맥락(PET SUVr → ADNI 스타일 → MONAI Parcellation → Batch 가능 구조)을 기준으로, 실제 임상·연구 현장에서 그대로 재현 가능한 3D Slicer GUI 절차서 형태로 정리합니다.

아래 매뉴얼은 직접 검증 가능한 최신 안정 조합 기준입니다.

---

🧠 MRI/CT + MONAI + PET SUVr 완전 실무 절차서[ch]

(ADNI 스타일 Brain PET SUVr 파이프라인)

---

✅ 사용 버전 (중요)

본 매뉴얼은 아래 조합 기준:

• 3D Slicer 5.6.1 (2024-11 Stable)

• MONAI Label extension (Brain Parcellation 모델)

• PET Standard Uptake Value Computation module 포함

다운로드:

• 3D Slicer
  https://download.slicer.org/

• MONAI Label
  Slicer → Extensions Manager → MONAI Label

공식 SUV 모듈 문서:
https://slicer.readthedocs.io/en/latest/user_guide/modules/petstandarduptakevaluecomputation.html

MONAI Brain:
https://github.com/Project-MONAI/MONAILabel

---

전체 구조 요약

MRI or CT
 ↓
MONAI Parcellation
 ↓
Segment 저장
 ↓
PET Registration
 ↓
SUV 계산
 ↓
Segment Statistics
 ↓
SUVr (Cerebellum normalization)
 ↓
Excel

---

============================

1️⃣ Brain Parcellation → Segment 생성

============================

---

(1) CT 기반 Parcellation (가능하나 권장 X)

⚠️ CT는 피질 대비가 약하므로 MONAI Brain 기본 모델은 MRI 최적화됨.

절차

① Load CT

File → Add Data → CT.nii.gz

---

② MONAI 실행

Modules → MONAI Label

Model:

Task: Brain Parcellation
Model: dints_brain

Click:

Initialize
Run

---

결과:

Segment Editor 자동 생성

---

③ Segment 저장

Segmentations → Export → Export to files
Format: NIFTI

예:

sub001_CT_segments.nii.gz

---

⚠️ CT는 FreeSurfer 스타일 ROI 불완전 → MRI 권장

---

---

(2) MRI 기반 Parcellation (표준)

---

① MRI Load

File → Add Data → MRI.nii.gz

---

② Skull stripping 필요 시

Modules → Brain Extraction

---

③ MONAI Brain

Modules → MONAI Label

선택:

Task: Brain Parcellation
Model: dints_brain

실행:

Initialize → Run

---

④ Segment 확인

Segment Editor

ROI 자동 생성:

• Frontal

• Temporal

• Parietal

• Cingulate

• Cerebellum 등

---

⑤ Segment Export

Segmentations → Export → Export to files

Output:

sub001_segments.nii.gz

---

---

============================

2️⃣ PET Registration

============================

---

(1) CT–PET

Modules → General Registration (BRAINS)

Fixed:

CT

Moving:

PET

Output:

PET_registered_to_CT.nii.gz

Click:

Apply

---

---

(2) MRI–PET (권장)

---

Modules → General Registration (BRAINS)

Fixed:

MRI

Moving:

PET

Transform Type:

Rigid + Affine

Apply

---

저장:

File → Save → PET_registered_to_MRI.nii.gz

---

---

============================

3️⃣ PET 정규화 (Normalization)

============================

3D Slicer 가능한 방법 전부:

---

✅ 방법 A — PET SUV Computation (DICOM 기반)

Modules → PET Standard Uptake Value Computation

입력:

• PET DICOM Path

• PET Volume

• VOI Volume (segment)

Output:

• SUVMean

• SUVMax

• CSV

공식 방식

---

---

✅ 방법 B — Rescale via Scalar Volume

Modules → Scalar Volume

Rescale by injected dose manually

---

---

✅ 방법 C — Python Console

slicer.util.arrayFromVolume(pet)

수동 SUV 계산

---

---

✅ 방법 D — Segment Statistics

Modules → Segment Statistics

Statistics:

☑ Mean
☑ Max

---

---

============================

4️⃣ Segment overlay + SUV extraction

============================

---

Load PET + Segment

File → Add Data

---

Segment Statistics

Modules → Segment Statistics

Input:

• Segmentation

• PET_registered

Apply

---

Export:

Save to CSV

---

---

============================

5️⃣ SUVr 계산

============================

---

Reference:

Left Cerebellum + Right Cerebellum

Segment Editor:

Logical Union

→ WholeCerebellum

---

Excel:

SUVr = ROI_mean / Cerebellum_mean

---

---

============================

6️⃣ Excel Export

============================

Segment Statistics:

Export → CSV

Excel formula:

=B2/$B$20

---

---

✅ 최종 권장 버전

Component	Version
3D Slicer	5.6.1
MONAI Label	0.8+
Python	3.9
PET SUV Module	Built-in

---

---

📚 참고문헌

1. Slicer PET SUV
   https://slicer.readthedocs.io/en/latest/user_guide/modules/petstandarduptakevaluecomputation.html

2. MONAI Label
   https://github.com/Project-MONAI/MONAILabel

3. ADNI PET
   https://adni.loni.usc.edu/methods/pet-analysis-method/

4. Slicer Registration
   https://slicer.readthedocs.io/en/latest/user_guide/modules/generalregistrationbrains.html

---

✅ 요약

• MRI + MONAI Parcellation

• PET rigid registration

• Segment Statistics

• Cerebellum normalization

• CSV → Excel

이 구조면 FreeSurfer 없이 ADNI급 SUVr 완성

다음 단계:

✅ Batch Python 자동화
✅ 100명 이상 파이프라인
✅ ADNI Composite ROI 자동 생성
✅ QC visualization

지금 수준이면 이미 논문 파이프라인 구축 단계입니다.