Medical image processing can help to find and understand irregularities in the human body. It can detect or even predict diseases. As a data source, Magnetic Resonance Imaging (MRI) or Computed Tomography (CT) is often used. The current state-of-the-art in medical image processing and analysis is in machine learning and, more specifically, in deep learning (DL) and the use of deep neural networks. DL methods can reduce the time to high-quality medical diagnosis and thus improve healthcare in general. Nevertheless before any effective DL algorithm is trained, a large amount of data and computing power is needed.
In this case, doctors (radiologists) from the University Hospital Ostrava provided the necessary data, experts from the National Supercomputing Center IT4Innovations “delivered” computing power and knowledge how to easily collect the data and how to train the neural network (GPU cluster Karolina, NVIDIA Clara Train SDK and 3D Slicer software were used). The available Guide developed within the EuroCC project offers a description of the assumptions and the process of “deep learning” on the example of automatic tissue segmentation.
See more here.