Dark Mode
RSNA Cervical Spine Fracture Metadata
Patient Health Records & Digital Health
Tags and Keywords
Trusted By
"No reviews yet"
Free
About
This dataset provides cleaned metadata extracted from DICOM training images, specifically curated for the RSNA 2022 Cervical Spine Fracture Detection competition. It includes various versions of metadata, such as cleaned original metadata, metadata for images with segmentations (including correct C1-C7 labels), and metadata with accurate machine learning predictions for vertebrae presence. Notably,
meta-train-with-vertebrae offers 88% accurate Random Forest predictions, while train-segmented features 95% accurate EffNetV2 predictions. Further refined versions like train-vert-fold4 incorporate cleaned segmentations, an image-plus-tabular model, and additional feature-engineered columns, culminating in train-vert, which presents ensembled predictions from train-segmented and train-vert-fold4. The dataset aims to support the development and evaluation of models for identifying cervical spine fractures.Columns
- StudyInstanceUID: A unique identifier for each patient study.
- Slice: The numerical index of the image slice within a patient's study.
- ImageHeight: The height of the image in pixels, consistently 512 pixels.
- ImageWidth: The width of the image in pixels, consistently 512 pixels.
- SliceThickness: The thickness of the CT slice in millimetres, varying from 0.5mm to 1mm.
- ImagePositionPatient_x: The x-coordinate of the top-left corner of the image in patient-centred coordinates.
- ImagePositionPatient_y: The y-coordinate of the top-left corner of the image in patient-centred coordinates.
- ImagePositionPatient_z: The z-coordinate of the top-left corner of the image in patient-centred coordinates.
- C1: A binary indicator (0 or 1) denoting whether the C1 vertebra appears in the slice.
- C2: A binary indicator (0 or 1) denoting whether the C2 vertebra appears in the slice.
- C3: A binary indicator (0 or 1) denoting whether the C3 vertebra appears in the slice.
- C4: A binary indicator (0 or 1) denoting whether the C4 vertebra appears in the slice.
- C5: A binary indicator (0 or 1) denoting whether the C5 vertebra appears in the slice.
- C6: A binary indicator (0 or 1) denoting whether the C6 vertebra appears in the slice.
- C7: A binary indicator (0 or 1) denoting whether the C7 vertebra appears in the slice.
Distribution
The dataset is primarily available in a CSV file format, specifically exemplified by
meta_segmentation.csv. This particular file is approximately 2.5 MB in size and contains 15 columns. Across its various metadata files, the dataset consists of around 29,800 records.Usage
This dataset is ideally suited for developing, training, and evaluating machine learning models focused on cervical spine fracture detection. It can be utilised for:
- Creating predictive models for identifying the presence of specific vertebrae (C1-C7) in CT slices.
- Enhancing image segmentation tasks by providing cleaned metadata and associated labels.
- Feature engineering and model development, including the use of both image and tabular data.
- Benchmarking and improving diagnostic accuracy for spine-related medical conditions.
- Participating in and researching solutions for medical imaging competitions like RSNA 2022.
Coverage
The dataset focuses on metadata from medical imaging (DICOM files) related to cervical spine fracture detection. While specific geographic or demographic coverage is not detailed, it pertains to clinical training data for medical diagnostic purposes. The dataset is static, with an expected update frequency of "Never."
License
CC0: Public Domain
Who Can Use It
This dataset is intended for a range of users involved in medical imaging and data science:
- Machine Learning Engineers and Data Scientists: For building and refining AI models for medical diagnosis.
- Medical Researchers: To study spine-related conditions and improve diagnostic techniques.
- Healthcare AI Developers: To integrate advanced diagnostic capabilities into healthcare systems.
- Academics and Students: For research, education, and projects in medical image analysis and computer vision.
- Participants of Medical Competitions: Particularly those focused on radiology and fracture detection.
Dataset Name Suggestions
- RSNA Cervical Spine Fracture Metadata
- Spine CT Scan Fracture Detection Data
- Cervical Vertebrae Metadata for AI
- RSNA 2022 Spine Fracture Train Data
- Cleaned Spine Fracture Image Metadata
Attributes
Original Data Source: RSNA Cervical Spine Fracture Metadata
QUALITY