Artificially intelligent MICE
|Coordinator||Nonpi Medical AB|
|Funding from Vinnova||SEK 300 000|
|Project duration||May 2018 - October 2018|
Purpose and goal
The project aims to develop and promote an automated, user-friendly feature for using artificial intelligence for advanced medical image analysis applications. This is to provide an intuitive tool to simplify advanced tasks for anyone working with medical image analysis. In step 1, the project has completed a prototype for application of machine learning directly in the MICE Toolkit and implemented basic network layers and machine learning functionality to design bespoke neural networks.
Expected results and effects
In step 1, the project has completed a prototype for the application of neural networks directly in the MICE Toolkit and implemented basic network layers and machine learning functionality for designing machine learning networks. The ambition is that this will be further developed into a fully automated solution for application as well as design and training of neural networks in an intuitive user-friendly drag´n´drop environment combined with MICE Toolkit´s powerful medical image analysis tools. However, it remains to be validated as a package solution in the MICE Toolkit.
Planned approach and implementation
The project´s approach is based on initially solving technical barriers such as development and implementation of functionality to apply all proven forms of machine learning (including deep learning and other AI models). The project has subsequently focused on the ability to intuitively and easily design own networks directly in the MICE Toolkit´s regular environment without the need for programming skills. After that, the project has focused on the ability to train own networks, which today requires interpreting to python code but will be implemented in the same way for MICE Toolkit