Datasets:
license: apache-2.0
tags:
- multimodal
- ophthalmology
- OCT
- benchmark
- medical
- visual question answering
🤗 OCT-Bench
To comprehensively evaluate Multimodal Large Language Models (MLLMs) on optical coherence tomography (OCT) image understanding, we introduce OCT-Bench, a comprehensive benchmark that follows the real-world clinical interpretation workflow from visual perception, medical cognition, to clinical reasoning.
OCT-Bench contains 10,076 expert-verified multiple-choice questions constructed from 4,137 OCT images collected from seven public datasets. The benchmark establishes a hierarchical capability taxonomy consisting of 3 capability dimensions, 9 capability groups, and 20 fine-grained tasks, covering imaging attributes, retinal anatomy, lesion characteristics, disease diagnosis, treatment planning, and prognostic management.
We benchmark 20 representative MLLMs, including proprietary models, open-source general-purpose models, and medical-domain models, providing a comprehensive evaluation of OCT understanding capabilities.
For detailed usage and instructions, please refer to the GitHub page.
You can download OCT-Bench. The expected directory structure is:
OCT-Bench ├── images │ ├── OCT5K │ ├── OCTDL │ └── ... └── VQA ├── T01_VQA.jsonl ├── T02_VQA.jsonl └── ...