Skip to main content

radiuma_suv: PET to SUVbw Conversion – Python Script and Library

radiuma_suv is the proposed library name for the PET-to-SUVbw conversion engine in src/radiuma_suv/core.py. The code converts PET pixels to body-weight-normalized standardized uptake values (SUVbw) using DICOM acquisition and radiopharmaceutical metadata. It supports conventional single-frame PET, Enhanced multi-frame PET, NumPy arrays with matching DICOM metadata, and folders containing one or several PET series. In-memory results are floating-point arrays; DICOM results are newly written, calibrated derived images.

Code Overview: radiuma_suv is a Python-based PET-to-SUVbw conversion tool that converts PET image pixel values into body-weight-normalized Standardized Uptake Values (SUVbw) using DICOM metadata. It supports conventional single-frame PET, Enhanced multi-frame PET, NumPy arrays, and DICOM directories containing multiple PET series.

The code provides three main interfaces:

  • radiuma_suv(input_array, input_metadata, parameter_estimation="OFF"): Converts a 2D/3D PET NumPy array into a floating-point SUVbw array and returns the corresponding metadata dictionary.
  • radiuma_suv_from_input(image_input, parameter_estimation="OFF", output_path=None): Automatically detects supported PET inputs and produces SUVbw outputs in the corresponding format (NumPy array or derived DICOM file/directory).
  • RadiumaSUVLoader(parameter_estimation="OFF", pet_metadata=None, output_path=None): Provides an object-oriented interface for PET-to-SUVbw conversion through the .convert(image_input) method.

Parameter Estimation: "OFF" uses existing DICOM metadata without filling missing fields, while "ON" fills eligible missing metadata using predefined configuration values. Original DICOM files are not overwritten.

Scientific caution: parameter_estimation="ON" can fill missing metadata with notebook-configured values, including patient weight, radionuclide activity, timing, and other acquisition details. Those values must not be assumed correct for another patient. Prefer "OFF" with complete verified source DICOM metadata. No clinical validation is established by the notebook.

🔍 Table of Contents

🎥 YouTube Tutorials

Coming soon

🧩 IBSI Standardization

We have three versions of the software/code for IBSI Compliance-SUV Calibrations.

  • Version 1: before the start of IBSI-SUV
  • Version 2: Calibrated with the first standardization attempt and with the first 17 DROs.
  • Version 3: Calibrated with the second standardization attempt and with all 58 DROs. IBSI calibration datasets are available for download

Report of Calibrations Calibration 1 (Just the first 17 DROs) • Use Code Version 2 and the first 17 DROs. o Result 1, called “Version 2_Callibration 1_radiuma_suv_PySERA.csv”  Note: We already submitted this result, but we include it in this collection for your easy access.

Calibration 2 (All 58 DROs) • Use Code Version 3 and all 58 DROs. o Result 2, called “Version 3_Callibration 2_ radiuma_suv_PySERA.csv”

• Use Code Version 2 and all 58 DROs. o Result 3, called “Version 2_Callibration 2_ radiuma_suv_PySERA.csv”

• Use Code Version 1 and all 58 DROs. o Result 4, called “Version 1_Callibration 2_ radiuma_suv_PySERA.csv”

Calibration 3 ( All 16 Real Patients: Final Validation Phase) • Use Code Version 3 and all 16 real patients (Validation phase). o Result 5, called “Version 3_Callibration 3_ radiuma_suv_PySERA.csv”

• Use Code Version 2 and all 16 real patients (Validation phase). o Result 6, called “Version 2_Callibration 3_ radiuma_suv_PySERA.csv”

• Use Code Version 1 and all 16 real patients (Validation phase). o Result 7, called “Version 1_Callibration 3_ radiuma_suv_PySERA.csv”

✨ Key Features

  • PET-to-SUVbw conversion using PET image pixels and matching DICOM metadata.
  • Conventional PET DICOM: single-frame PET images and conventional PET slice series.
  • Enhanced PET DICOM: multi-frame images with functional group metadata.
  • NumPy arrays: convert in memory using a corresponding DICOM metadata dictionary.
  • Hierarchical PET directories: recursively identify PET DICOM groups and process independent series separately.
  • Type-preserving API: NumPy input returns a NumPy array; DICOM input creates derived DICOM outputs.
  • Optional metadata completion: parameter_estimation="ON" fills selected missing values using notebook settings.
  • Derived DICOM output: new DICOM identifiers, calibrated signed 16-bit pixels, and output validation against computed SUVbw values.
  • Source protection: output paths are checked to avoid overwriting original DICOM files.
  • Object-oriented convenience wrapper: RadiumaSUVLoader.

