An electrocardiogram (ECG) is a noninvasive study used for the diagnosis of cardiac arrhythmias (CAs). The identification of a cardiac arrhythmia depends on its classification. This classification has been approached through different strategies, both mathematical and computational. In this work, a new computational model based on the particle swarm optimization (PSO) algorithm and convolutional neural network (CNN) is proposed for the classification of five classes of CAs obtained from the MIT-BIH Arrhythmia Dataset (MITDB). The goal of the PSO is to optimize the hyperparameters that define the layered architecture of a CNN, to increase the accuracy and decrease the categorical cross-entropy (CE) error. The proposed model found a satisfactory layered architecture in 17.68 hours, obtaining an Accuracy of 98% and 97%, a CE of 0.044968 and 0.084768, in training and testing, respectively. These results demonstrate that the proposed model is reliable and also represents an innovative approach because it allows dispensing with the manual selection of the hyperparameters of the layered architecture of a CNN.

Una arritmia cardíaca es un latido irregular del corazón que se traduce en un impulso eléctrico anormal, y su tipo se define por el ritmo y duración. Su clasificación ha sido abordada en diferentes campos de la ciencia, destacando el uso de algoritmos de aprendizaje profundo. La presente investigación, utilizó un modelo híbrido entre Redes Neuronales Convolucionales y el algoritmo metaheurístico de Optimización por Enjambre de Partículas; para la clasificación de arritmias cardíacas. El metaheurístico se encargó de optimizar la arquitectura de capas de la red neuronal, a través de la minización de la pérdida durante el entrenamiento y prueba. Los datos se obtuvieron del MIT-BIH Arrhythmia dataset, donde se describen cinco categorías de arritmias. Los resultados logrados demostraron que el metaheurístico es un algoritmo confiable en la búsqueda de la mejor arquitectura de capas, logrando obtener una exactitud del 97%, lo que significa que el uso de técnicas metaheurísticas es una opción que se debe tomar en consideración a la hora de optimizar el rendimiento de las redes neuronales convolucionales.

Background: Periodontal diseases continue to be highly prevalent worldwide, and their early detection represents a clinical challenge, especially when based on non-standardized intraoral photographs. The present study develops an artificial intelligence-based prototype for the automatic classification of periodontal health, gingivitis, and periodontitis using Red, Green, Blue (RGB) images obtained in real conditions. Methods: A dataset comprising 1552 (306 healthy, 1019 with gingivitis, and 227 with periodontitis) was constructed by integrating proprietary clinical photographs with a public repository. A patient-level stratified split was enforced to prevent data leakage, ensuring that all images from the same patient remained within a single partition. This proposal uses EfficientNet-B2, which includes two-phase training, balanced focal loss, weighted sampling, CutMix/MixUp augmentation, and centered anatomical cropping to improve generalization across varied images. Results: The final model achieved an accuracy of 0.833, a macro F1-score (F1) of 0.832 [95% CI: 0.789–0.874], and a macro Area Under the Curve (AUC) of 0.962 [95% CI: 0.946–0.976] on an independent test set. A seven-configuration ablation study showed that each training component contributes to improved performance, and a baseline comparison with ResNet-50 demonstrated the superiority of EfficientNet-B2. Five-fold cross-validation with patient-level grouping yielded consistent results (F1 = 0.832 ± 0.016, AUC = 0.950 ± 0.005). Conclusions: These results demonstrate that EfficientNet-B2 is useful for assessing periodontal health using readily available RGB photographs, with potential for early detection, clinical triage, and remote assessment in modern dentistry.