Collective intelligence assisting healthcare

Date :
Changed on 16/04/2020
Mathematicians from Inria Lille-Nord Europe spent the best part of three years collaborating with a multinational from the health sector. Their research led to the development of a tool to help design medical analysis laboratories employing optimisation algorithms, which manufacturers from a range of different sectors are interested in using. Let’s take a closer look at an example of successful technology transfer between Inria and Normand Info (Beckman Coulter)
Illustration santé
© Jakkapan - AdobeStock


Specialising in the manufacture of scientific devices for use in biology and medicine, Beckman-Coulter is a company which develops a range of services designed to improve the quality of care that healthcare professionals are able to give to their patients. One service they offer is providing all of the information needed to make an accurate diagnosis: Beckman-Coulter has made designing, installing and using medical analysis laboratories a key area of focus for development. With the Inria Lille-Nord Europe Research Centre and its optimisation experts, the American company found a scientific partner capable of helping them meet their goals.

Optimising production: a complex mathematical problem

Various methods are used within analysis laboratories: they handle between 50,000 and 100,000 samples on a daily basis, using a dozen or so machines operated by roughly the same number of specialists. What is the best way of using these resources, while reducing processing timescales and ensuring high quality analysis - bearing in mind that laboratories operate over periods of decades? This is a difficult question to answer, given the vast array of parameters and the many different ways they can be combined.

Mathematicians from the Bonus* team have been working for around thirty years on “large-scale optimisation” methods that could be of use to Beckman-Coulter. These mathematical tools can be used to solve problems with large numbers of variables or constraints, such as those encountered in the field of “operational research” or in “machine learning”.

An expert in optimisation and head of the Dolfin team (now Bonus) until 2017, El-Ghazali Talbi explains: “These methods enable haulage and logistics companies to select the best possible warehouse locations, to optimise vehicle fleets and to reduce delivery timescales; telecommunications operators to install antennae and to carry out maintenance; and energy companies to design production and distribution networks, varying sources and responding to fluctuations in demand, etc.”

Algorithms inspired by the natural world

The researchers from Bonus place particular emphasis on “evolutionary algorithms”, a class of mathematical methods inspired by solutions found by certain colonies of animals in response to problems with large numbers of parameters (see inset). “Our research has already led to multiple collaborations with the world of industry, in the sectors we’ve mentioned: La Redoute, the Cnes*, EDF, the Onera*, Orange and T-Mobile have already sought assistance from us.”

However, designing and optimising a medical analysis laboratory presented the scientists with new challenges. “We developed a comprehensive tool which can be used to take decisions on ‘strategic’ objectives (over a decade), ‘tactical’ objectives (over a month) or ‘operational’ objectives (which impact upon daily activities). This might include, for example, which machine to buy and where to position it on-site (a strategic challenge), how to configure the machine in order to carry out analysis (a tactical decision) and in what order samples should be assigned to it for analysis (an operational question). Given that these different questions and the corresponding timescales are interdependent, there are a huge number of different possible combinations.”


Beckman-Coulter/Inria: a partnership that gets results

Illustration santé
© - AdobeStock

In order to arrive at the right solution, the researchers involved in this ambitious project (with a budget of €500,000 over three years and a PhD student, an engineer and an expert from Inria) worked in close collaboration with their client's trade experts. “Another way in which the project is original, from a scientific perspective, is in the response we came up with: during Sohrab Faramarzi’s PhD research, we combined optimisation techniques with mathematical models simulating analysis machines. Designed specifically for this project, these simulations benefited from the expertise of personnel from Beckman-Coulter with whom we were in discussions on a weekly basis. Without their support, the models developed would not have met the level of precision required by the optimisation algorithms.”


This close collaboration led to multiple patents being filed, in addition to the development of tools designed for the configuration or reconfiguration of installations and for optimising the way in which laboratories are run. “The solutions put forward can also be transposed onto other production systems and we are expecting further collaborations with the world of industry”, explains El-Ghazali Talbi.

* Bonus (Big Optimization aNd Ultra Scale computing) – is a joint Inria Lille-Nord Europe research team, with support from the University of Lille and the Lille Centre for Research into IT, Signals and Automation. The team is made up of 5 permanent researchers and around a dozen or so PhD students and postdoc researchers.

Cnes: Centre national d’études spatiales (The French National Centre for Space Studies)

Onera: Office national d’études et de recherches aérospatiales (The French National Aerospace Research Centre)

From swarm intelligence to large-scale optimisation

© Marina Foteeva - AdobeStock


Whether it’s bringing back food for the colony, identifying the position of flowers for gathering pollen or nectar or adapting in flight during long migrations: ants, bees and birds find solutions to complex problems by coordinating the work of each individual, while minimising the amount of effort required. Mathematicians study this “living intelligence” in order to devise algorithms capable of effectively solving certain optimisation problems. “Evolutionary algorithms”, developed through this approach, are used by the researchers from the Bonus team for large-scale problems.