These are implementation capabilities, not independent quantitative or clinical validation claims.

📦 Library Installation via pip

After publishing the intended package under the distribution name radiuma_suv:

python -m pip install radiuma_suv

For local development in the directory containing pyproject.toml:

python -m pip install -e .

📚 Library Usage

💻 Script Execution

The radiuma_suv code supports two execution modes: Standalone Mode and Embedded Mode.

1. Standalone Mode

Run the Python script directly to convert PET DICOM images into SUVbw.

Step 1: Choose the input path and options at the CLI (do not edit the installed library):

radiuma_suv PATH_TO_PET_DICOM_FILE_OR_FOLDER --parameter-estimation OFF

Step 2: Alternatively execute using Python:

python -m radiuma_suv PATH_TO_PET_DICOM_FILE_OR_FOLDER --parameter-estimation OFF

Output: Creates calibrated SUVbw DICOM files in a new _processed location without overwriting original images.

2. Embedded Mode

Import the conversion functions into another Python application, pipeline, or GUI.

Option A: DICOM file or directory

from radiuma_suv import radiuma_suv_from_input

output, metadata = radiuma_suv_from_input(
    image_input="PET_Input",
    parameter_estimation="OFF",
    output_path="SUV_Output"
)

Option B: NumPy array with DICOM metadata

from radiuma_suv import radiuma_suv

suv_array, metadata = radiuma_suv(
    input_array=pet_array,
    input_metadata=pet_metadata,
    parameter_estimation="OFF"
)

Option C: Object-oriented integration

from radiuma_suv import RadiumaSUVLoader

converter = RadiumaSUVLoader(parameter_estimation="OFF")
output, metadata = converter.convert("PET_Input")

Note: DICOM inputs automatically provide metadata through their headers. NumPy inputs require a corresponding DICOM metadata dictionary. parameter_estimation controls metadata replacement:

Option Behavior
"OFF" Does not fill missing parameters from the configured notebook defaults.
"ON" Fills eligible missing metadata fields using configured defaults; existing values are not generally replaced.

The notebook has configurable values for patient weight, age, height, sex; administered dose, half-life and radiopharmaceutical timing; acquisition and series times; PET units; SUV type; decay correction; rescale slope and intercept; image geometry; and other fields. These settings are not additional keyword arguments of radiuma_suv_from_input.

The source implements unit conversion paths involving BQML, GML, CM2ML, and CNTS. Successful processing depends on the necessary metadata and calibration factors. The BQML converter handles DecayCorrection values ADMIN, START, and NONE with different reference-time logic. The source also includes body-composition helper functions used for certain GML SUV normalization variants.

📥 Installation

Dependencies imported by the notebook are numpy, pydicom, and SimpleITK; other referenced Python modules are standard-library modules. In the original notebook the installation line is %pip install numpy pydicom SimpleITK; for a published library, declare these in pyproject.toml instead.

🌐 GitHub Installation

https://git.talkfluent.ir/admin/radiuma_suv/src/branch/main

git clone https://github.com/YOUR_USERNAME/radiuma_suv.git
cd radiuma_suv
python -m pip install -e .

macOS/Linux Installation

python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install -e .

Windows Setup

In PowerShell inside the project directory:

python -m venv .venv
.venv\Scripts\Activate.ps1
python -m pip install --upgrade pip
python -m pip install -e .

