RadGenome-Chest CT is a comprehensive, large-scale, region-guided 3D chest CT interpretation dataset built upon CT-RATE. It includes 197 organ-level segmentation masks, 665,000 multi-granularity grounded reports, and 1.3 million grounded visual question-answering (VQA) pairs. The dataset was created by extending the original CT-RATE dataset with detailed segmentation masks, anatomically structured reports, and VQA pairs. The segmentation masks provide visual clues for interpretation, while the grounded reports and VQA pairs link text to specific anatomical regions in the CT volumes. All grounded reports and VQA pairs in the validation set have been manually verified to ensure quality. RadGenome-Chest CT aims to advance the development of multimodal medical foundation models by enabling models to generate texts based on segmentation regions, which was previously unattainable with existing datasets. The dataset will be released to support further research and development in this field. The dataset includes a hierarchical anatomical structure, detailed segmentation masks, and structured reports that facilitate the generation of grounded visual question-answering pairs. The dataset also includes a wide range of question types, including abnormality, presence, location, and size, which are used to train models for medical image analysis. The dataset has been validated with a distribution of normal and abnormal cases, and word clouds of frequent abnormalities and disorders are provided for analysis. The dataset is expected to significantly advance the development of multimodal medical AI models, enabling them to generate texts based on segmentation regions and improve interpretability and patient care.RadGenome-Chest CT is a comprehensive, large-scale, region-guided 3D chest CT interpretation dataset built upon CT-RATE. It includes 197 organ-level segmentation masks, 665,000 multi-granularity grounded reports, and 1.3 million grounded visual question-answering (VQA) pairs. The dataset was created by extending the original CT-RATE dataset with detailed segmentation masks, anatomically structured reports, and VQA pairs. The segmentation masks provide visual clues for interpretation, while the grounded reports and VQA pairs link text to specific anatomical regions in the CT volumes. All grounded reports and VQA pairs in the validation set have been manually verified to ensure quality. RadGenome-Chest CT aims to advance the development of multimodal medical foundation models by enabling models to generate texts based on segmentation regions, which was previously unattainable with existing datasets. The dataset will be released to support further research and development in this field. The dataset includes a hierarchical anatomical structure, detailed segmentation masks, and structured reports that facilitate the generation of grounded visual question-answering pairs. The dataset also includes a wide range of question types, including abnormality, presence, location, and size, which are used to train models for medical image analysis. The dataset has been validated with a distribution of normal and abnormal cases, and word clouds of frequent abnormalities and disorders are provided for analysis. The dataset is expected to significantly advance the development of multimodal medical AI models, enabling them to generate texts based on segmentation regions and improve interpretability and patient care.