jilei-hao / avrspt-dataservice

AVRSPT-DataService

Prerequisite

Install

https://fastapi.tiangolo.com/#installation

Run

• Set directory to serve in config.ini
• cd src
• `uvicorn main:app --reload --host x.x.x.x --port xxxx

Reference

https://fastapi.tiangolo.com/#run-it

avrspt-dataservice's Issues

Model Data Optimization

Description

The mesh directly generated by MarchingCubes filter from the 4DCTA segmentation image has a very large file size, due to the dense distribution of vertices and edges. It is not optimal for web application which we want to reduce the amount of data transfer between client and server. For mobile app, large data size also occupies too much memory, which may lead to out of memory issue for the entire application. The structure of the model is relatively simple, not containing detailed features that requires large amount of vertices to represent. Therefore we want to explore a way to decimate mesh by removing unnecessary vertices and edges, and then smooth the mesh to make it looks as similar to the undecimated mesh as possible.

Requirement

It's an open-ended task. The optimization can be done on different levels: Downsampling of the image, Marching Cubes, Decimation and Smoothing. Start by understanding the background and issue, and get familiar with the tools that medical image scientists use daily. Then some scripts and pipelines can be written to experiment with different algorithms and configurations.

• Download ITK-SNAP from http://www.itksnap.org/pmwiki/pmwiki.php
• Clone the AVRSPT-Converter project and study the source code
• Create pipelines to explore different algorithms and configurations
• Final Product, A data pipeline and one or many set of parameters to generate decimated but smooth meshes.

Resource

Project

• Data Service: https://github.com/jilei-hao/AVRSPT-Converter
• Box: https://upenn.box.com/s/bx5cqs6eovc7jex9zxuplpvqecpqkbi8

Iso-Contour Algorithms

• vtkDiscreteMarchingCubes example: https://gitlab.kitware.com/vtk/vtk/-/blob/v9.2.0/Examples/Medical/Cxx/GenerateModelsFromLabels.cxx
• Brand new iso-contour algorithm: https://discourse.vtk.org/t/new-isocontouring-filter-for-discrete-labled-volumetric-data-vtksurfacenets3d/10614
• Existing IsoContour Algorithm: https://github.com/jilei-hao/cmrep/blob/dev/czi-128/src/util/RealImageToMesh.cxx#L177-L203

Mesh Smoothing

• Laplacian Smoothing: https://vtk.org/doc/nightly/html/classvtkSmoothPolyDataFilter.html
• Taubin Smoothing: https://vtk.org/doc/nightly/html/classvtkWindowedSincPolyDataFilter.html

Rohan Project Kick-off