CORDA: Covid ​R​adiographic images ​D​ata-set for ​A​I

step del progetto corda

CORDA: Covid ​Radiographic images ​Data-set for ​A​I

CORDA has started as a collaboration between the Radiology 2 unit of Città della Salute e della Scienza and the EIDOSLab group in our department and is open to further collaborations. 

logo: EIDOSLab
logo: asl TORINO
Logo: Azienda ospedaliero-universitaria (città dell salute e della scienza di Torino)
logo: A.O ordine Mauriziano Torino


In the CORDA project we investigate the possibility of using widespread Chest X-Ray (CXR) imaging for screening COVID-19 patients under-going radiology examination. 
Our goal is to investigate the potential of using deep learning to recognize COVID from CXR images. Currently, there is no clinical evidence that CXR images could effectively be used in the fight against COVID. At the same time, there are currently few, small public datasets that can be fed into powerful deep learning models.


We are currently pursuing the following goals:
• Build the CORDA dataset: retrospectively select CXRs performed at Radiology 2 Unit in patients with fever or respiratory symptoms (cough, shortness of breath, dyspnea) undergoing nasopharingeal swab to rule out COVID-19 infection;
• Training state-of-the-art deep learning models for image classification using all the available data (CORDA and other public datasets);
• Discuss the results in the interdisciplinary team to drive our effort toward clinically-useful purposes, targeting the post-emergency scenario.

Our study presented on the University Research Forum.


Our preliminary results are discussed in the following pre-print:
Tartaglione, E., Barbano, C.A., Berzovini, C., Calandri, M. and Grangetto, M.,Unveiling COVID-19 from Chest X-ray with deep learning: a hurdles race with small data, ArXiv preprint