Deep Learning for Early Detection of Pneumonia: Systematic Review

Abstract

Pneumonia is a leading cause of illness and death worldwide, and early diagnosis is essential to improve patient outcomes. However, in many hospitals, the lack of advanced technology and the use of manual methods to detect the disease make accurate identification difficult. In this context, Deep Learning applied to the analysis of medical images, such as chest X-rays, is presented as a promising solution to improve both diagnostic speed and accuracy. This article reviews the literature on the use of Deep Learning models in early detection of pneumonia, showing that these models outperform or match the performance of expert radiologists. The PICOC methodology and the PRISMA diagram were used to select 30 relevant studies from the last five years. Although progress has been made, challenges remain, such as variable image quality and adaptation to different clinical contexts. The findings suggest that Deep Learning contributes to the reduction of diagnostic times and improves accuracy, but there are still areas to be improved for optimal integration into clinical practice. These tools are expected to become more adaptable and continue to improve their accuracy over time