**Radiology Objects in COntext Version 2 (ROCOv2)** is an updated and expanded multimodal image dataset for medical image analysis. It consists of 79,789 radiological images and associated captions and medical concepts extracted from the PMC Open Access subset. This dataset is an extension of the original ROCO dataset, adding 35,705 new images and providing manually curated concepts for imaging modalities, including additional anatomical and directional concepts for X-rays. The dataset is suitable for training image annotation models, multi-label image classification using Unified Medical Language System (UMLS) concepts, and pre-training of medical domain models. It has been used in the concept detection and caption prediction tasks of ImageCLEFmedical Caption 2023. The dataset includes images from various imaging modalities and covers a wide range of anatomical regions, making it valuable for developing and evaluating models for tasks such as image caption generation and image retrieval. The creation process involved downloading and filtering images from the PMC Open Access Subset, extracting captions, and manually curating concepts. The dataset is available on Zenodo and can be used for various applications, including training models for structured medical reporting and multi-task learning.**Radiology Objects in COntext Version 2 (ROCOv2)** is an updated and expanded multimodal image dataset for medical image analysis. It consists of 79,789 radiological images and associated captions and medical concepts extracted from the PMC Open Access subset. This dataset is an extension of the original ROCO dataset, adding 35,705 new images and providing manually curated concepts for imaging modalities, including additional anatomical and directional concepts for X-rays. The dataset is suitable for training image annotation models, multi-label image classification using Unified Medical Language System (UMLS) concepts, and pre-training of medical domain models. It has been used in the concept detection and caption prediction tasks of ImageCLEFmedical Caption 2023. The dataset includes images from various imaging modalities and covers a wide range of anatomical regions, making it valuable for developing and evaluating models for tasks such as image caption generation and image retrieval. The creation process involved downloading and filtering images from the PMC Open Access Subset, extracting captions, and manually curating concepts. The dataset is available on Zenodo and can be used for various applications, including training models for structured medical reporting and multi-task learning.