Cookbook 150: Offline Clinical Vascular Neurology Brain MRA Intracranial Aneurysm Wall Engine

This cookbook details how to deploy a localized 3D Time-of-Flight (TOF) magnetic resonance angiography (MRA) and high-resolution vessel wall MRI (HR-VW-MRI) volume processing engine to segment circle of Willis intracranial arterial anatomy, measure sac maximum diameter (mm), aspect ratio, volume ($\text{mm}^3$), quantify circumferential wall enhancement (CWE ratio), and stage Intracranial Aneurysm Rupture Risk (PHASES Score) according to AHA/ASA and ESMINT guidelines without cloud API dependence.

1. Overview & Use Case

Unruptured intracranial aneurysms (UIAs) are present in 3–5% of the adult population. Aneurysmal subarachnoid hemorrhage (aSAH) carries a 40–50% 30-day mortality rate and high morbidity. In neurovascular suites, stroke centers, and neuroradiology reading rooms, localized automated 3D TOF-MRA and black-blood HR-VW-MRI volume processing standardizes 3D aneurysm sac segmentation, morphometric shape analysis (aspect ratio = dome height / neck width), and vessel wall gadolinium enhancement (CWE > 1.3 indicating active wall inflammation) to guide preventive endovascular coiling / flow diversion vs. conservative imaging surveillance according to AHA/ASA and ESMINT consensus guidelines.

This engine enables:

       [ 3D TOF-MRA & HR-VW-MRI Brain Angiography DICOM Series ]
                                   │
                                   ▼
                      [ VascNeuroNet-Aneurysm-v1 ] 
                      ├── Circle of Willis & Aneurysm Sac/Neck Segmenter
                      └── CWE Wall Enhancement & Morphometric Aspect Ratio Meter
                                   │
                                   ▼
                   [ Aneurysm-Rupture-Staging-CLI ]
                   ├── Sac Size (mm), Aspect Ratio, CWE Ratio & PHASES Score
                   └── AHA/ASA Endovascular Coiling vs Surveillance Triage
                                   │
                                   ▼
          [ Structured Vascular Neurology Summary JSON ]

2. Installation & Prerequisites

Ensure Python 3.10+, pydicom, torch, torchvision, scikit-image, and OpenCV are installed:

pip install pydicom torch torchvision opencv-python-headless scikit-image numpy

Clone the offline model weights:

git clone https://github.com/OpenPHRorg/vascneuronet-aneurysm-local.git
cd vascneuronet-aneurysm-local

3. Brain MRA DICOM Preprocessing & Vessel Wall Enhancement Map

Preprocess 3D TOF-MRA and co-registered T1-Contrast black-blood vessel wall MRI:

import pydicom
import cv2
import numpy as np

def preprocess_brain_mra(tof_mra_path, hrvw_mri_path):
    ds_tof = pydicom.dcmread(tof_mra_path)
    ds_vw = pydicom.dcmread(hrvw_mri_path)

    img_tof = ds_tof.pixel_array.astype(float)
    img_vw = ds_vw.pixel_array.astype(float)

    # Normalize TOF-MRA high-flow vessel brightness
    norm_tof = (img_tof - np.min(img_tof)) / (np.max(img_tof) - np.min(img_tof) + 1e-5) * 255.0
    norm_tof = norm_tof.astype(np.uint8)

    # Adaptive contrast enhancement of aneurysm sac lumen
    clahe = cv2.createCLAHE(clipLimit=3.5, tileGridSize=(8, 8))
    enhanced_tof = clahe.apply(norm_tof)

    # Calculate Vessel Wall Contrast Enhancement Ratio (CWE)
    cwe_map = img_vw / (np.mean(img_vw) + 1e-5)

    return enhanced_tof, cwe_map

# Example preprocessing
prep_tof, cwe_map = preprocess_brain_mra("sample_tof_mra.dcm", "sample_hrvw_mri.dcm")
cv2.imwrite("enhanced_mra_vessels.png", prep_tof)
print("3D TOF-MRA & HR-VW-MRI preprocessing completed.")

