Pca Guided Idif

This module implements PCA (Principal Component Analysis)-guided methods for generating Image-Derived Input Functions (IDIF) from 4D-PET data. IDIFs are used in PET (Positron Emission Tomography) imaging workflows for kinetic modeling and quantification. The methods provided in this module focus on data-driven approaches based on PCA, offering tools for selecting relevant voxels and refining input functions.

The module is designed for flexibility, offering tools for to apply PCA-based analyses over 4D-PET data and masks. The classes are structured to allow reuse and extension for new functionalities.

Classes

• PCAGuidedIdifBase: An abstract base class that supports data preparation, PCA operations, and IDIF computation from PCA projections.

• PCAGuidedTopVoxelsIDIF: Implements IDIF generation based on selecting the top PCA component voxels.

• PCAGuidedIdifFitterBase: A base class for deriving optimal voxel fit masks using minimization routines and customizable terms (e.g., smoothness, noise).

• PCAGuidedIdifFitter: A concrete implementation of PCAGuidedIdifFitterBase, providing default term definitions.

Dependencies

• numpy: For numerical operations and array manipulations.

• ants: For reading and handling medical image data.

• lmfit: Used for robust fitting and parameter optimization.

• sklearn: For PCA implementation.

• Other project-specific utilities:

  • ScanTimingInfo: Handles timing information from NIfTI images.

  • temporal_pca_analysis_of_image_over_mask(): Performs PCA on dynamic image data over a given mask.

  • extract_roi_voxel_tacs_from_image_using_mask(): Extracts voxel TACS (Time-Activity Curves) from the input image.

Notes

• The classes contain abstract (NotImplementedError) methods, intended to be overridden by derived classes as per the specific use case. These serve as stubs for users wishing to extend the module.

• While the module supports large dynamic datasets, care should be taken with memory usage when handling large voxel masks or PCA decomposition.

Todo

• Write a tutorial to to use the class to generate an IDIF and use the IF to perform kinetic analysis

Classes

PCAGuidedIdifBase	A base class for PCA-guided Image-Derived Input Function (IDIF) generation.
PCAGuidedTopVoxelsIDIF	Class for calculating a PCA-guided IDIF by averaging over the top voxels of a selected principal component (PC).
PCAGuidedIdifFitterBase	Base class for calculating the PCA-guided IDIF by fitting to find the best voxels
PCAGuidedIdifFitter	Class to calculate the PCA-guided Image-Derived Input Function (IDIF) by fitting to find the best voxels.