Mi SciELO
Servicios Personalizados
Articulo
Indicadores
- Citado por SciELO
Links relacionados
- Similares en SciELO
Compartir
Ingeniería Electrónica, Automática y Comunicaciones
versión On-line ISSN 1815-5928
Resumen
CRUZ CORZO, Hanlert; PEREZ DIAZ, Marlen y LOPEZ CABRERA, José Daniel. Potential of YOLOv5 to detect lung nodules on chest x-rays. EAC [online]. 2023, vol.44, n.1, pp. 31-46. Epub 06-Dic-2023. ISSN 1815-5928.
Chest radiography is one of the most widespread methods for the identification of pulmonary nodules. However, they are difficult to interpret due to their low contrast and the amount of overlapping anatomical structures in the thoracic region. Computer-aided diagnostic (CAD) systems increase the effectiveness of diagnoses and reduce the workload of specialists. This research proposes a CAD system based on artificial intelligence, where chest X-rays are used for the detection of pulmonary nodules. A network with a YOLO method was used on which transfer learning techniques and three training strategies were applied. Image sets were created from four international databases. The network was trained and validated, and an external test was carried out for the best model obtained, based on a fifth highly difficult database of a different origin than the previous ones. The best model was obtained with the training for the set of segmented images, including the mediastinal area, with a sensitivity of 68% in the external test, which is comparable to previus studies. The approach presents potential and also the advantage to visualize and filtering the model confidence.
Palabras clave : chest x-ray; lung nodule; artificial intelligence; deep learning; YOLO.