar generation system in order to study how works applying several cluster methods followed by a regression technique to predict the output temperature of the system. With the objective of defining the quality of each clustering method two possible solutions have been implemented. The first one is based on three unsupervised learning metrics (Silhouette, Calinski-Harabasz and Davies-Bouldin) while the second one, employs the most common error measurements for a regression algorithm such as Multi Layer Perceptron.

Abstract Cardiovascular diseases state as one of the greatest risks of death for the general population. Late detection in heart diseases highly conditions the chances of survival for patients. Age, sex, cholesterol level, sugar level, heart rate, among other factors, are known to have an influence on life-threatening heart problems, but, due to the high amount of variables, it is often difficult for an expert to evaluate each patient taking this information into account. In this manuscript, the authors propose using deep learning methods, combined with feature augmentation techniques for evaluating whether patients are at risk of suffering cardiovascular disease. The results of the proposed methods outperform other state of the art methods by 4.4%, leading to a precision of a 90%, which presents a significant improvement, even more so when it comes to an affliction that affects a large population.

Eating disorders (ED) are among the most widespread mental illnesses in our society today. This research work presents the study of deep learning models applied to the domain of eating disorders. For this purpose, a collection of messages from the social network Twitter was compiled using web scraping techniques. After collecting a total amount of 1,085,957 tweets, a subset of 2,000 tweets was manually classified. This classification made it possible to differentiate tweets written by people who suffer or have suffered from an ED from those written by people who have not suffered from an ED. After this, 6 predictive models based on Bidirectional Encoder Representations from Transformers (BERT) were created and a comparison was made by evaluating which model scored the best. The best scoring model was RoBERTa using the pre-trained roberta-base model with an accuracy of 87.5%.

In this work, a descriptor called a ZIBO (absolute Zernike moments with Invariant Boundary Orientation) that describes the shape of objects using the module of Zernike moments and the edge features obtained from an almost rotational invariant version of the Edge Gradient Co-occurrence Matrix (EGCM) is proposed. The two descriptors obtained, the Zernike module as global descriptor and the new version of EGCM as local one, are used to characterize images from three different datasets, Kimia99, MPEG2 and MPEG7. Later on, the concatenation of both local and global descriptors was evaluated using kNN with City block and Chi-square distance metrics. Also, the descriptors are assessed separately with a weight-based method, being the results obtained compared with the ones reached by the baseline method, ZMEG (Zernike Moment Edge Gradient). Using MPEG7, which is the most challenging dataset, and the weight-based classifier, this proposal obtained a success rate of 78.29%, outperforming the 75.86% achieved by ZMEG method. With the MPEG2 dataset, results were even better with an 81.00% of success rate against 77.25% of ZMEG.