💻 Python Script - Command Line Interface (CLI)

The notebook contains an example if __name__ == "__main__": block but does not implement an argument-parsing CLI or a radiuma_suv terminal command. Create a short Python script when you want command-line execution:

# convert_pet.py
from radiuma_suv import radiuma_suv_from_input
path, headers = radiuma_suv_from_input("pet_scan.dcm", parameter_estimation="OFF")
print(path)
python convert_pet.py

📦 Library Installation via pip

python -m pip install radiuma_suv

This will install the intended project only after it is actually published under that name on PyPI.

📂 Data Structure Requirements

The notebook processes PET DICOM input without masks. Examples of appropriate layouts:

Single PET image:

PET/
└── scan.dcm

Conventional PET image series:

PET_Series/
├── slice_001.dcm
├── slice_002.dcm
└── slice_003.dcm

Hierarchical directory:

Patients/
├── patient_001/
│   └── PET/
│       ├── slice_001.dcm
│       └── slice_002.dcm
└── patient_002/
    └── PET/
        └── enhanced_pet.dcm

The source searches directories recursively, reads DICOM headers, selects Modality=PT, and groups series by their SeriesInstanceUID and folder. A DICOM series spanning different source folders is rejected by the writer. The corresponding output folders preserve relative structure with _processed naming.

Output must not be inside the source directory. Existing destinations are not overwritten. Non-DICOM files may be skipped during directory discovery, but .nii, .nii.gz, .nrrd, and .nhdr are not valid radiuma_suv PET inputs. For NumPy, pass an in-memory ndarray and DICOM metadata; a .npy path is not accepted by radiuma_suv_from_input.

📋 radiuma_suv Parameters Reference

Public function parameters

Parameter Type Default Description
input_array numpy.ndarray Required PET array passed to radiuma_suv.
input_metadata dict Required Corresponding PET DICOM metadata dictionary.
image_input numpy.ndarray, string, path-like Required NumPy array, PET DICOM file, or PET DICOM directory.
parameter_estimation str "OFF" "ON" or "OFF" only.
output_path string / path-like / None None Destination for DICOM outputs; not valid with NumPy input.
pet_metadata dict or None None RadiumaSUVLoader metadata for a NumPy input.

Selected configured DICOM fields

These are example literal settings present in the notebook, not a universal dataset or recommended patient parameters.

Metadata field Notebook setting Meaning
PatientWeight "75.5" Patient weight, kg
PatientAge "045Y" Age in DICOM AS format
PatientSize "1.75" Height, m
PatientSex "M" DICOM sex code
RadionuclideTotalDose "370000000" Radionuclide activity; representation-specific unit handling must be verified
RadionuclideHalfLife "6586.2" Half-life, s
RadiopharmaceuticalStartDateTime "20261010093015" Radiopharmaceutical administration start
ActualFrameDuration "180000" Frame duration, ms
FrameReferenceTime "90000" Frame reference time, ms
AcquisitionDate "20261010" DICOM DA
AcquisitionTime "100000" DICOM TM
Units "BQML" PET units
DecayCorrection "START" Decay reference
RescaleSlope "1.0" Pixel calibration slope
RescaleIntercept "0.0" Pixel calibration intercept
Rows, Columns 256, 256 Example image size
PixelSpacing [4.0, 4.0] Row/column spacing, mm
SliceThickness "3.0" Slice thickness, mm
ImagePositionPatient [0.0, 0.0, 0.0] Image coordinates, mm
ImageOrientationPatient [1, 0, 0, 0, 1, 0] Orientation cosines

The source defines many additional correction, geometry, timing, and vendor-related fields. Consult the notebook's User Controlled Parameters section for the complete set. These values should not be inferred or substituted without acquisition-specific evidence.