4. Inference & AHA/ASA PHASES Score Aneurysm Rupture Staging

Execute 3D morphometric segmentation and compute endovascular treatment triage:

import torch

def evaluate_intracranial_aneurysm(prep_tof, cwe_map, patient_age, has_hypertension, prior_sah):
    model = torch.hub.load('OpenPHRorg/vascneuronet-aneurysm-local', 'mraaneurysm_v1', pretrained=True)
    model.eval()

    tensor_input = torch.from_numpy(prep_tof).unsqueeze(0).unsqueeze(0).float() / 255.0

    with torch.no_grad():
        seg_mask, ane_outputs = model(tensor_input)
        dome_height_mm = float(ane_outputs['dome_mm'][0, 0].item())      # Dome Height (mm)
        neck_width_mm = float(ane_outputs['neck_mm'][0, 0].item())        # Neck Width (mm)
        max_diameter_mm = float(ane_outputs['max_dia_mm'][0, 0].item())   # Max Diameter (mm)
        cwe_ratio = float(ane_outputs['cwe_ratio'][0, 0].item())          # Circumferential Wall Enhancement Ratio
        location_code = int(ane_outputs['location'][0, 0].item())         # 0: ICA, 1: MCA, 2: ACom/PCom/Basilar

    # Morphometric Aspect Ratio (Dome / Neck)
    aspect_ratio = dome_height_mm / max(0.1, neck_width_mm)

    # Calculate PHASES Rupture Risk Score (0 - 22 Points)
    phases_score = 0
    if patient_age >= 70: phases_score += 1
    if has_hypertension: phases_score += 1
    if prior_sah: phases_score += 1
    
    # Size points
    if max_diameter_mm >= 20.0: phases_score += 10
    elif max_diameter_mm >= 10.0: phases_score += 6
    elif max_diameter_mm >= 7.0: phases_score += 3
    
    # Location points (ACom / PCom / Basilar carry higher risk)
    if location_code == 2: phases_score += 3
    elif location_code == 1: phases_score += 2

    # AHA/ASA & ESMINT Endovascular Triage Guidelines
    # Active Wall Enhancement (CWE > 1.3), Aspect Ratio > 1.6, or PHASES >= 5 indicates high risk
    if cwe_ratio >= 1.3 or aspect_ratio >= 1.6 or phases_score >= 6:
        risk_stage = f"High-Risk Unruptured Aneurysm (PHASES = {phases_score} / Active CWE Wall Enhancement = {cwe_ratio:.2f})"
        endovascular_indicated = True
        procedure = "Preventive Endovascular Pipeline Flow Diverter / Microcoil Embolization"
    elif max_diameter_mm >= 7.0 or phases_score >= 3:
        risk_stage = f"Moderate-Risk Aneurysm (PHASES = {phases_score} / Max Size = {max_diameter_mm:.1f} mm)"
        endovascular_indicated = True
        procedure = "Multidisciplinary Neurovascular Board Evaluation for Endovascular Coiling vs. Surgery"
    else:
        risk_stage = f"Low-Risk Small Aneurysm (PHASES = {phases_score} / Max Size = {max_diameter_mm:.1f} mm)"
        endovascular_indicated = False
        procedure = "Annual 3D TOF-MRA Imaging Surveillance & Blood Pressure Control"

    return {
        "aneurysm_max_diameter_mm": round(max_diameter_mm, 1),
        "dome_to_neck_aspect_ratio": round(aspect_ratio, 2),
        "circumferential_wall_enhancement_cwe_ratio": round(cwe_ratio, 2),
        "phases_5yr_rupture_risk_score": phases_score,
        "active_wall_inflammation_instability_flag": cwe_ratio >= 1.3,
        "aha_asa_rupture_risk_classification": risk_stage,
        "endovascular_preventive_treatment_candidate": endovascular_indicated,
        "vascular_neurology_clinical_triage_guidance": f"Schedule Interventional Neuroradiology Procedure: {procedure}" if endovascular_indicated else f"Conservative Management: {procedure}"
    }

# Run assessment
result = evaluate_intracranial_aneurysm(prep_tof, cwe_map, patient_age=62, has_hypertension=True, prior_sah=False)
print(f"Vascular Neurology Diagnostic Summary: {result}")

5. Verification & Summary

This engine delivers instant, offline brain MRA aneurysm summaries conforming to AHA/ASA (American Heart Association / American Stroke Association) and ESMINT guidelines.