liveBook · Manning

Coordinate system transformation

As explained in Deep Learning with PyTorch, Second Edition

The mapping information used to convert nodule coordinates into array indices for the CT volume.

Coordinate system transformation maps nodule positions from patient coordinates to the index, row, and column coordinates used by a CT volume’s 3D array tensor. section 12.1

The transformation connects annotation data to model-ready samples:

  • Combine each annotated nodule coordinate with its scan’s transformation information. section 12.1
  • Convert the position into I, R, and C voxel coordinates. section 12.1
  • Use those coordinates to crop a small 3D CT slice around the nodule. section 12.1

Nodule-centered CT sample

A nodule annotation and its series UID identify the relevant scan; transforming the annotation yields the voxel center used to crop the model input. section 12.1

How Deep Learning with PyTorch, Second Edition covers it

How deep it goes
Defines the coordinate conversion and its cropping role, while deferring detailed mechanics to a later treatment.
Examples
Uses annotated lung nodules, transformation matrices, I/R/C coordinates, and small 3D CT crops.
What it assumes
Assumes raw CT arrays and patient-coordinate annotations while introducing the need for coordinate conversion.
What you can do after
Apply it to map annotated nodule centers into CT voxels and crop model-ready 3D samples.
How it connects
Ties patient coordinates to CT volumes, voxel coordinates, 3D array tensors, annotations, and PyTorch Dataset samples.
Cover of Deep Learning with PyTorch, Second Edition
liveBookmanning.com
Read Deep Learning with PyTorch, Second Edition on liveBook Get the book at manning.com

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Deep Learning with PyTorch, Second Edition. The section links point to the parts of the book it draws on.