CONTEXT
Mediterranée Infection is a research laboratory in virology and infectious diseases. It is the world leader on the study of virus and bacterias. Méditerranée Infection receives hundreds of thousands of analysis performed on patients in a large number of hospitals in the PACA region in France. These analysis require a manual processing involving a visual interpretation, which is costly, slow and subject to human error.
CHALLENCE
Their challenge was to automatize at least 95% of the images analysis through an automatic image processing system. But also, the objective was to reduce costs and processing time. Finally, the client wanted to make the laboratory staff focus on border-line cases and more valuable tasks.
RESULT
- The antibiogram images result from placing antibiotics disks on a bacteria strain culture in a Petri box – A lot of images present imperfections (spill, blurry areas) – The antibiotics used are recognized by OCR
- Machine learning system that has recently been patented – 98% of the phenotypes recognized – Continuous improvement by reinjecting the disambiguation input from the laboratory staff
