Authors

Inspiration

Instantechnologies is an innovative startup born from the experience of its founders, over 25 years in the fields of semantics and artificial intelligence; it is part of the Tecnopolo di Mirandola - the second most important biomedical district in the world. He has decided to make his own skills, technologies and proprietary algorithms available to help him find solutions to deal with the virus that is currently raging in Italy, Europe and the world.

Challenges we ran into

Hack the Virus

SOLUTIONS TO HELP TO FACE THE VIRUS: 1) Creation of predictive models for the forecast of infections 2) Creation of a tool to identify patients through image recognition of RX and TC

... to be made available to support both medical and governmental analysis and strategic processes to deal with the ongoing emergency Covi

What it does

1) Creation of predictive models for the forecast of infections

The objective of this first analysis is to estimate the short and long-term impact of the infection on each Italian province initially and extendable to the rest of the world, in order to provide indications on the possible need for reception over time. , on the possibility of anticipating and setting up structures and personnel as well as allowing work on further forms of prevention and containment.

2) Creation of a tool to identify patients through image recognition of RX and TC

Use of semantic, AI and image recognition algorithms to analyze both pulmonary CT and chest X-rays of patients for the diagnosis and monitoring of Covid-19 pneumonia in a few minutes (between scanning and analysis) unlike the 3 days required for the swab.

How we built it

1)Creazione di modelli predettivi per la previsione dei contagi

Diversamente da quanti propongono analisi previsionali, IT ha messo a punto un modello di AI che non basa le proprie previsioni sull'andamento dei dati storici.

1) Creation of predictive models for the forecast of infections

Unlike those who propose forecast analyzes, IT has developed an AI model that does not base its forecasts on the trend of historical data.

After a careful initial analysis, the model was developed considering the semantic meaning of the data, a feature that distinguishes IT, profiling citizens' behaviors and therefore starting from the causes of diffusion rather than effects.

To exemplify the functioning of the model, the probabilities of contracting the virus by each citizen and the possibility of it becoming a vehicle of contagion were considered. People were divided into asymptomatic and in need of care considering the incubation times and the number of possible contacts maintained by each of them in the different periods.

In the initial stages, the automatic intelligent analysis proposed some dozens of possible scenarios that have drastically reduced over the days thanks to the feedback received from the feedback of real trends. After 5 days, the system has reached a maturity which today allows us to provide estimates of the diversity of contagion progressions in every area of ​​Italy. The system continues to hone its forecasting capacity every day and the same model is currently being transposed to other countries.

2) Creation of a tool to identify patients through image recognition of RX and TC

The artificial intelligence system is able to process the image and provide a response in a few seconds starting from the analysis of the images of the pulmonary CT and / or chest x-ray. This tool is designed to detect lung abnormalities in patients with the first symptoms of Covid-19 before they degenerate into respiratory problems or pneumonia. The timing in identifying the symptoms and the time of elaboration of the analysis are fundamental factors in the identification of the disease and in the hospital course.

Accomplishments that we're proud of

1) Creation of predictive models for the forecast of infections

As anticipated, the implemented system offers a point of view that we would like to direct in a complete and preferential way towards national, regional and provincial emergency managers. Secondly, a simplified public display, limited and supervised by the competent authorities, also without long-term forecasts, can represent a useful indicator to inform and make citizens aware of short-term progress and developments.

2) Creation of a tool to identify patients through image recognition of RX and TC

We are specializing this tool in collaboration with the Department of Health Physics of the Mantua and Cremona hospitals.

What we learned

1) Creation of predictive models for the forecast of infections

The model is dynamic, that is, the external changes of the social system are considered (implementation of restrictions, habits, activation of forms of protection, transport restrictions, etc.) from which behavioral changes derive. The ecosystem considered is therefore very different from that of a traditional AI analysis in that it assesses and foresees social changes for which historical data are not available.

2) Creation of a tool to identify patients through image recognition of RX and TC

This tool allows hospitals to be more effective in the diagnosis and management of wards, thus reducing the number of patients who need to be admitted to the intensive care unit.

What's next for AI tools to predict contagion and identify disease(by CT&XR)

1) Creation of predictive models for the forecast of infections

Improvement of Data Visualization and r system spead to Intarantional Governative Organization, with private access to the dashboard.

2) Creation of a tool to identify patients through image recognition of RX and TC

We want share this project with Medical Organization and Hospital

Try It out

Hackathons

Technologies

ai, imagerecognition, property-solutions, semantic

Devpost Software Identifier

255068