Multimodal Orbscan IIz Dataset: 3,000 Axial Power Anterior Maps with OCR-Extracted Clinical Parameters from 3,000 Unique Patients

Description

This dataset contains 3,000 anonymized Orbscan IIz Axial Power Anterior corneal topography maps acquired between 2007 and 2025 from 3,000 unique patients at a single ophthalmology center in Iraq. It is a paired multimodal dataset comprising processed topography images and their corresponding OCR-extracted clinical parameters, including Sim K values, astigmatism, mean power, steep and flat axes, and related corneal measurements. All images were de-identified, cropped to the central corneal map region, and resized to 1024 × 1024 pixels. A manually verified subset of the OCR-extracted metadata was reviewed by domain experts to assess extraction quality. The dataset is intended for research in artificial intelligence, corneal topography analysis, medical image computing, and multimodal learning.

Steps to reproduce

1. Original Orbscan IIz screenshots were collected from routine clinical examinations. 2. Personal identifiers were removed and replaced with anonymized labels. 3. Images were cropped to the central corneal map region and resized to 1024 × 1024 pixels. 4. Clinical parameters were extracted using OCR (Tesseract). 5. A subset of the extracted metadata was manually verified by domain experts. 6. Images were linked to metadata using unique filenames.

Institutions

• University of Tikrit

  Salah ad Din, Tikrit

Categories

Licence