Services on Demand
Journal
Article
Indicators
Cited by SciELO
Access statistics
Related links
Cited by Google
Similars in SciELO
Similars in Google
Share
Ingeniería y Desarrollo
Print version ISSN 0122-3461On-line version ISSN 2145-9371
Abstract
ASTUDILLO DELGADO, VICTOR MANUEL and REVELO LUNA, DAVID ARMANDO. Support for the Diagnosis of Pneumonia and Detection of Pulmonary Opacities Using Semantic Instance Segmentation in Chest X-ray Images. Ing. Desarro. [online]. 2021, vol.39, n.2, pp.259-274. Epub Apr 23, 2022. ISSN 0122-3461. https://doi.org/10.14482/inde.39.2.621.367.
Pneumonia is a disease that has caused many deaths worldwide. Pneumonia detection methods: blood tests, sputum test, CT scan, and chest X-rays, generally require a specialist doctor for analysis. The objective of this research was to generate a model, based on semantic instances and deep learning (Mask-RCNN), that allows for support in the diagnosis of pulmonary opacities and pneumonia, using i2 o24 images of chest X-ray radiographs. 3 experiments were carried out, where the conditions of the dataset images were varied (expi: patients with pneumonia, exp2: healthy, and, also, pneumonia patients, exp3: healthy, and, also, pneumonia patients, plus data augmentation). For all 3 experiments, a histogram equalization pre-processing technique was performed. In order to evaluate the performance of the models, the parameters Intersection on the Union, Precision, Recall, Fi score, and Accuracy were used. In the identification of pulmonary opacities in the images, and for the classification of patients with pneumonia from healthy ones, Recall was found to be the best performing metric for experiment i, for both opacity and pneumonia detection.
Keywords : convolutional neural networks; data augmentation; junctional intersection; radiological society of North America; ResNet5o.