OCT Vision

Ophthalmology Assistance for Contact Lens Fitting and Corneal Analysis

Engineering Project - Bachelor's Degree in Computer Engineering
Universidade da Coruña

OCT Vision is a specialized tool designed to automate and enhance the measurement of the relationship between the cornea and contact lenses using Anterior Segment Optical Coherence Tomography (AS-OCT) images.

It provides clinicians with precise data to facilitate the calculation and adaptation of lenses, particularly for patients with irregular corneas (e.g., keratoconus, corneal transplants).

• Automated Segmentation: Precisely detects corneal and contact lens surfaces using optimized image processing pipelines.
• Metric Suite:
  • Vertical Clearances: Direct measurement of the sagittal gap.
  • Euclidean Clearance: Minimum distance between surfaces for safety analysis.
  • Zonal Metrics: Analysis of central vs. peripheral thickness.
• Interactive Visualization: Real-time cross-sectional measurements with an interactive thickness profile.
• Demo Mode: Built-in library of diverse clinical cases for training and validation.

• Backend: Flask (Python) with a focus on custom computer vision algorithms.
• Computer Vision: OpenCV, NumPy, and SciPy for edge detection, morphological processing, and polynomial interpolation.
• Frontend: Vanilla JavaScript (Modern ES6+), Tailwind CSS for a premium medical interface.
• Testing: Robust test suite using Pytest for core geometry and processing services.

• Python 3.11+
• uv (Recommended for dependency management)

uv sync

uv run python run.py

Then visit http://localhost:5000 in your browser.

You can easily deploy OCT Vision using Docker and Docker Compose. This ensures all system dependencies (OpenCV, etc.) are correctly configured.

1. Build and Start:

   docker compose up --build -d

2. Access: The application will be available at http://localhost:5000.

1. Build:

   docker build -t oct-vision .

2. Run:

   docker run -p 5000:5000 oct-vision

PYTHONPATH=. uv run python3 -m pytest

The application follows a multi-stage processing pipeline:

1. Preprocessing: Image enhancement (CLAHE) and noise reduction.
2. Segmentation: Canny edge detection and morphological grouping.
3. Analysis: Least-squares polynomial fitting to derive smooth corneal/lens models.
4. Computation: Geometric algorithms to find vertical and minimum Euclidean distances.

This project is licensed under the GNU General Public License v3.0.

Developed with precision for the ophthalmology community.