ET | MT | NES | IST | POL | BMT
Motivation
Coherent fiber bundle (CFB)-based endoscopes offer a minimally invasive imaging to access deep tissue in real time, such as brain tumor. The honeycomb artifacts caused by structure of fiber cores can be removed by a deep neural networks (DNN) with an enhanced spatial resolution. Furthermore, a physics-based fiber imaging model allows training the DNN in a self-supervised manner without given medical tissue samples. However, the performance of the fiber imaging model for different fiber types and at different wavelengths is not fully studied.
In this work, the physics-based model for optical CFB imaging is to be
further investigated. Experiments and data collection using different optical fiber types (Sumita and Fujikura) and at different wavelengths (blue, green and red) needs to be implemented. Subsequently, fiber imaging modeling and DNN-based image reconstruction will be implemented.
(a) Fiber endoscopy for brain cancer diagnostics
(b) Reconstruction for different imaging wavelength/CFBs
Keywords: Optical experiment, fiber bundle imaging, deep neural network, Python, PyTorch
Tasks:
- Experiments of CFB imaging and data acquisition
- Implementation of modeling CFB imaging for different fibers/wavelengths
- Implementation of image reconstruction for different fibers/wavelengths
- Result evaluation and documentation
Contact
Tijue Wang tijue.wang@tu-dresden.de