Algorithms to Aid in the Diagnosis of Lung Diseases Using Artificial Intelligence

EasyChair Preprint 7857

Abstract

The world has been hit by a pandemic caused by the coronavirus that has impacted all of humanity. The disease caused by this virus affects several systems of the human body, among them the respiratory system is the one that suffers the most, leading to the formation of consolidations and ground-glass opacity in the lungs. These symptoms can be detected in radiology exams. To assist the medical community in diagnosing Covid-19, several works were developed with the aim of generating new tools and methods. In this work, a Covid-19 detection system and probabilistic Grad-CAM generation was developed, using the neural network models ResNet50V2, DenseNet121, InceptionResnetV2 AND VGG-19. The results of the two models were compared using the precision and specificity of each one for Covid-19, where the DenseNet121 network obtained precision values of 99.28% and the ResNet50V2 a specificity of 99.72%, which were higher than those obtained in the reference literature.

Keyphrases: Artificial Inteligence, COVID-19, Segmentation, deep learning, machine learning

@booklet{EasyChair:7857,
  author    = {Gustavo Chichanoski and Maria Bernadete França},
  title     = {Algorithms to Aid in the Diagnosis of Lung Diseases Using Artificial Intelligence},
  howpublished = {EasyChair Preprint 7857},
  year      = {EasyChair, 2022}}