Artificial Intelligence in Medical Imaging
Prerequisites
- Some knowledge on inverse problems
- Deep learning for image reconstruction
- Basic knowledge of PyTorch or TensorFlow
Description
Deep learning for medical imaging is an active research area in imaging science, which studies the reconstruction of images from data acquired with a medical device, such as an MRI or CT scan. In some cases, it also analyzes the decision-making and diagnosis based on the reconstructed images. Medical imaging tasks are mathematically formulated as inverse problems, meaning that the mathematical machinery needed for solving them is the same as in inverse problems, i.e., data-driven and model-based regularization, combining classical apriori information obtained via first principles or learned from data. Learning from data is particularly hard in medical imaging, mainly due to the lack of publicly available annotated medical data. In addition, inverse problems in medical imaging tend to be severely ill-posed, for example, limited-angle tomography, making them particularly hard to solve using solely deep learning.
References
- Inverse problems in biomedical imaging: modeling and methods of solution (https://iopscience.iop.org/article/10.1088/1361-6420/ac28ec/meta)
- Template-Based Image Reconstruction from Sparse Tomographic Data (https://link.springer.com/article/10.1007/s00245-019-09573-2)
- An Introduction to X-ray tomography and RadonTransforms (https://sites.tufts.edu/tquinto/files/2021/01/sc-article.pdf)