Alberto Fernández-Villar’s research while affiliated with Centro de Investigación Biomédica en Red de Enfermedades Respiratorias and other places

Background and Objectives: The association between syncope and sleep apnea (SA) has been scarcely investigated. Dysfunction of the autonomic nervous system (ANS) may represent a shared pathophysiological mechanism. This study aimed to determine the prevalence of SA in patients with syncope of unclear cause (SUC), identify potential associated factors, and evaluate nocturnal heart rate variability (HRV) as a marker of ANS function. Materials and Methods: A prospective cohort study was conducted in adult patients diagnosed with SUC. Nocturnal cardiorespiratory polygraphy was performed to detect the presence of SA. A range of variables potentially associated with SA was collected. Both SA diagnosis and HRV parameters were assessed using the Embletta® MPR polygraph system. Results: A total of 156 patients were enrolled (57% male), with a mean age of 64 years and a mean body mass index of 27.5 kg/m² (range: 24.8–32.2). Hypertension was present in 46% of the cohort. The overall prevalence of SA was 78.2% (95% CI: 71.7–84.4%), with 28.7% classified as severe. Age (OR = 1.04; 95% CI: 1.01–1.07) and BMI (OR = 1.17; 95% CI: 1.06–1.28) were independent predictors of SA. Mean RR interval was significantly lower in patients with SA compared to those without (942 ms vs. 995 ms; p = 0.04). No significant differences in HRV parameters were observed between the two groups. Conclusions: This study found a high prevalence (nearly 78%) of SA among adult patients with SUC, particularly in individuals over 50 years of age and those who were overweight. However, this association could not be predicted based on clinical variables alone. No significant differences in nocturnal HRV were detected between patients with SUC with and without SA.

Background: COPD is a chronic disease characterized by frequent exacerbations that require hospitalization, significantly increasing the care burden. In recent years, the use of artificial intelligence-based tools to improve the management of patients with COPD has progressed, but the prediction of readmission has been less explored. In fact, in the state of the art, no models specifically designed to make medium-term readmission predictions (2–3 months after admission) have been found. This work presents a new intelligent clinical decision support system to predict the risk of hospital readmission in 90 days in patients with COPD after an episode of acute exacerbation. Methods: The system is structured in two levels: the first one consists of three machine learning algorithms —Random Forest, Naïve Bayes, and Multilayer Perceptron—that operate concurrently to predict the risk of readmission; the second level, an expert system based on a fuzzy inference engine that combines the generated risks, determining the final prediction. The employed database includes more than five hundred patients with demographic, clinical, and social variables. Prior to building the model, the initial dataset was divided into training and test subsets. In order to reduce the high dimensionality of the problem, filter-based feature selection techniques were employed, followed by recursive feature selection supported by the use of the Random Forest algorithm, guaranteeing the usability of the system and its potential integration into the clinical environment. After training the models in the first level, the knowledge base of the expert system was determined on the training data subset using the Wang–Mendel automatic rule generation algorithm. Results: Preliminary results obtained on the test set are promising, with an AUC of approximately 0.8. At the selected cutoff point, a sensitivity of 0.67 and a specificity of 0.75 were achieved. Conclusions: This highlights the system’s future potential for the early identification of patients at risk of readmission. For future implementation in clinical practice, an extensive clinical validation process will be required, along with the expansion of the database, which will likely contribute to improving the system’s robustness and generalization capacity.

