Tumor lesions are the leading causes of death worldwide. Early detection of tumors can improve the overall survival of patients. Manual annotation of tumors is highly laborious and time-consuming. We plan to develop an end-to-end automated framework for the segmentation of tumors and their metastases from whole-body PET/CT scans using deep learning.
The current piece of code is not very organized without any documentation (Hard to read and understand). It is also not very effcient in terms of memory and speed.
My intent here is to convert this piece of code into a python package by having separate modules, classes, functions and add proper documentations (wherever necessary). I am also planning to use high performance computing (such as GPU acceleration) and parallel programming (multi-processing) to fastract some of the computations. Finally, I also plan to optimize the overall framework for 3D tumor segmentation. Additionally, I am also planning to include better visualization tools (related to the General Problem Statement), that will be helpful during visualization.
- Transforming ugly looking code to python libraries.
- Adding documentation.
- Adding proper test function.
- Optimizing the code.
- Making the code easy to read.
- Packaging.