New diagnostic technology for COVID-19 in chest X-rays

Researchers from the Institute of New Technologies - UNINOVA - Prof. Dr. José Barata, Prof. Dr. Ricardo Peres, and Leandro Filipe, have achieved a significant breakthrough in the diagnosis of COVID-19 by utilizing Vision Transformers (ViTs), a form of Artificial Intelligence (AI) that analyzes images by dividing them into smaller sections, enabling a more comprehensive understanding of the images and, consequently, intelligent decision-making based on this analysis.

This study achieved an impressive accuracy of 93.8% when analyzing chest X-rays through the application of ViTs, surpassing traditional AI models such as Convolutional Neural Networks. The implemented method provides a deeper context of the images, crucial for accurate diagnosis. This not only streamlines the process but also enhances the reliability of the results, assisting healthcare professionals in making informed decisions more rapidly. Additionally, the study incorporated Explainable AI (XAI) techniques to enhance the decision-making process based on AI. These techniques allow visualization of the model's decision process, highlighting the areas of the image that most influence its decision. The use of XAI in medicine increases transparency and trust in automated diagnoses by physicians and nurses.

While the results are promising, the researchers advocate for the continuous integration of AI technologies in healthcare to better prepare for future pandemics. This advancement represents tangible hope in the fight against COVID-19 and opens doors to a more efficient and accurate era in medical diagnosis.