📊 Parameter Compatibility

Feature / argument NumPy + metadata Single PET DICOM DICOM directory
parameter_estimation ✓ ✓ ✓
output_path Not allowed ✓ ✓
Floating-point array return ✓ No No
Derived DICOM output No ✓ ✓
Independent series grouping No No ✓

📚 API Reference

radiuma_suv(input_array, input_metadata, parameter_estimation="OFF")

Converts an in-memory PET NumPy array to SUVbw. Returns (suv_array, output_metadata).

radiuma_suv(input_array, input_metadata, parameter_estimation="ON")

Converts an in-memory PET NumPy array to SUVbw, it replace missed parameters in input image DICOM header with default vlues defined in code. Returns (suv_array, output_metadata).

radiuma_suv_from_input(image_input, parameter_estimation="OFF", output_path=None)

Detects input representation and returns a matching result: a NumPy array or a new DICOM output file/folder path, plus corresponding metadata. DICOM source objects are not deliberately overwritten.

radiuma_suv_from_input(image_input, parameter_estimation="ON", output_path=None)

Detects input representation and returns a matching result: a new DICOM output file/folder path, it replace missed parameters in input image DICOM header with default vlues defined in code. DICOM source objects are not deliberately overwritten.

from radiuma_suv import RadiumaSUVLoader
converter = RadiumaSUVLoader(parameter_estimation="OFF", output_path="pet_output.dcm")
output, metadata = converter.convert("pet_input.dcm")

📊 Output Structure

Input First return value Second return value
NumPy array Floating-point SUVbw NumPy array Metadata dictionary
Single PET DICOM file New DICOM file path Output header dictionary
PET DICOM folder, one series Output directory path List of header dictionaries
Folder, multiple series Output root directory path Nested header lists

DICOM representation: The source encodes SUVbw into calibrated signed 16-bit pixels. For conventional PET, the output RescaleSlope is needed to recover actual SUVbw. For Enhanced PET, use its real-world value mapping. The writer checks the saved file's pixel shape and reconstruction against calculated values within quantization tolerance. This is not an independent reference SUV validation.

📁 Supported File Formats

Format Supported Notes
Conventional PET DICOM .dcm Yes Single frame
Enhanced PET DICOM .dcm Yes Multi-frame Enhanced PET SOP class
Extensionless DICOM in a folder Detected if readable Must identify as PET (Modality=PT)
PET DICOM directory Yes Recursive series grouping
.npy path No Load with np.load() and pass array

🎯 Library Examples

Example 1: Single DICOM

from radiuma_suv import radiuma_suv_from_input
result, meta = radiuma_suv_from_input("pet.dcm", parameter_estimation="OFF")
print(result)
from radiuma_suv import radiuma_suv_from_input
result, meta = radiuma_suv_from_input("pet.dcm", parameter_estimation="ON")
print(result)

Example 2: Multiple PET folders

from radiuma_suv import radiuma_suv_from_input
out, headers = radiuma_suv_from_input(
    "PET_Input", parameter_estimation="OFF", output_path="SUV_Output"
)
from radiuma_suv import radiuma_suv_from_input
out, headers = radiuma_suv_from_input(
    "PET_Input", parameter_estimation="ON", output_path="SUV_Output"
)

Example 3: NumPy data

import numpy as np
from radiuma_suv import radiuma_suv
pixels = np.load("pet_pixels.npy")
# Load the genuine DICOM metadata for the same scan.
suv, meta = radiuma_suv(pixels, matching_pet_dicom_metadata, "OFF")
import numpy as np
from radiuma_suv import radiuma_suv
pixels = np.load("pet_pixels.npy")
# Load the genuine DICOM metadata for the same scan.
suv, meta = radiuma_suv(pixels, matching_pet_dicom_metadata, "ON")

⚡ Performance Tips

  1. Use the folder interface when processing multiple PET DICOM series.
  2. Keep output outside the input directory and ensure the output path does not exist.
  3. Avoid making unnecessary copies of large arrays when working with NumPy.
  4. Use real acquisition-specific metadata for correct dose and decay calculations.
  5. Validate numerical values after conversion and DICOM quantization.
  6. The notebook contains no benchmarking results, so no speed claims are made here.

❓ Troubleshooting

Message / problem Recommended check
ModuleNotFoundError: No module named 'radiuma_suv' Confirm the package was created, installed, and imported from the active Python environment.
parameter_estimation must be "ON" or "OFF" Use the exact supported values.
Input must be PET DICOM (Modality=PT) Confirm the file really contains PET DICOM.
Non-Enhanced multi-frame PET is unsupported Input format is outside the implemented multi-frame path.
No PET DICOM series found Verify DICOM readability, modality, and folder contents.
Output directory must not be nested within input folder Choose an independent output directory.
Refusing to overwrite existing output Choose another destination.
Missing activity, weight, half-life, or timing Review source metadata; do not blindly apply configured defaults.
Unexpected SUVbw values Recheck units, calibration, corrected dose, decay time, and actual acquisition data.

