This project is an open, modular pipeline designed to analyze a patient’s brain MRI and return two high-value outputs:
- Whether MS-like lesions are present
- Which cognitive, emotional, or physical symptom domains those lesions are most likely to affect
The goal is not diagnosis.
The goal is patient-owned insight — giving individuals clarity on their MRI scans over time while keeping a clinician in the loop.
The repository includes the full processing pipeline: ingestion, harmonization, preprocessing, lesion detection, atlas overlap, symptom-domain mapping, and human-readable reporting.
- Folder structure
README.TXT - This readme file
bussiness_modals - PDF of 90 day implementation plan
data - Patient Data ( Dicom MRI )
grants - grant information
images - images used in this readme file
libary - code for functionality
enviroment.yml - enviroment needed for python
pitch_decks - varrious pitch decks (clincial, executive, general)
proposals - proposals ex Kettering Health
.gitignore - stuff to ignore
app.py - graphql endpoint
test_app.py - test for app
msDetector.py - main code for ms detector
test_msDetector.py - test for main code
MS patients often undergo years of MRI scans without clear explanations of:
- What changed
- Why it matters
- What symptoms to expect
This tool fills that gap by providing:
- Longitudinal lesion tracking
- Template-aligned, harmonized brain images
- Region-based lesion interpretation
- Symptom inference based on neuroscientific mapping
- Simple, patient-readable summaries
This is the foundation for a future clinical-grade, one-click MRI intelligence engine.
A 6-stage neuroimaging + AI workflow:
- Converts DICOM → NIfTI
- Standardizes voxel spacing and orientation
- Ensures comparability across scanners
File: 01_ingest_and_harmonize.py
- N4 bias-field correction
- Skull stripping / brain extraction
- Registers to MNI152 template
- Produces clean, normalized volumes for AI
Files:
02_preprocess_and_register.py
Includes: MNI152_T1_1mm.nii.gz templates
- UNet / nnU-Net–style deep learning
- Generates voxel-level lesion masks
- Returns lesion presence, volume, and distribution
Files:
03_train_lesion_model.py
04_inference_and_longitudinal_report.py
- Aggregates MRI results across dates
- Computes lesion count, lesion load, and emergence of new lesions
- Builds patient-level progression profiles
Included in: 04_inference_and_longitudinal_report.py
Maps lesions → brain regions → symptom domains.
Intersects lesion masks with anatomical atlases.
File: 4_1_loadAtlasAndComputeRegionOverlaps.py
Maps affected regions to cognitive/emotional/motor functions.
File: 4_2_mapRegionsToSymptomDomains.py
Example domains:
- Executive function
- Emotional regulation
- Sensory processing
- Motor coordination
- Memory systems
Generates clinical-style interpretation:
Lesions Detected: Yes Most likely affected domains:
Executive function
Emotional regulation
Sensory processing
File: 4_3_scanSummary.py
A single call runs the entire pipeline:
from analyze_brain_for_symptoms_image import analyze_brain_for_symptoms
has_lesions, symptoms = analyze_brain_for_symptoms("example.nii.gz")
print("Lesions present:", has_lesions)
print("Likely symptoms:", symptoms)
Example Output:
Lesions present: True
Likely symptoms: ['executive dysfunction', 'emotional regulation challenges']📁 Repo Structure
ms_detection/
│── 01_ingest_and_harmonize.py
│── 02_preprocess_and_register.py
│── 03_train_lesion_model.py
│── 04_inference_and_longitudinal_report.py
│── 4_1_loadAtlasAndComputeRegionOverlaps.py
│── 4_2_mapRegionsToSymptomDomains.py
│── 4_3_scanSummary.py
│── analyze_brain_for_symptoms_image.py
│── MNI152_T1_1mm.nii.gz
│── data/
└── README.md
- To run a single-image analysis:
'''python analyze_brain_for_symptoms_image.py /path/to/your_mri.nii.gz'''
No GPU required for demonstrating the pipeline structure
- A real trained checkpoint
- A stub for demonstration / IP review
🏥 Clinical Vision A future version of this tool supports: Upload MRI (NIfTI or DICOM) Auto-run lesion detection Auto-map regions → symptoms Auto-generate PDF summaries Longitudinal tracking across years
Optional clinician review for safety
A clinical-grade version would include:
One-click PDF export Time-series lesion progression Symptom-domain probability mapping Secure patient data handling (HIPAA-compliant)
🧩 Why This Matters
MS imaging is complex. Patients deserve clarity.
This tool offers:
A sense of control
Explanations for symptoms
A way to track change over time
Better communication with clinicians
It is not diagnostic — it is empowerment through understanding.
📚 Future Work
Train/finetune models on public MS datasets
Expand symptom ontology
Add FLAIR + diffusion support
Build a web app wrapper
Integrate one-click PDF export
Deploy as fog/edge compute module for patient-owned devices
👤 Author
Ali Payne Entrepreneur, engineer, and researcher exploring patient-centered neurotechnology.
Pipeline Diagram
┌───────────────────────────┐
│ Patient-owned MRI data │
│ (DICOM from hospitals) │
└───────────┬──────────────┘
│
▼
┌───────────────────────────────────────┐
│ 1. Ingest & Harmonize │
│ - DICOM → NIfTI │
│ - Standardize orientation/voxel size│
└───────────────────┬──────────────────┘
│
▼
┌────────────────────────────────────────────┐
│ 2. Preprocess & Normalize │
│ - Bias correction (ANTs N4) │
│ - Brain extraction │
│ - Register to MNI │
└───────────────────┬──────────────────────┘
│
▼
┌────────────────────────────────────────────┐
│ 3. AI Lesion Detection │
│ - UNet / nnU-Net │
│ - Output: voxelwise lesion masks │
└───────────────────┬──────────────────────┘
│
▼
┌────────────────────────────────────────────┐
│ 4. Lesion Quantification & Longitudinal │
│ - Lesion count, volume, location │
│ - Time-series progression │
└───────────────────┬──────────────────────┘
│
▼
┌────────────────────────────────────────────┐
│ 5. Symptom–Lesion Mapping │
│ - Atlas intersection │
│ - Map regions → cognitive/emotional fxn │
│ - Generate likely symptom domains │
└───────────────────┬──────────────────────┘
│
▼
┌────────────────────────────────────────────┐
│ 6. Clinical Review & Reporting │
│ - Overlays + regional labels │
│ - Narrative summary │
└───────────────────────────────────────────┘
# Single-Function Example
image_path = "/mnt/data/your_mri_image.nii.gz"
has_lesions, symptoms = analyze_brain_for_symptoms(image_path)
print("Lesions present:", has_lesions)
print("Likely symptoms:", symptoms)
# Example Output:
Lesions present: True
Likely symptoms: [
"executive dysfunction",
"emotional regulation issues",
"anxiety-like symptoms",
"interoceptive disturbance",
"memory issues"
]