Multi-Organ Gynaecological Magnetic Resonance Imaging for Brachytherapy-based Oncology

This dataset introduces a Multi-Organ Gynaecological Magnetic Resonance Imaging for Brachytherapy-based Oncology (MOGaMBO) dataset, a novel magnetic resonance imaging dataset aimed to advance research in applications of computational intelligence in brachytherapy diagnosis and organ segmentation for cervical cancer treatment. The dataset comprises high-resolution T2-weighted 3D MR scans from 94 patients with locally advanced cervical cancer (stages IB2–IVA), adhering to FIGO guidelines for interstitial and intra-cavitary brachytherapy. The imaging was performed using a 1.5T GE Signa Explorer scanner, with acquisition parameters TR and TE set to optimal values for soft-tissue contrast at 2600ms and 155ms, respectively, combined with a pixel resolution of 0.5 × 0.5 mm² and 30–50 slices per scan. To ensure dosimetric consistency, bladder volume was standardized via Foley catheterization during imaging. The critical organs-at-risk—urinary bladder, rectum, sigmoid colon, and femoral heads, were manually contoured by expert radiation oncologists using the open-source ITK-SNAP platform, ensuring precise region-of-interest annotations. The dataset underwent rigorous deidentification to protect patient privacy, removing all demographic and identifiable information. MOGaMBO provides a standardized, privacy-compliant resource for developing and validating medical image segmentation or representation learning algorithms, and brachytherapy-related research tools. This dataset addresses a critical gap in accessible, multi-organ imaging resources for the gynaecological brachytherapy dataset, with applications in treatment planning, and AI-driven clinical research. If you use this dataset, then cite the following work in the appropriate format BibTex: @misc{manna2025, title={Federated Self-Supervised Learning for One-Shot Cross-Modal and Cross-Imaging Technique Segmentation}, author={Siladittya Manna and Suresh Das and Sayantari Ghosh and Saumik Bhattacharya}, year={2025}, eprint={2503.23507}, archivePrefix={arXiv}, primaryClass={cs.CV}, url={https://arxiv.org/abs/2503.23507}, } MLA: Manna, Siladittya, et al. "Federated Self-Supervised Learning for One-Shot Cross-Modal and Cross-Imaging Technique Segmentation." arXiv preprint arXiv:2503.23507 (2025). APA: Manna, S., Das, S., Ghosh, S., & Bhattacharya, S. (2025). Federated Self-Supervised Learning for One-Shot Cross-Modal and Cross-Imaging Technique Segmentation. arXiv preprint arXiv:2503.23507. Chicago: Manna, Siladittya, Suresh Das, Sayantari Ghosh, and Saumik Bhattacharya. "Federated Self-Supervised Learning for One-Shot Cross-Modal and Cross-Imaging Technique Segmentation." arXiv preprint arXiv:2503.23507 (2025). Harvard: Manna, S., Das, S., Ghosh, S. and Bhattacharya, S., 2025. Federated Self-Supervised Learning for One-Shot Cross-Modal and Cross-Imaging Technique Segmentation. arXiv preprint arXiv:2503.23507.本数据集为基于近距离放射治疗的妇科多器官磁共振成像（Multi-Organ Gynaecological Magnetic Resonance Imaging for Brachytherapy-based Oncology，简称MOGaMBO）数据集，是一款新型磁共振成像数据集，旨在推动计算智能在宫颈癌治疗的近距离放射治疗诊断与器官分割领域的应用研究。该数据集包含94例局部晚期宫颈癌（FIGO分期IB2–IVA期）患者的高分辨率T2加权三维MR扫描影像，符合国际妇产科联盟（FIGO）制定的间质与腔内近距离放射治疗指南标准。影像采集采用1.5T GE Signa Explorer扫描仪，采集参数重复时间（TR）与回波时间（TE）分别设置为2600ms与155ms，以获得最优软组织对比度；单次扫描的像素分辨率为0.5 × 0.5 mm²，层厚数量为30至50层。为保证剂量学一致性，成像过程中通过弗利导尿管置管（Foley catheterization）对膀胱体积进行标准化处理。高危危及器官——膀胱、直肠、乙状结肠与股骨头，由资深放射肿瘤学家基于开源ITK-SNAP平台进行手动勾画，确保感兴趣区域（region-of-interest，ROI）标注的精准性。该数据集已完成严格的去标识化处理以保护患者隐私，移除了所有人口统计学信息与可识别身份的相关数据。MOGaMBO可为医学图像分割或表征学习算法的开发与验证，以及近距离放射治疗相关研究工具的研发，提供一套标准化、符合隐私合规要求的科研资源。本数据集填补了妇科近距离放射治疗领域可获取的多器官影像资源的关键空白，可应用于治疗规划与AI驱动的临床研究。 若使用本数据集，请以合适的格式引用以下文献： BibTex格式： @misc{manna2025, title={Federated Self-Supervised Learning for One-Shot Cross-Modal and Cross-Imaging Technique Segmentation}, author={Siladittya Manna and Suresh Das and Sayantari Ghosh and Saumik Bhattacharya}, year={2025}, eprint={2503.23507}, archivePrefix={arXiv}, primaryClass={cs.CV}, url={https://arxiv.org/abs/2503.23507}, } MLA格式： Manna, Siladittya, 等. "Federated Self-Supervised Learning for One-Shot Cross-Modal and Cross-Imaging Technique Segmentation." arXiv预印本 arXiv:2503.23507 (2025). APA格式： Manna, S., Das, S., Ghosh, S., & Bhattacharya, S. (2025). Federated Self-Supervised Learning for One-Shot Cross-Modal and Cross-Imaging Technique Segmentation. arXiv预印本 arXiv:2503.23507. 芝加哥格式： Manna, Siladittya, Suresh Das, Sayantari Ghosh, and Saumik Bhattacharya. "Federated Self-Supervised Learning for One-Shot Cross-Modal and Cross-Imaging Technique Segmentation." arXiv预印本 arXiv:2503.23507 (2025). 哈佛格式： Manna, S., Das, S., Ghosh, S. and Bhattacharya, S., 2025. Federated Self-Supervised Learning for One-Shot Cross-Modal and Cross-Imaging Technique Segmentation. arXiv预印本 arXiv:2503.23507.