Covid@Keydiag, a collaboration to assess quickly the severity of Covid infection

The CovidAKeydiag project was born out of the circumstances of the epidemic but takes its roots several years earlier, from the meeting between two people: Dr Jean-Noël Ravey, head of the osteo-articular imaging department of Grenoble University Hospital, and Lionel Reveret, expert at Inria in anatomical motion modeling. Functional anatomy was the common space for dialogue, and the aim was to document the needs of the field to guide research. Between the researcher waiting for an echo and an outlet for his work, and the physician who undergoes a lot of digital tools because he cannot participate in their design, they converged on a common vision based on concreteness, complementarity, and trust. Since then, the researcher has systematized a methodology drawing directly from the field, and the physician wanted to regain control over his tools by creating a company publishing medical software (Keydiag).

This software is at the start of the project: it models medical reasoning by a path of key criteria, through structured reports, which allow radiologists to work quickly and well, while accessing knowledge on a rapidly evolving and easily deployable web support: assets that are perfectly adapted to the epidemic context.

His team developed in a few days a module specific to Covid-19 disease, made available free of charge to French radiologists, in order to widen the bottleneck represented by the exponential flow of patients to carry out such an examination: the chest scanner is a key step in documenting serious or equivocal forms of Covid, and the emerging disease is therefore little known.

But if the application is a workflow tool, it also collects valuable metadata: the technical terms present in the reports. These could become valuable indicators of the number and severity of cases, in the complete anonymity of patient data (which does not pass through the platform), provided that they could be aggregated and represented in an ad hoc format: we had at our fingertips a scientific and health monitoring tool, in real time, on the epidemic front line.

Today, the project is progressing very satisfactorily with a group dedicated to exploiting data from Covid scanners, involving the Keydiag team (including a data scientist) reinforced by Inria research engineers, including Lionel Reveret, who is coordinating the collaboration in tandem with Jean-Noël Ravey. The landmarks were quickly found because of the previous relationship. An export and classification work on Keydiag ontology including the different versions made over time made it possible to have more than 7,500 thoracic scanner reports, each indexing 575 variable values. The sampling is correlated with national data, allowing its epidemiological interest to be validated. An audit was carried out to ensure the safety of the platform despite its accelerated deployment, as well as an expertise of the regulatory framework. Finally, a web-based data dashboard will be available very soon, allowing to accelerate the work on data mining: 3 scientific publications are in preparation. The main objective was to visualize the data from Keydiag reports for 2 reasons: to better understand the disease on one hand, and to follow its evolution in real time based on specific medical criteria on the other hand. It is therefore in the process of becoming a reality.

After a phase of team building and role allocation, we alternate individual work and regular synchronisation steps by videoconference. Knowing each other has allowed us to quickly find the right tempo and avoid meeting up: we already had the Covid emergency, we keep the oxygen to move forward!

 

Inria is perfectly fulfilling its investment in applied research: in this project people are invested in quickly producing a deliverable applicable to the reality of the Covid crisis. When everyone gets down to work and gets straight to the point, it's really useful and ultimately resource-saving because the effort is well distributed.

This model of collaboration is really interesting because it is oriented towards the immediate needs of the field without sacrificing the professionalism necessary to avoid nonsense in this sensitive field that is health. It demonstrates that not only can we obtain quick and concrete results by taking calculated risks, doing a service to patients and physicians, but also bring solutions with very high potential in the future.

Finally, the Covid opens our eyes to what is important, to what really helps, and what works. It calibrates people and technology. This Covid mission from Inria, seen from my side on this CovidAKeydiag project, it provokes a deconfinement of intelligence, an experience of collective intelligence.

Dr Jean-Noël Ravey, CEO of Keymaging