ENHANCE-PET/PUMA - Githubissues

PUMA 1.0 🐾 - One Image multiple perspectives 🎭

🌈 PET Reimagined: From Monochrome to a Palette of Possibilities with PUMA 1.0

Step into the vibrant new world of PET imaging with PUMA 1.0 🌟, where traditional monochrome scans are transformed into a dynamic spectrum of diagnostic data. 🎨

With PUMA 1.0, experience the power of multiplexing: a process that fuses multiple tracer images into a single, multicolored composite that reveals not just the presence of disease but its multifaceted physiological context. 🔄 The multiplexed approach provides a more comprehensive view, helping clinicians uncover nuanced insights about tumors and their microenvironments. 🔍

Whether you're in a high-tech lab or a remote clinic, PUMA 1.0 delivers high-quality, multiplexed PET scans that are as rich in detail as they are in color. 🏥 It’s not just an upgrade—it’s a whole new way of seeing PET data, turning every image into a detailed map of insight and opportunity. 🗺️

🎉 Key Features

🚀 Versatile and Powerful

Run PUMA 1.0 on any device, any OS, from x86 to ARM64 (hello, Apple Silicon fans!). Whether you’re on a high-end GPU or a humble CPU, PUMA adapts to your needs without breaking a sweat.

🔍 Precision Meets Speed

Harness the power of AI with MOOSE-driven segmentations and ‘greedy’ library’s wizardry for diffeomorphic image registration. PUMA 1.0 nails the perfect balance of sharp accuracy and zippy performance.

🎨 Art of Imaging

Why settle for ordinary when you can visualize in vibrant RGB? Each color shines a spotlight on a different tracer, turning complex data into a vivid, easy-to-interpret display. With processing times ranging from just 5 to 12 minutes, you're all set for a speedy yet thorough diagnostic journey.

🚀 Why PUMA?

https://github.com/ENHANCE-PET/PUMA/assets/48599863/03368642-a288-44cb-8eaf-4833380a26c8

Requirements ✅

Before stepping into the future with PUMA 1.0, here's what you need for an optimal experience:

Operating System: Windows, Mac, or Linux - PUMA 1.0 is versatile and works across all these platforms.
Memory: Make sure your system has enough memory (8-16 GB) to run the tasks smoothly.
GPU: You would benefit from a cuda-enabled GPU (NVIDIA), 8 GB or more, or MPS (Apple silicon) - if you like quick results ⚡️. But if you don't fret not, PUMA runs a cpu-tuned version if there is no GPU.
Python: PUMA 1.0 operates the best with Python 3.10.

Once these specifications are met, you're all set to experience PUMA 1.0's capabilities.

Installation Guide 🛠️

Installation is a breeze on Windows, Linux, and MacOS. Follow the steps below to start your journey with PUMA 1.0.

For Linux and MacOS 🐧🍏

Create a Python environment named 'puma-env' or as per your preference.
```
python3 -m venv puma-env
```

Activate the environment.

source puma-env/bin/activate  # for Linux
source puma-env/bin/activate  # for MacOS

Install PUMA 1.0.
```
pip install pumaz # for Linux and MacOS
```
Apple M1 Specific Installation: If you are installing PUMA on an Apple Silicon device (e.g., Apple M1), follow this step. Do not do this if you are installing PUMA on Linux!
```
pip install git+https://github.com/LalithShiyam/pytorch-mps.git
```

Congratulations! You're all set to start using PUMA 1.0.

For Windows 🪟

Create a Python environment, e.g., 'puma-env'.
```
python -m venv puma-env
```
Activate the environment.
```
.\puma-env\Scripts\activate
```
Install PUMA 1.0.
```
pip install pumaz
```

You're now ready to experience the precision and speed of PUMA 1.0.