❓ Get Help

  • Documentation: Refer to this README for installation, configuration, supported formats, and usage examples.
  • Execution: See the Script Execution section for standalone and embedded usage.
  • Troubleshooting: Check the Troubleshooting section for common errors and solutions.
  • Issues and Bug Reports: GitHub Issues.
  • Technical Support: Contact the development team using the details in the Contact or Maintenance section.

🕒 Version History

v1.0.1 
- Conventional PET DICOM to SUVbw
- Enhanced multi-frame PET to SUVbw
- NumPy input and metadata output
- Recursive DICOM series processing

📬 Contact

radiuma_suv is available free of charge. For access, questions, or feedback:

Dr. Mohammad R. Salmanpour (Team Lead)
📧msalman@bccrc.ca | m.salmanpoor66@gmail.com, m.salmanpour@ubc.ca


🛠️ Maintenance

For technical support and maintenance inquiries, please contact:

Dr. Mohammad R. Salmanpour (Team Lead)
msalman@bccrc.ca – m.salmanpoor66@gmail.com – m.salmanpour@ubc.ca

Ramin Afrah (Maintainer)
afrah.ramin@gmail.com

👥 Authors

  • Dr. Mohammad R. Salmanpour (Team Lead, Fund Provider, Evaluator, Medical Imaging Expert, Backend Development, Code Refactoring, Debugging, Library Management, IBSI Standardization, and Activation of the radiuma_suv library) – msalman@bccrc.ca, m.salmanpoor66@gmail.com, m.salmanpour@ubc.ca
  • Ramin Afrah (Development, Code Refactoring, Debugging, and Library Management) - afrah.ramin@gmail.com
  • Dr. Mehrdad Oveisi (Evaluator, Software Engineer, and Advisor)
  • Dr. Arman Rahmim (Fund Provider, Medical Imaging Expert, Evaluator, and Advisor)

📚 Citation

Coming soon

📜 License

This open-source software is released under the MIT License, which grants permission to use, modify, and distribute it for any purpose, including research or commercial use, without requiring modified versions to be shared as open source. See the LICENSE file for details.

Support

  • Issues: GitHub Issues
  • Documentation: This README and the included guides
  • Examples: See examples/basic_usage.py

Acknowledgment

Metadata

Release files for radiuma-suv 1.0.1

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for radiuma-suv 1.0.1
File Size Uploaded Provenance Actions
radiuma_suv-1.0.1.tar.gz 44.6 kB Without provenance Details for radiuma_suv-1.0.1.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for radiuma-suv 1.0.1
File Interpreter ABI Platform Actions
radiuma_suv-1.0.1-py3-none-any.whl Python 3 none any Details for radiuma_suv-1.0.1-py3-none-any.whl

Total release size: 79.2 kB

Release files / radiuma_suv-1.0.1.tar.gz

Download URL radiuma_suv-1.0.1.tar.gz
Size 44.6 kB
Tags Source
SHA-256 checksum
How to use checksums
4d32db167193dcc011437614a052c6b422abfb6032818eee7a17ce86507988ea
BLAKE2b-256 checksum
How to use checksums
0dfb59097993d5b39facb0f3ebdd92dfecd9a3ea4a9e8cf79219174d35329c6d
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.13.7

Release files / radiuma_suv-1.0.1-py3-none-any.whl

Download URL radiuma_suv-1.0.1-py3-none-any.whl
Size 34.5 kB
Tags Python 3
SHA-256 checksum
How to use checksums
0210c6065216a6d134b861aa6b499d874e6647b8851526b65693afedc9cb00c0
BLAKE2b-256 checksum
How to use checksums
2812ceba1f89bb05b9ede436c078d9705954593340bf89dfa9aa1edefca9381a
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.13.7

Security for radiuma-suv 1.0.1

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page