DeepTumorVQA 2.0: 3D Abdominal CT Diagnostic VQA Benchmark

DeepTumorVQA v2 is a 3D abdominal-CT diagnostic Visual Question Answering benchmark containing 438,000 total QA pairs. The dataset includes 10,000 curated benchmark pairs and a 428,000-pair training pool, along with pre-extracted 2D and video modalities and 20,000 agent training trajectories. It was created by the tumor-vqa organization and was last updated in May 2026.

Use Cases

• Train diagnostic AI models based on the 438,000 QA pairs.
• Benchmark VQA model performance on the 10,000 curated clinical question-answer pairs.
• Develop and evaluate agent-based systems using the 20,000 agent training trajectories with tool-use traces.
• Explore multimodal learning with pre-extracted 2D and video representations of CT scans.

Strengths

• Contains 438,000 total QA pairs, providing a substantial training pool.
• Includes a curated benchmark of 10,000 QA pairs for standardized evaluation.
• Offers multiple data modalities: 3D CT scans, pre-extracted 2D slices, and video.
• Provides 20,000 agent training trajectories with tool-use traces for interactive AI research.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for individual components is unknown, which may limit suitability assessment.
• Freshness should be verified as the last update timestamp is from the future (2026-05-25).

Provenance

Source

tumor-vqa (organization on Hugging Face)

Collection Method

Likely curated from clinical CT scans and annotated with diagnostic questions and answers.

Freshness

Last updated 2026-05-25 06:19:50