It is rewarding that a project you have been involved and have given birth has been awarded by an European Prize. In this case, Tecnalia has been awarded for the development of a histological image based search engine that is able to extract and capture the visual information that is clinically relevant.
This is really useful for biomedical researchers that need to access biological tissue samples that are stored in biobanks among the world. This search allows them locate candidate tissues by means of their visual appearance as you can see in this video:
The system analyzes the histological images and extracts the subtle architectural, textural and morphological features that are inherently included on an image. It is noteworthy that each of this histological images consists of a pyramidal ‘google maps similar structure’ that describes each tissue. This image can weight more that 2GB. Biosimil extract relevant information from the image describing the discriminative patterns and thus, reducing a 2GB image into a small one dimensional vector. This vector is indexed on a large scalable database for similarity search… as you can note, euclidean distance is not the most suitable measurement of similarity among these vectors… we will discuss about this another day…
It is still lot of work to do, improvements we want to integrate into the system, larger scale pilots… but we are in the good path and eager to continue working on Biosimil 🙂