The results of the CIRCA project research have been published in Scientific Reports!
We are very pleased to announce that the findings of the work carried out within the CIRCA project have been published in the prestigious journal Scientific Reports (https://www.doi.org/10.1038/s41598-025-26895-0). CIRCA is an online tool (circa.aei.polsl.pl) developed to support the analysis of chest X-ray images for diagnosing COVID-19 or pneumonia of other origins.
A key innovation of the tool is the quantitative assessment of medical-image heterogeneity by identifying radiomically distinct subtypes within each class, which ensures better model generalization. CIRCA employs a hierarchical decision system that aggregates predictions from image-based and radiomic-feature-based models after preliminary segmentation of the lung region. On an independent test set, CIRCA achieved a classification accuracy of 86% (normal class) and 83% (COVID-19). The system demonstrated an accuracy approximately 10% higher than the COVID-Net algorithm. The developed algorithm can be adapted for the analysis of images of other complex, heterogeneous lung diseases.
