Taylor658/stereotactic-radiosurgery-k1-with-segmentation

🎯 Stereotactic Radiosurgery Dataset (SRS)

🏥 400 synthetic patient records describing the clinical, imaging, segmentation, and treatment-planning metadata of a stereotactic radiosurgery workflow, delivered as a single CSV with placeholder file paths.

⚠️ Disclaimer: This is a metadata-only synthetic dataset. It contains no real patients, no image files, and no segmentation files. Every record is generated; paths in the imaging and segmentation columns are placeholders that do not resolve to any file. Not for clinical use, treatment planning, or any patient-facing decision.

📋 At a Glance

📝 Overview

The dataset models the shape of an SRS data pipeline, from patient intake through imaging, target delineation, and plan parameters, so that tooling and models can be prototyped against a realistic schema before access to protected clinical data is arranged. It is intended for research and development of AI methods in SRS, not as a source of clinical ground truth.

📊 Dataset Summary

💡 Fields

🏥 Clinical

• —Patient demographics: age, sex, weight, height (synthetic)
• —Medical history: comorbidities, previous treatments (synthetic labels)
• —Tumor details: histology, grade, and stage (synthetic labels)

🧠 Imaging Metadata

• —Modality: CT, MRI, or PET
• —Protocol: contrast-enhanced or non-contrast
• —Scanner descriptors: manufacturer and model (synthetic labels)
• —Voxel geometry: isotropic or anisotropic descriptors

🎯 Segmentation Metadata

• —Gross Tumor Volume (GTV), Clinical Target Volume (CTV), Planning Target Volume (PTV)
• —Placeholder NIfTI-style paths (e.g., /simulated/path/SIM-0001_GTV_segmentation.nii)
• —Synthetic inter-observer variability descriptor

⚙️ Treatment Plan Metadata

• —Beam arrangement label, prescribed dose (Gy), DVH summary fields
• —Optimization objective labels (target coverage, organ-at-risk sparing)

🔍 Example Record

📌 Note: The SIM- prefix on every identifier marks the record as synthetic. Column names above are illustrative; see the CSV header for the exact schema.

🚀 Usage

Load the Dataset

from datasets import load_dataset

dataset = load_dataset("Taylor658/stereotactic-radiosurgery-k1-with-segmentation", split="train")
print(dataset[0])

🎯 Intended Usage

• —🧪 Pipeline prototyping: build and test ingestion, validation, and cohort-selection code against an SRS-shaped schema before connecting to a protected clinical source
• —🏷️ Tabular modeling exercises: classification or regression on synthetic metadata fields (e.g., beam arrangement from histology and stage) to develop and debug training code
• —📐 Schema design: a starting point for defining the metadata a real SRS dataset should carry

🚫 Out of Scope

• —Training or validating any model intended for clinical contouring, dose prediction, or treatment optimization; no image or dose data is present
• —Patient outcome prediction; outcomes in the file are synthetic labels with no relationship to real treatment response
• —Any use that treats the records as representing real people

⚠️ Limitations

🧬 Synthetic Content

• —Every field is generated; there is no underlying patient, scan, contour, or plan
• —Distributions of age, histology, dose, and other fields are template choices, not epidemiological or clinical distributions

📁 No Imaging or Segmentation Files

• —Path columns are placeholders and do not resolve
• —The dataset cannot support image-based tasks

⚙️ Plan Fields Are Labels

• —Dose, beam, and DVH fields are summary values, not the output of a treatment planning system
• —No dose distribution, structure set, or plan file is included

🏥 Not for Clinical Use

• —Nothing in this dataset should inform patient care, planning, or QA

📜 License

Released under CC BY-NC 4.0. You may share and adapt the material for non-commercial purposes with attribution.

🙌 Citation

@misc{srs_synthetic_metadata_2026,
  title  = {Stereotactic Radiosurgery Dataset (SRS): Synthetic Metadata},
  author = {Taylor, A.},
  year   = {2026},
  howpublished = {\url{https://huggingface.co/datasets/Taylor658/stereotactic-radiosurgery-k1-with-segmentation}},
  note   = {Metadata-only synthetic dataset with placeholder imaging and segmentation paths.}
}

🧑‍💻 Contributing

• —💬 Open a Discussion to propose additional metadata fields, histology categories, or plan descriptors

👨‍🚀 Author: A Taylor · 🤗 Hugging Face: hf.co/Taylor658 · 🐙 GitHub: ATaylorAerospace