Image Dataset on Eye Diseases Classification (Uveitis, Conjunctivitis, Cataract, Eyelid) with Symptoms and SMOTE Validation

Published: 12 December 2024| Version 2 | DOI: 10.17632/n9zp473wfw.2
Contributors:
, Marzia Ahmed

Description

Dataset Description: This dataset contains images and corresponding symptom descriptions for five types of eye diseases: Uveitis, Conjunctivitis, Cataract, Eyelid Drooping, and Normal. The dataset is intended for use in medical image analysis and machine learning model development. It includes image data (.jpg format) along with detailed descriptions of the diseases and their symptoms. Diseases Included: Normal: No abnormalities, clear vision, no redness or swelling. Uveitis: Eye redness, pain, blurred vision, sensitivity to light, and floating spots. Conjunctivitis: Redness, itching, tearing, discharge, and crusting of the eyelids. Cataract: Cloudy or blurred vision, difficulty seeing at night, sensitivity to glare. Eyelid Drooping: Drooping eyelids, swelling, irritation, and lumps on more than one eyelid. Symptoms: Each disease is associated with its symptoms as listed above. These symptoms are designed to help researchers and practitioners in the classification and diagnosis of these diseases based on visual and textual information. All the symptoms and dataset has been checked and corrected from a Professor from Bangladesh Eye Hospital. Data Collection: Images: The images were collected from online sources (via Google search) using disease-specific keywords. Data Quality Control: Duplicate images were removed, and the dataset was carefully reviewed for accuracy. Balanced Dataset: To ensure a balanced distribution of images across all diseases, SMOTE (Synthetic Minority Over-sampling Technique) was applied. The final dataset includes an equal number of images for each disease category (649 images per disease). Dataset Validation: Before SMOTE: Cataract: 544 images Conjunctivitis: 357 images Eyelid Drooping: 525 images Normal: 649 images Uveitis: 223 images After SMOTE: All diseases now have 649 images, resulting in a balanced dataset. The use of SMOTE ensures that all disease categories are equally represented, mitigating the risk of model bias due to class imbalance. File Formats: Images: JPEG format (.jpg) Dataset Usage: This dataset is intended for research purposes, specifically in medical image classification tasks. It can be used to train, test, and validate machine learning models that aim to identify and diagnose eye diseases based on images and symptoms. Ethics & Consent: The images used in this dataset have been gathered from publicly available online sources, and appropriate ethical considerations were taken into account. The dataset has been anonymized, and no personally identifiable information is included.

Files

Institutions

Daffodil International University

Categories

Cataract, Uveitis, Eyelid Disease, Image Classification, Conjunctivitis

Licence