🧠 PET-ki-proc

Automated processing pipeline for dynamic [18F]-DOPA PET and Ki estimation.

Figure: Overview of PET-ki_proc pipeline. Some visual elements were created with BioRender.com; others are KI generated.

---

🧾 Processing Overview

1. Motion Correction

   • Correct PET data from 5 min post-injection using FSL mcflirt.

2. Denoising (optional)

   • Apply Chambolle's total variation denoising (via scikit-image).

3. Mask Preparation

   • Register T1w image to PET space (FSL flirt).
   • Normalize to MNI-152 (FSL flirt + fnirt).
   • Register striatum & cerebellum atlases to native space.
   • Optionally restrict striatum atlas to GM mask.
   • Apply absolute or percentile threshold & binarize.

4. Patlak Fitting

   • Estimate Ki using Gjedde-Patlak linear modeling using cerebellum as reference.

---

💻 Usage

Run for one subject:

run_PET_ki_proc.py -subject sub-001 -session 01 -denoising=True -gm_mask=True -ref_frame=30

Run for all subjects in a session:

run_all.py 01 -denoising=True

---

⚙️ Arguments

Argument	Description
-subject	Subject ID (e.g., sub-001)
-session	Session ID (e.g., 01)
-subject_path	Path to BIDS dataset
-ref_frame	Reference frame for motion correction (default: last-frame, e.g., -ref_frame=30)
-denoising	Apply denoising (True/False, default: True)
-gm_mask	Restrict striatum atlas to GM mask (True/False, default: True)
-weight	Denoising weight (default: 100)
-num_iter	Max number of iterations for denoising (default: 1000)
-atlas	Atlas for masking (see below)
-thr_striatum	Absolute threshold for striatum mask (e.g., -thr_striatum=0.6)
-thrp_striatum	Percentile threshold for striatum mask (e.g., -thrp_striatum=0.6)
-thr_cerebellum	Absolute threshold for cerebellum mask (e.g., -thr_cerebellum=0.9)
-thrp_cerebellum	Percentile threshold for cerebellum mask (e.g., -thrp_cerebellum=0.9)

---

📥 Input

• BIDS-formatted dataset
• Decay-corrected dynamic PET image:
  sub-XXX/ses-YY/pet/sub-XXX_ses-YY_trc-18FDOPA_rec-acdyn_pet.nii
• T1w anatomical image:
  sub-XXX/ses-YY/anat/sub-XXX_ses-YY_T1w.nii

📤 Output

in sub-XXX/ses-YY/pet/patlak_fit/

• Ki map: patlak/*_ki_map_thr0.4.nii.gz
• R² map: *patlak/_r2_map_thr0.4.nii.gz
• Results sheet: *_OGI_3_thr0.6_denoised_PET_ki.xlsx
• Summary PDF: *_OGI_3_thr0.6_denoised_PET_ki.pdf

---

🗺 Available Atlases

• OGI_3: Oxford-GSK-Imanova (3 subregions)
• OGI_7: Oxford-GSK-Imanova (7 subregions)
• HO: Harvard-Oxford subcortical atlas
• CIT: CIT168 subcortical atlas
• add your own atlas to pwt_ki_proc/app/data/atlas

---

Installation

Option 1: 🐳 Docker

Run with Docker (recommended for easiest setup).

Pull the image:

docker pull juliasbrain/pet_ki_proc

Run docker:

docker run --rm -it -v /path/to/preprocData:/app/preprocData juliasbrain/pet_ki_proc run_PET_ki_proc.py -subject sub-001 -session 01 -denoising=True
docker run --rm -it -v /path/to/preprocData:/app/preprocData juliasbrain/pet_ki_proc run_all.py 01 -denoising=True

Option 2: 🐍 Python package

Install Python package via pip (requires FSL installed on your machine).

Install:

pip install git+https://github.com/juliasbrain/PET_ki_proc.git

Run:

run_PET_ki_proc -subject sub-001 -session 01 -denoising=True -subject_path /path/to/preprocData

run_all 01 -denoising=True -subject_path /path/to/preprocData

---

📚 Summary pdf

---

📎 Citation

Please cite this repo when using the application. https://github.com/juliasbrain/PET_ki_proc