CaDIS: Cataract dataset for surgical RGB-image segmentation

Area of Science:

• Ophthalmology
• Computer Vision
• Medical Imaging

Background:

• Video feedback is vital for surgical procedures, providing surgeons with key sensory information.
• Accurate scene understanding is essential for computer-assisted interventions and post-operative surgical analysis.
• Semantic segmentation, identifying instruments and structures, is fundamental for scene understanding in surgery.

Purpose of the Study:

• To introduce a novel dataset for semantic segmentation of cataract surgery videos.
• To benchmark the performance of state-of-the-art deep learning models on this new dataset.
• To enhance the development of computer-assisted interventions and surgical video analysis tools.

Main Methods:

• Development and release of a new dataset for semantic segmentation of cataract surgery videos.
• Benchmarking of several state-of-the-art deep learning models for semantic segmentation.
• Utilizing deep learning techniques for image analysis and semantic segmentation.

Main Results:

• A new, publicly available dataset for cataract surgery video semantic segmentation has been created.
• Performance benchmarks for various deep learning models on the dataset provide insights into current capabilities.
• The dataset complements existing resources like the CATARACTS challenge dataset.

Conclusions:

• The new dataset will facilitate advancements in computer-assisted interventions and surgical video analysis.
• Benchmarking results offer a baseline for future research in semantic segmentation for ophthalmic surgery.
• Public availability of the dataset promotes further research and development in the field.