SPECT

Data used in the tutorials can be downloaded here.

Introduction

The use cases below are basic introductions for how to use the library, and demonstrate complete reconstruction pipelines for SIMIND (Monte Carlo) and DICOM (clinical) phantom data using OSEM. They are a good starting point for learning the library.

SIMIND Introduction

SIMIND Intro tutorial

DICOM Introduction

DICOM Intro tutorial

Available Algorithms

The tutorials below demonstrates some of the available reconstruction algorithms of the library, and borrows code from the introduction tutorials above.

Various Algorithms

SIMIND Intro tutorial

Additional Use Cases

The tutorials below demonstrate some additional use cases for SPECT reconstruction, such as reconstructing and stitching multi-bed positions, reconstructing multiple photopeaks, and estimating uncertainty in reconstructed images

Multi Bed Positions

multi bed

Multi Photopeak Reconstruction

dual peak

Uncertainty Estimation

uncertainty spect

Hybrid Monte Carlo Reconstruction (Simulated data)

uncertainty spect

Hybrid Monte Carlo Reconstruction (Real data)

uncertainty spect 2

Unique Systems

Some SPECT systems have unique properties that require special handling. The tutorials below demonstrate how to handle these systems.

StarGuide

multi bed

Advanced PSF Models

The tutorials below demonstrate how to use PSF models obtained via the SPECTPSFToolbox for Ac-225 reconstruction, though they can be adapted for other isotopes.

Ac225 SIMIND

simind

Ac225 DICOM

Ac225 DICOM

Cardiac Functionality

The tutorials below demonstrate reconstruction and reorientation of cardiac SPECT data.

Cardiac Reconstruction and Reorientation

cardiac

Useful Code Snippets

The following contains some useful code snippets you may require when working with PyTomography.

Select Code Snippet... Open multiple simind regions / energy windows at once Resample projection data before reconstruction

import os
from pytomography.io.SPECT import simind

path = '/disk1/pytomography_tutorial_data/simind_tutorial/'

# SIMIND simulation files for simulation of only a liver region
photopeak_path_liver = os.path.join(path, 'multi_projections', 'liver', 'photopeak.h00')
upperscatter_path_liver = os.path.join(path, 'multi_projections', 'liver', 'lowerscatter.h00')
lowerscatter_path_liver = os.path.join(path, 'multi_projections', 'liver', 'upperscatter.h00')

# SIMIND simulation files for simulation of all anatomical regions
organs = ['bkg', 'liver', 'l_lung', 'r_lung', 'l_kidney', 'r_kidney','salivary', 'bladder']
headerfiles = [os.path.join(path, 'multi_projections', organ, 'photopeak.h00') for organ in organs]
headerfiles_lower = [os.path.join(path, 'multi_projections', organ, 'lowerscatter.h00') for organ in organs]
headerfiles_upper = [os.path.join(path, 'multi_projections', organ, 'upperscatter.h00') for organ in organs]

# 1. To load multiple energy windows for one region, we provide the paths as a list
photopeak_liver = simind.get_projections([photopeak_path_liver, upperscatter_path_liver, lowerscatter_path_liver])
print(photopeak_liver.shape)

# 2. To load a single energy window and combine multiple regions, we provide the list as [[<list>]]. This will scale each SIMIND set of projections by the corresponding activities (combining projections together):
activities = [2500, 450, 7, 7, 100, 100, 20, 90] # MBq
photopeak_allregions = simind.get_projections([headerfiles], weights=activities)
print(photopeak_allregions.shape)

# 3. To load multiple energy windows for multiple regions, we provide the list as [[<list>], [<list>], [<list>]]:
activities = [2500, 450, 7, 7, 100, 100, 20, 90] # MBq
projections_allregions = simind.get_projections([headerfiles, headerfiles_lower, headerfiles_upper], weights=activities)
print(projections_allregions.shape)

from pytomography.io.SPECT import simind
from pytomography.io.SPECT.shared import subsample_metadata, subsample_amap, subsample_projections
import os

# CHANGE THIS TO WHERE YOU DOWNLOADED THE TUTORIAL DATA
PATH = '/disk1/pytomography_tutorial_data'

data_path = os.path.join(PATH, 'simind_tutorial', 'lu177_SYME_jaszak')
photopeak_path = os.path.join(data_path,'tot_w4.h00')
attenuation_path = os.path.join(data_path, 'amap.h00')

object_meta, proj_meta = simind.get_metadata(photopeak_path)
photopeak = simind.get_projections(photopeak_path)
amap = simind.get_attenuation_map(attenuation_path)

# Projection data can be subsampled in pixels (via average pooling) and angles via
photopeak_subsampled = subsample_projections(
    projections = photopeak,
    N_pixel = 2, # from 128x128 to 64x64
    N_angle = 2, # every 2nd angle
    N_angle_start = 1, # angle to start at
)

# If you subsample the photopeak, you also have to adjust the metadata:
object_meta_subsampled, proj_meta_subsampled = subsample_metadata(
    object_meta = object_meta,
    proj_meta = proj_meta,
    N_pixel = 2,
    N_angle = 2,
    N_angle_start = 1,
)

# And the attenuation map (if you use attenuation correction):
attenuation_map = subsample_amap(
    amap = amap,
    N_pixel = 2,
)

# You can then proceed to use the updated metadata/projections to build the system matrix and likelihood as normal