Artificial Intelligence In Medical Imaging: A Meta-Analysis On Impacts In Early Diagnosis

Abstract

Artificial intelligence (AI) has profoundly transformed radiology and diagnostic imaging, offering computational tools capable of identifying patterns with high accuracy. This meta-analysis synthesizes the findings of three scientific studies indexed in SciELO, covering the use of machine learning and deep learning algorithms in the diagnosis of eye diseases, pulmonary nodules, breast cancer, and various lesions in imaging exams. The analyzed studies demonstrate that systems based on convolutional neural networks (CNNs) can achieve diagnostic accuracies equal to or greater than those of human experts in specific visual pattern recognition tasks. The study by Abed and Al-Bakry (2024) demonstrated 99.9% accuracy in classifying eight eye diseases through fundoscopy. The works of Santos et al. (2019) and Koenigkam-Santos et al. (2019) consolidate the theoretical foundations of AI applied to radiology, addressing everything from computer-assisted diagnosis to radiomics and precision medicine. The integrated analysis of these findings points to consistent benefits in increasing diagnostic sensitivity, reducing false negatives, and optimizing clinical workflow, especially in cancer screening. It is concluded that AI represents an essential complementary resource to the work of the radiologist, with the potential to expand access to early diagnosis, although further multicenter studies with prospective data are still needed for validation in diverse clinical scenarios.