Background and Objectives: The diversity of patients with malignant pleural effusion (MPE) due to non-small cell lung cancer (NSCLC) as well as the variability in mutations makes it essential to improve molecular characterization. Objective: Describe clinical, pathological, and molecular characteristics MPE in a Caucasian population. Materials and Methods: Retrospective study of patients with NSCLC diagnosis who had undergone a molecular study from 1 January 2018–31 December 2022. Univariate analysis was performed to compare patient characteristics between the group with and without MPE and molecular biomarkers. Results: A total of 400 patients were included; 53% presented any biomarker and 29% had MPE.PDL1, which was the most frequent. EGFR mutation was associated with women (OR:3.873) and lack of smoking (OR:5.105), but not with MPE. Patients with pleural effusion were older and had lower ECOG. There was no significant difference in the presence of any biomarker. We also did not find an association between the presence of specific mutations and MPE (22.4% vs. 18%, p = 0.2), or PDL1 expression (31.9% vs. 35.9%, p = 0.3). Being younger constituted a protective factor for the presence of MPE (OR:0.962; 95% CI 0.939–0.985, p = 0.002), as well as ECOG ≤ 1 (OR:0.539; 95% CI 0.322–0.902, p = 0.01). Conclusions: This is the first study that describes the clinical, pathological, and molecular characteristics of MPE patients due to NSCLC in a Caucasian population. Although overall we did not find significant differences in the molecular profile between patients with MPE and without effusion, EGFR mutation was associated with a tendency towards pleural progression.

Objective: This study aims to employ machine learning (ML) tools to cluster patients hospitalized for acute exacerbations of chronic obstructive pulmonary disease (COPD) based on their diverse social and clinical characteristics. This clustering is intended to facilitate the subsequent analysis of differences in clinical outcomes. Methods: We analysed a cohort of patients with severe COPD from two Pulmonary Departments in north-western Spain using the k-prototypes algorithm, incorporating demographic, clinical, and social data. The resulting clusters were correlated with metrics such as readmissions, mortality, and place of death. Additionally, we developed an intelligent clinical decision support system (ICDSS) using a supervised ML model (Random Forest) to assign new patients to these clusters based on a reduced set of variables. Results: The cohort consisted of 524 patients, with an average age of 70.30±9.35 years, 77.67% male, and an average FEV1 of 44.43±15.4. Four distinct clusters (A-D) were identified with varying clinical-demographic and social profiles. Cluster D showed the highest levels of dependency, social isolation, and increased rates of readmissions and mortality. Cluster B was characterized by prevalent cardiovascular comorbidities. Cluster C included a younger demographic, with a higher proportion of women and significant psychosocial challenges. The ICDSS, using five key variables, achieved areas under the ROC curve of at least 0.91. Conclusions: ML tools effectively facilitate the social and clinical clustering of patients with severe COPD, closely related to resource utilization and prognostic profiles. The ICDSS enhances the ability to characterize new patients in clinical settings.

Objetivo Diseñar un sistema inteligente de soporte a la decisión clínica (SISDC) para predecir la mortalidad a un año tras una exacerbación aguda de EPOC. Material y métodos Se utilizó una base de datos con 524 pacientes, recopilada entre 2018 y 2022 en los hospitales Álvaro Cunqueiro y Lucus Augusti, que incluye variables clínicas, demográficas y sociales. Para el desarrollo del SISDC, primero se abordó un proceso de selección de características mediante Regresión Lasso (RL), para identificar las variables con mayor poder predictivo. Posteriormente, se desarrolló un contenedor estructurado pseudo-simbólico que encapsula la información de cada paciente. Su construcción contempla dos pasos: (1) se definió una estrategia para crear un lenguaje común sonoro basado en las características del paciente, y (2) se calculó el espectrograma asociado al sonido, generando una imagen por paciente. Estas imágenes, que actúan como pseudo-símbolos al integrar datos y conocimiento, fueron analizadas mediante una Red Neuronal Convolucional, SqueezeNet, entrenada para calcular el Riesgo de Mortalidad. El análisis y procesamiento de los datos se realizaron en Python© y MATLAB©, reservando un 10% para pruebas. Resultados La RL permitió seleccionar las variables más relevantes, ordenadas según importancia: oxigenoterapia domiciliaria, número de actividades elaboradas para las que necesita ayuda, edad, IMC, disnea mMRC, número de actividades básicas para las que necesita ayuda, número de hospitalizaciones año previo, FEV1 porcentual, vivienda propia y tabaquismo activo. Posteriormente, se implementó una aplicación. En el conjunto de prueba, se alcanzó un área bajo la curva ROC próxima a 0.85. Conclusiones La propuesta es eficaz para predecir la mortalidad a un año, integrando variables de distintas esferas. Tiene el potencial para mejorar la planificación de los cuidados al final de la vida. No obstante, en el futuro será necesario realizar nuevos estudios para validar el SISDC.

