NVIDIA Launches NV-Reason-CT for 3D CT Analysis
NVIDIA has introduced NV-Reason-CT, a vision language model designed for 3D CT analysis, offering state-of-the-art performance and detailed reasoning capabilities.
NVIDIA announced the launch of NV-Reason-CT, a vision language model (VLM) specifically designed for 3D CT analysis. This model combines a full 3D vision transformer encoder with a Qwen3.5-4B language model trained to generate chain-of-thought reasoning that mirrors the systematic review process of radiologists.
NV-Reason-CT achieves state-of-the-art results on the CT-RATE benchmark, with a Macro-F1 of 0.614 and a Macro-AUROC of 0.871, surpassing existing 3D contrastive and fused 2D/3D baseline models. NIH radiologists have validated the clinical plausibility of the model's structured reports and reasoning traces, noting that the step-by-step thinking process enables trust and auditability.
The model is designed to generate detailed structured reports and emulate the systematic review process of radiologists, covering 30 chest and 29 abdominal abnormalities. It also supports multistep conversational follow-up, allowing clinicians and researchers to ask follow-up questions and probe the model's reasoning at any stage.
NV-Reason-CT is an open research and development foundation, not an autonomous diagnostic system or a cleared clinical product. It is intended for researchers and developers building specialized CT analysis applications, and it integrates with other NVIDIA Medical AI models for segmentation and synthetic data generation.
Source: nvidia-developer
