Laguerre Gauss Kernel for COVID-19 Medical Decision Making From Chest Tomography

Abstract

As March 22nd Colombia had 235 confirmed cases of COVID-19. We already had decided to put together of national mathematicians and engineers along with medical doctors to quickly answer to potential needs on the development of a tool for decision-making support. As well known, several attempts to provide a deep learning solution to early diagnosis tools were not successful firstly for the lack of formalism and medical validity of their findings and secondly, because as known for the mathematics and engineers a learning machine must fulfill bounds for learning depending on its complexity. Consequently with no database fulfilling neither the medical standards nor the guarantees of learning, these attempts are far to be useful. Both medical and mathematical conditions for an artificial intelligence learning machine are going to be satisfied during our work. Looking for the construction of a solid approach to the problem, this paper contains the Laguerre-Gauss kernel applied to computerized
tomography images looking for feature extraction in order to provide a preprocessing tool that reduces the complexity of the learning machine, and in consequence reducing the dataset needed to reach the medical doctors standards of sensitivity and specificity. We developed it in Paniagua et al, 2016, 2017, Paniagua, 2018, and proposed in Quintero et al, 2020 as a tool for deep learning in Medical Applications. Our results are promising and will be proved and tested on several deep learning algorithms and the interpretability of the particular structures in lungs generated by the virus.