Self-learning decision support tool for fast and accurate cancer evaluation in digital pathology
| Reference number | |
| Coordinator | ContextVision AB - R & D |
| Funding from Vinnova | SEK 3 729 170 |
| Project duration | October 2015 - September 2018 |
| Status | Completed |
| Venture | Eurostars |
Purpose and goal
The purpose of the project was to, based on market needs, unique technology and knowledge in image analysis, develop a number of decision support tools for pathologists, so that diagnosis and prognosis of cancer could be facilitated. The objective was to develop a toolbox of decision support algorithms that can simplify and enhance the pathologist´s work in assessing tissue samples of the most common cancers, prostate, breast, colon and skin. The target is met as we now have begun the pre-launch of the first product - INIFY Prostate.
Expected results and effects
The result of the project is that a new business unit has been formed within Contextvision for commercialization of the development work, about ten new employees have been recruited and the first product is now being introduced in Europe. We have proven our unique skills in deep learning and image analysis through participation in two breast cancer Challenges, Camyleon17 and TUPAC, where we ranked second in the world on both. The project has so far produced 15 scientific publications, presentations and posters.
Planned approach and implementation
The collaboration with our partner eHealth unit within the HES-SO University of Sierre, Switzerland has worked exceptionally well, as shown in the outcome. We have had regular project meetings monthly, whereof two times a year physically on-site. SCRUM methodology has been used to control the project with daily reconciliation. A patent application for a method of using objective data instead of subjective evaluations as a basis for training the DL networks has been submitted and published.