MIA-COV19D is a competition organized by the University of Lincoln to develop a novel solution to effectively detect COVID-related diseases in CT scans of patients’ lungs. The competition has been organized within the context of ICCV 2021, the premier international computer vision event within the academic community.
Our data scientists in synbrAIn have been working extensively to such a task, developing product-ready solutions that have been already integrated in MS HUMANAID (as mentioned in this post). In particular, our work is in strict collaboration with our partner SenticLab, which joined the MIA-COV19D competition with great results.
SenticLab developed a customized solution for detecting COVID-19 in CTs based on deep learning, providing an implementation that achieves a macro F1-score higher than 90%, consistently outperforming the competition baseline set at 70%. Results showed a very slight difference compared to the one obtained by the winning team, but in general this approach seems to be promising, and both our teams are working together to integrate new improvements.
We are proud of such an achievement, especially considering the importance that such findings might bring to all of humanity. It is also for these reasons that SenticLab team has described the whole approach in a paper that is already available on arXiv.
These results reconfirm the position of our consortium as spearhead in the field of AI-assisted medical diagnosis. We believe the artificial intelligence can bring exceptional and concrete advantages to our society, representing an almost indispensable tool for the human-machine cooperation.