Imaging is a discipline that has a strong observer dependency, especially in the way it is conducted today. For example, someone will review the images and make an assessment on whether there is malignancy or not. Or someone will a draw Region of Interest (ROI) based on their visual perception of a boundary between two different tissues. Furthermore, the vast majority of imaging data is currently used qualitatively, not quantitatively, which makes it hard to combine it with non-imaging data for making evidence-based decisions.
Learning to See (L2S) is Preclinical IT’s effort to address the above challenges by:
1. Curating comprehensive preclinical datasets that involve imaging and their associated non-imaging data
2. Automating image analytics to remove observer dependencies
3. Extract numerical indicators of biological activity (biomarkers) from multiple flavors of data and assess their predictive and translational potential
The data is at the center of our approach to the above objectives.