Artificial intelligence against COVID-19 Pandemic using Chest X-ray Images

This talk will dive into the AI for COVID initiative, a multicentre research project aimed at supporting the development and promoting the use of innovative AI-based methods to predict clinical outcomes of SARS-CoV-2-related disease. In the talk we will first discuss three AI-based approaches that use features extracted from CXR images, either handcrafted or automatically learnt by convolutional neuronal networks (CNNs), which are then integrated with the clinical data in a multimodal fashion. Furthermore, the talk will also present another application of the same repository, which is used to test a new late fusion approach combining the outputs of several state-of-the-art CNNs. It is driven by a two-objective function that constructs an optimum ensemble determining which and how many base learners should be aggregated, by maximizing the accuracy and the diversity of the ensemble itself.