Deep learning for automated image analysis in clinical drug development trials
Purpose and goal
The unmet need for more precise and tailored methods for decision-making in drug development is big. Antaros supports, using advanced quantitative image analysis based on e.g. magnetic resonance imaging (MRI), drug developers in clinical development phase to understand how their drug works. With the Vinnova´s grant "Start your AI journey", Antaros Medical wants to streamline and improve certain image analysis methods for studies of non-alcoholic fatty liver disease (NAFLD/NASH). The end goal is a more automated image analysis of MR images on liver.
Expected results and effects
By implementing the project, the already established image analysis methods become more efficient and less operator dependent, which leads to an even higher precision. This is key for biomarkers guiding drug development decisions. The acquired knowledge is also expected to be useful for most future studies of kidneys, heart and brain.
Planned approach and implementation
We start our AI journey by, with expertise from Uppsala University, setting up and training convolutional neural networks (CNNs) to perform these segmentations. The networks would be trained using a large number of existing segmentations and are expected to generate results that only need minor corrections during quality control.