GIGAVISION: System for Marking Tumor Regions in Gigapixel Histological Images

To this day, biopsy is the definitive diagnostic test for the detection of different types of cancer. A sample of the tissue under study is taken, processed (using fixation, inclusion, cutting, and staining techniques), and analysed by the pathologist under a microscope. Due to the current heavy workload and thanks to Whole Slide Image (WSI) technology, a common practice is the digital scanning of such samples that allows an automatic analysis of them using image processing and artificial intelligence algorithms. For the training phase of these predictive models, samples labelled or annotated by expert pathologists are required. Images scanned using WSI technology are gigapixel images that take up 1-2 GB of disk memory. That is why loading them statically into memory becomes unfeasible, thus making it difficult to access and view these images. To solve this problem, GIGAVISION has been developed, a system that allows the marking of tumour regions without computational load. This makes it easier for pathologists to mark images. Among its features are: – It is a web tool that allows the visualization of gigapixel images from anywhere in the world. It is a collaborative tool that encourages the teamwork of pathologists from different countries and thus allows for various opinions on the annotation of tumor patterns. – The tool allows you to increase the resolution of the images without this implying a loss of quality. Each time a certain zoom level is chosen, the application dynamically loads the image at that magnification level at full resolution. – The web tool includes various functionalities to be able to perform the annotation task in a comfortable and simple way for the user, either using a tablet pen or using the mouse pointer. Free dial and dot mesh marking functionality is included. – Finally, the tool communicates with a mySQL database that keeps the data up-to-date and organized. When you make an annotation and press the save annotation button, it is automatically stored in the table relating to the user who made the annotation.