Ophthalmic Image Registration Application

A professional Python application for longitudinal ophthalmic image registration and comparison, supporting DICOM and standard image formats with multiple alignment methods.

Version 2.3.0 | Updated: December 2025

Overview

This application provides robust image registration with multiple alignment options:

Alignment Methods

• AKAZE: Fast and accurate feature-based alignment (recommended default)
• ORB: Fastest option for quick previews
• SIFT: Classic algorithm for detailed images
• Manual 3-Point: For difficult cases where automatic methods fail

Supported Modalities

• Fundus photography
• IR (Infrared) en face images
• FAF (Fundus Autofluorescence) images
• Cross-modality registration (IR↔FAF) with CLAHE preprocessing
• Other planar retinal modalities

Features

• DICOM Support: Full DICOM loading with metadata preservation
• Pixel Spacing: Accurate mm measurements from DICOM tags
• Measurement Unit Toggle: Switch between mm and pixels for measurements
• CLAHE Preprocessing: Contrast Limited Adaptive Histogram Equalization for cross-modality registration
• Multimodality Options: Grayscale, invert, contrast, brightness adjustments
• Batch Processing: Register multiple follow-up images at once
• Batch Export: Export all registered images to DICOM, PNG, TIFF, or JPEG
• Reference Baseline: Lock baseline for longitudinal studies
• Measurement Tools: Distance and area measurements via right-click menu
• Registration Points Visualization: Up to 50 spatially distributed points with connecting lines
• Multiple Comparison Modes: Side-by-side, Overlay, Difference, Checkerboard, Split View

Architecture

ophthalmic_registration/
├── core/
│   ├── image_data.py          # Image data structures and metadata
│   └── exceptions.py          # Custom exceptions
├── io/
│   ├── dicom_loader.py        # DICOM image loading with metadata
│   ├── standard_loader.py     # Standard format loading (PNG, TIFF, etc.)
│   └── image_io.py            # Unified image I/O interface
├── preprocessing/
│   └── pipeline.py            # Preprocessing pipeline (CLAHE, etc.)
├── registration/
│   ├── feature_aligner.py     # Feature-based alignment (AKAZE, ORB, SIFT)
│   ├── sift_aligner.py        # SIFT-specific alignment
│   └── registration_pipeline.py  # Unified registration pipeline
├── measurement/
│   └── spatial.py             # Spatial calibration and measurements
├── visualization/
│   └── comparison.py          # Visualization and comparison tools
├── export/
│   └── output.py              # Export utilities (DICOM, PNG, TIFF, JPEG)
├── gui/
│   ├── main_window.py         # Main application window
│   ├── image_viewer.py        # Image viewer with zoom/pan/measurements
│   ├── comparison_view.py     # Multi-mode comparison widget
│   ├── controls_panel.py      # Registration settings panel
│   ├── batch_registration.py  # Batch registration dialog
│   ├── batch_export_dialog.py # Batch export dialog
│   ├── manual_registration.py # Manual 3-point registration
│   └── styles.py              # Modern dark/light themes
└── utils/
    └── logging_config.py      # Logging configuration

Installation

pip install -r requirements.txt

Optional Dependencies

For additional preprocessing options:

pip install scikit-image  # For advanced filters

Quick Start

GUI Application

Launch the graphical user interface:

python run_gui.py

Basic Workflow

1. Load Baseline: File → Open Baseline (Ctrl+O)
2. Load Follow-up: File → Open Follow-up (Ctrl+Shift+O)
3. Register: Click "Register" button or press Ctrl+R
4. Compare: Switch to "Compare" tab to view results
5. Export: File → Export Results (Ctrl+E)

Measurement Tools

Right-click on any image to access measurement tools:

• Measure → Distance: Click two points to measure distance
• Measure → Area: Click multiple points, double-click to complete polygon
• Measurements display in mm if DICOM pixel spacing is available

Batch Processing

1. Load baseline image
2. Registration → Batch Registration (Ctrl+B)
3. Select multiple follow-up images
4. Process all at once

Batch Export

1. Register one or more follow-up images
2. Registration → Batch Export Registered Images (Ctrl+Shift+E)
3. Choose format (DICOM, PNG, TIFF, JPEG)
4. Select additional outputs (overlays, difference maps, reports)

Python API

from ophthalmic_registration import RegistrationPipeline, ImageLoader

# Load images
loader = ImageLoader()
baseline = loader.load("baseline_fundus.dcm")
followup = loader.load("followup_fundus.dcm")

# Register images
pipeline = RegistrationPipeline()
result, registered = pipeline.register_and_apply(baseline, followup)

# Export
from ophthalmic_registration import ExportManager
exporter = ExportManager(output_dir="./results")
exporter.export_registration_results(baseline, followup, registered, result)

Modules

Image I/O (io/)

• Unified loading for DICOM and standard formats
• Automatic metadata extraction (pixel spacing, orientation)
• Intensity normalization for consistent processing

Preprocessing (preprocessing/)

• Grayscale conversion
• CLAHE contrast enhancement
• Vessel enhancement filters
• Resolution normalization

Registration (registration/)

• Feature-based: AKAZE, ORB, SIFT with RANSAC
• Supports: Translation, Euclidean, Affine, Homography motion models
• CLAHE preprocessing for cross-modality alignment

Measurement (measurement/)

• Pixel-to-mm conversion using DICOM pixel spacing
• Distance and area measurements
• Displayed directly on image viewer

Export (export/)

• DICOM export with metadata preservation
• Lossless formats: PNG, TIFF
• Lossy format: JPEG with quality control
• Overlay images, difference maps, transform JSON, reports

License

For research and clinical evaluation purposes.

Author

Created by Li Fan, 2025

References

• AKAZE: Alcantarilla, P.F. et al. (2013). Fast Explicit Diffusion for Accelerated Features in Nonlinear Scale Spaces
• CLAHE: Zuiderveld, K. (1994). Contrast Limited Adaptive Histogram Equalization