Action List

CCTV-ML

Completed Actions

• convert notebook_07 to cctv_nb_07.py Jira task
• added saving the trained/load model feature.
• save predictions.csv and shape files based on the data i.e (pred, train test).
• document every function with Docstrings
• generate Sphinx Document for the cctv_nb_07.py
• Added "How to Run" section for notebook_07 Link.
• Added all mermaid diagrams from notebook_1 to notebook_7.
  • you can find the mermaids under Vishwanatha/CCTV link
• keep metro utility code.

Pending Actions for notebook_07

• Future Scope:
  • SQLite integration
    • csv's are converted to sqlite db link
  • LGBM model with CUDA
    • Explored LGBM with CUDA core in Google Colab and ran it in notebook_07 link.
    • implementation of LGBM with CUDA core for notebook_07 can be found in DailyLog
  • Run any utility
    • currently able to run 2 utilities i.e dnv and metro
  • Integrate weights and biases logging into notebook_07.

RKSP

Completed Actions

• Created directory tree structure.
  • all files are under vishwanatha/documents/RKSP/itk
• Used 3D Slicer for initial visualization. link
• Used Blender for STL creation to feed Unreal and clean the mesh.
• Imported into Unreal Engine.
• Explored VTK C++ DLL.
• Tried converting into web-based using WebAssembly.
• Switched to vtk.js.
• Successfully ran all the node-based vtk.js examples.
• Switched to VolView.
• Successfully debugged VolView which uses Vue.js.
• Documented in development practices.
• Tried Horus.
  • Deployed on macOS.
• Documented tools used for RKSP link
• Visualize bones using ITK and VTK (2dtobones.py).
• Implement voxel erasure using different shapes:
  • Cube form (cube.py)
  • Spherical form (drill.py)
  • Cone form (drillcone.py)
• Enhance image rendering with ITK using one axis (itkdemo.py).
• Convert subdirectories of DICOM images into a single .nrrd file (nrrrdto3d.py).
• Load one plane to 3D (oneplaneto3d.py).
• Generate 3D volume using ITK (dicomtovtk.py).
• Implement thresholding techniques to visualize bones more clearly (thresolding.py).
• Render 3D volume using STL (stlwithvtl.py).

Pending Actions for RKSP

• Explored and tried MONAI SDK, no output visualized.

• Convert 2D DICOM to 3D using three planes (2dto3d.py).

• Attempt using converted DICOM files (2dto3dwithdicom.py) - failed, needs improvement.

  • Address and improve failed outputs:
    • combined_axial.dcm
    • combined_coronal.dcm
    • combined_sagittal.dcm

• Improve Streamlit viewer (view.py).

  • needs improvement unable to parse the loaded files

LOE-COE Action List

Completed Actions

• Did initial review. documented link
• Restored deven-loe-coe-fork repository and created a virtual environment with Python 3.8.
• Tried running streamlit app.py additional info
• Tried running the debugger.
  • Added "How to Run" section for Loe-coe-app Link
• Tried running different utilities.
  • succesfully ran cov utility

Pending Actions

• Inputs are stored in sqlite.db; NOTE : should change the name.
• Implement transfer learning.
• Streamlit should support model training.
• Add functionality for saving the model.
• Develop logic to use pretrained models.
• differentiate LOE and COE in streamlit-UI