Usage Guide 📚

Start your journey with PUMA 1.0 by using our straightforward command-line tool. It requires the directory path containing different tracer images, and each image should be stored in separate folders. Here's how you can get started:


   pumaz \
       -d   <path_to_image_dir>              # Directory path containing the images to be analyzed
       -ir  <regions_to_ignore>              # Regions to ignore: arms, legs, head, none
       -m                                    # Optional: Enable multiplexed RGB image output
       -cs  <color_selection>                # Optional: Custom color selection for RGB output (requires -m)
       -c2d <convert_back_to_dicom>          # Optional: Once set, the generated nifti images will be converted back to DICOM

<path_to_image_dir> refers to the parent directory containing different tracer images in their respective sub-directories.
-ir specifies the regions to be ignored during registration. If you don't want to ignore any regions, use none. If you want to ignore the arms, legs, or head during registration, pass the corresponding regions delimited by a ,. For example: -ir head,arms to ignore the head and arms.
-m will activate the output of a multiplexed RGB image of the combined tracer images.
-cs, when passed along with -m, PUMA will ask you to provide a custom order of color channels for the corresponding tracer images. That way, you can freely decide which tracer image is associated with which channel.
-c2d, when set the generated aligned nifti images will be converted back to DICOM.

For assistance or additional information, you can always type:

pumaz -h

Example usage:

Apply PUMA to images in a directory, ignoring arms and legs, with multiplexed RGB output and custom colors:

pumaz -d /path/to/images -ir arms,legs -m -cs -c2d

Directory Structure and Naming Conventions for PUMA 📂🏷️

PUMA 1.0 requires your data to be structured in a certain way. It supports DICOM directories and NIFTI files. For NIFTI files, users need to ensure that the files are named with the correct modality tag at the start.

Required Directory Structure 🌳

Here is the directory structure that PUMA 1.0 expects:

📁 Parent_Directory
│
└───📂 Tracer1 # can be named anything
│   │
│   └───📁 PET_DICOM_Directory or 🗃️ PT_xxxx.nii.gz # If it's DICOM, the folder name can be anything, but if nifti use a prefix 'PT' for PET
│   │
│   └───📁 CT_DICOM_Directory or 🗃️ CT_xxxx.nii.gz # If it's DICOM, the folder name can be anything, but if nifti use a prefix 'CT' for CT
│
└───📂 Tracer2
│   │
│   └───📁 PET_DICOM_Directory or 🗃️ PT_xxxx.nii.gz
│   │
│   └───📁 CT_DICOM_Directory or 🗃️ CT_xxxx.nii.gz
...

└───📂 Tracer3
    │
    └───📁 PET_DICOM_Directory or 🗃️ PT_xxxx.nii.gz
    │
    └───📁 CT_DICOM_Directory or 🗃️ CT_xxxx.nii.gz

Naming Conventions 🏷️

For DICOM directories, no specific naming is required.
For NIFTI files, the file should start with the DICOM modality tag (e.g., 'PT' or 'CT') followed by the desired name. For example, 'PT_MySample.nii.gz'.

Note: All the PET and CT images related to a tracer should be placed in the same directory named after the tracer.

🚀 Benchmarks

A Note on QIMP Python Packages: The 'Z' Factor 📚🚀

All of our Python packages here at QIMP carry a special signature – a distinctive 'Z' at the end of their names. The 'Z' is more than just a letter to us; it's a symbol of our forward-thinking approach and commitment to continuous innovation.

Our PUMA package, for example, is named as 'pumaz', pronounced "puma-see". So, why 'Z'?

Well, in the world of mathematics and science, 'Z' often represents the unknown, the variable that's yet to be discovered, or the final destination in a series. We at QIMP believe in always pushing boundaries, venturing into uncharted territories, and staying on the cutting edge of technology. The 'Z' embodies this philosophy. It represents our constant quest to uncover what lies beyond the known, to explore the undiscovered, and to bring you the future of medical imaging.

Each time you see a 'Z' in one of our package names, be reminded of the spirit of exploration and discovery that drives our work. With QIMP, you're not just installing a package; you're joining us on a journey to the frontiers of medical image processing. Here's to exploring the 'Z' dimension together! 🚀