Objetivo Identificar perfiles clínico-sociales en pacientes hospitalizados por agudización de EPOC mediante Machine Learning (ML) no supervisado, evaluar su impacto en el consumo de recursos y pronóstico, y desarrollar un Sistema Inteligente de Soporte a la Decisión Clínica (SISDC) que permita asignar nuevos pacientes a los grupos con un conjunto reducido de variables. Material y métodos Se analizó una cohorte de 524 pacientes hospitalizados por AEPOC entre 2018 y 2022 en dos hospitales (Álvaro Cunqueiro y Lucus Augusti), con datos demográficos, clínicos y sociales. Se utilizaron k-prototypes y el método del codo para identificar cuatro clústeres, evaluando sus diferencias con pruebas estadísticas. Posteriormente, se desarrolló el SISDC basado en Random Forest y Recursive Feature Elimination, el cual se implementó en una aplicación web. Resultados Se identificaron cuatro clústeres. El clúster A, el más frecuente, incluye a varones con una enfermedad menos grave, que residen en zonas rurales, con buenas relaciones sociales y baja dependencia, mostrando un consumo moderado de recursos y un pronóstico favorable. El clúster B, formado por pacientes mayores, principalmente varones, con comorbilidades cardiovasculares y un alto uso de oxigenoterapia, residentes en áreas urbanas, presentó un riesgo intermedio de reingreso y mortalidad, respaldado por una mejor situación económica y social. El clúster C, el más joven, con una mayor proporción de mujeres, bajo IMC, tabaquismo y ansiedad, mostró un índice considerable de reingresos, aunque sin un pronóstico marcadamente desfavorable. Finalmente, el clúster D, compuesto por pacientes más dependientes y con relaciones limitadas al núcleo familiar, mostró el peor pronóstico y mayor consumo de recursos. El SISDC demostró una alta capacidad predictiva, asignando nuevos pacientes con solo 5 variables y alcanzando AUC superiores a 0.9. Conclusiones El uso de técnicas de ML permitió identificar perfiles heterogéneos de pacientes, subrayando la importancia de los factores sociales.

Introducción y objetivo La pandemia de COVID-19 ha transformado drásticamente la visión sobre la integración de la tecnología en la medicina, poniendo a prueba la capacidad de los sistemas de salud para tomar decisiones rápidas y efectivas, especialmente en servicios de urgencias desbordados. Presentamos un Sistema Inteligente de Soporte a la Decisión aplicado en el manejo de pacientes sospechosos de padecer neumonía COVID-19 a su llegada al servicio de urgencias. Material y método Se despliegan dos módulos que discurren de forma secuencial, basados en Redes Neuronales Convolucionales: (fig.1) El primero: DIAGNÓSTICO: escala 0-100, emplea DenseNet-121, para identificar si se existe neumonía en la Radiografías de Tórax (RT) El segundo: SEVERIDAD, utiliza COVID-Net CXR-S, busca cuantificar la severidad de la opacidad del espacio aéreo en la RT (ASOS): escala 0-24. Esta arquitectura plantea la Generación de Alertas y Toma de Decisiones: • Presenta neumonía y 0 ≤ ASOS < 5: seguimiento • Presenta neumonía y 5 ≤ ASOS < 15: hospitalización. • Presenta neumonía y 15 ≤ ASOS ≤ 24: ingreso en UCI. Base de RT Povisa: 2438 RT:1450 neumonía (59.5%) /988 no neumonía (40.5%).533 RT para pruebas test en el primer módulo y 200 para el segundo. Resultados Los resultados obtenidos en el conjunto de prueba: (fig. 2) AUC > 0.9 para módulo DIAGNÓSTICO. Root Mean Squared Error próximos a 6 para módulo SEVERIDAD. Conclusiones Estos hallazgos subrayan la capacidad del sistema para actuar como aliado fiable en la lucha contra el COVID-19, mejorando significativamente la gestión de pacientes en entornos de urgencia. En el futuro, se plantea la necesidad de expandir y profundizar en la validación de este sistema en otras patologías, además de desarrollar técnicas de explicabilidad