About the Project: RetinaMyopiaAI

Inspiration:
The inspiration for this project came from the growing prevalence of myopia worldwide and the need for early, accurate diagnosis. Many patients experience progressive vision loss due to delayed detection. I wanted to leverage AI and retinal imaging to create a tool that can assist ophthalmologists and improve patient outcomes.

What I Learned:
Through this project, I gained hands-on experience in:

Medical image processing, including retinal fundus image enhancement and segmentation.
Machine learning and deep learning, particularly CNNs (Convolutional Neural Networks) for image classification.
Clinical relevance of retinal biomarkers, such as optic disc size, cup-to-disc ratio, and peripapillary atrophy.
Data annotation and preprocessing, ensuring images are normalized and labeled for training.

How I Built It:
The project was implemented in Python using libraries such as OpenCV, scikit-image, and PyTorch.

Data Collection: Retinal fundus images were collected from public datasets.
Preprocessing: Images were resized, normalized, and enhanced to highlight key retinal structures.
Model Development: A CNN-based classifier was trained to detect myopia severity levels (mild, moderate, high).
Evaluation: Model performance was measured using accuracy, precision, recall, and F1-score.

Challenges Faced:

Image quality variability: Different fundus cameras produced images of varying resolution and brightness, requiring extensive preprocessing.
Limited labeled data: Annotated myopia images were scarce, necessitating data augmentation and careful training.
Class imbalance: Severe myopia images were less frequent, which required weighted loss functions and oversampling techniques.

Impact:
RetinaMyopiaAI demonstrates how AI can assist in early detection of myopia using fundus images. This project combines medical knowledge, computer vision, and machine learning to create a practical tool for ophthalmology.

Example of a simple retinal measurement (using LaTeX):

The optic disc radius ( r ) can be converted from pixels to millimeters:

[ r_{mm} = r_{px} \times 0.025 ]

Where ( r_{px} ) is the radius measured in pixels. This helps quantify structural changes in myopic eyes.