LearnFromCOVID-19

LearnFromCOVID-19

Illustrative project image
Approved by mentor

Medical tags

Clinical need
Remote or self-diagnosis
Area
Preventive medicine
Technology
Software
Project keywords
Machine learning, Artificial Intellingence (AI), DesignCompetition2020
Device classification
IIa

Project description

An early and rapid AI-diagnostic tool based on clinical features for the identification of new potential COVID-19 cases


Introduction

During this period of global emergency due to COVID-19 epidemic, the development of technological solutions is crucial to help and improve the efficiency of the overload healthcare system. The proposed project LearnFromCOVID-19 answers to this need by providing a machine learning algorithm that identifies new potential COVID-cases thanks to the smart learning of the clinical status of already established patients.


Description


The aim of the project is to develop a clinical diagnostic tool, based on artificial intelligence, to provide a rapid and early diagnosis of the disease to identify and remotely monitor the state of COVID-19 patients. In this way, the system overcomes the problem of the high number of patients to be managed, the lack of  swabs and the urgent need to reduce the spread of infection. 

In order to do that, this project consists of three key points :

  • Creation of a database of clinical features of suspected cases  previously recorded in hospitals;
  • Development of a machine learning tool to predict the swab test results;
  • Creation of an user-interface for new subject with suspicion of COVID-19.


The database will contain the following clinical features: age, gender, epidemiological factors, comorbidities, clinical assessed symptoms and vital signs. These data will be gather from the medical records  of  COVID-19 suspected cases who underwent swab test at the hospital,  with the utmost respect for privacy.  

The data collected will be used to train the machine learning algorithm and to evaluate its performance. 

Once found the best classification rule, an user-friendly interface will allow to use  the algorithm as follow: the user uploads independently, from his home, the clinical features reported above, in particular  the vital signs will be easily recorded using simple tools available in pharmacy (e.g. thermometer, spO2 sensor, blood pressure sensor,...). Once found the best classification rule, an user-friendly interface will allow to use the algorithm as follow: the user uploads independently, from his home, the clinical features reported above, in particular the vital signs will be easily recorded using simple tools available in pharmacy (e.g. thermometer, spO2 sensor, blood pressure sensor,...). This tool will be widely distributed to the population as a web app in order to predict the potential new cases and hence report the subjects who need to undergo the swab test. The family doctor gives the credentials (username and password) to his patients to login to the web app. The doctor is able to monitor each patient whatever the prediction is positive or negative suggesting him the following steps. The doctor has the access to the patient's clinical history and he can remotely monitor specific clinical features evaluating if the patient needs to redo the test or not.


In this way, the population under clinical control will strongly increase and the number of swabs will be targeted only to the potential positive cases predicted by the algorithm.

Moreover, this innovative tool will help to stop the spreading of the disease, indeed a rapid and early diagnosis of a potential COVID-19 case will allow him to take the right measures even before undergoing the swab test.  

Last but not least, thanks to this system the work of all the medical personnel which fight everyday on the front-line will be lightened.