COVID PATTERNS

COVID-19 Situation in Mexico

Since the health emergency caused by the SARS-CoV-2 virus, Mexico has experienced an economic, social and health crisis never seen before.

According to data obtained by the Secretary of Health of the mexican government, more than 248,167 deaths have accumulated as of March 2020.
As a result of this situation, the scientific community has been actively involved in the search for solutions for every aspect affected by the COVID-19 pandemic.

Main Objective

Our goal is to develop a neural network capable of predicting the possible mortality from COVID-19 disease in mexican population.

To achieve this, we evaluate clinical data of the patient as well as the lung damage observed in the chest tomography.

Project Development Areas

Neural Network

We are adapting neural net architectures to identify damage patterns specific to COVID disease through a multimodal input.

We aim to expand model input to take categorical and continuous clinical features from a dataset of mexican COVID affected patients mixed with information extracted from CT scans to provide better disease prognosis.

Clinical Data Analysis

A compilation of previously selected clinical variables was carried out according to their importance in the prognosis of COVID-19 mortality.

Approximately 300 patient records were used to perform a statistical analysis and evaluate their versatility in the prediction model. Five machine learning analytical models were implemented to assess the accuracy of the prediction model.

Computed Tomography Analysis

We are exploring both explainable and highly abstract feature extraction techniques to find key relationships between COVID mortality and information available in CT-Scans.

We are developing a LungNet model tailored to the mexican population to automatically extract lung damage percentages.