I want to know which deep learning techniques (supervised or unsupervised or merged) are popular in medical image processing and what can be the future challenges in this context.
Your question is a bit general but simply they are popular since they have the best performance over most of the conventional approaches in medical image processing.
Thanks on worthy insights Saeid Soheily-Khah Mubashir Ahmad Amirreza Mahbod Mohamed Abdel-Nasser
Supervised or unsupervised technique, which one is more successful? Do you consider time and space if getting better results on medical image analysis ?
Many deep learning techniques are currently being used for image processing of medical images. Some of the examples are
1. U-net based architectures are currently being used for image segmentation as well as image reconstruction. Sometimes the models have been used even for image super-resolution also e.g. PET imaging.
2. For classification of cancerous vs non-cancerous cells, various architectures have been used. Recently proposed Capsnet architecture has been quite popular for MNIST dataset and people are trying to use it for other data-sets.
3. CNN architectures are also being used for regression problems for getting the cellularity index of images.
4. Various problems related to fusion of medical images, enhancement and super-resolution of medical images has been done using simple CNN based architectures.