RSNA Lumbar Degenerative Imaging Spine Classification - RSNA/AI-Challenge-Data GitHub Wiki

RSNA assembled this dataset in 2024 for the RSNA Lumbar Spine Degenerative Classification AI Challenge (https://www.kaggle.com/competitions/rsna-2024-lumbar-spine-degenerative-classification/). The goal of this competition was to identify medical conditions affecting the lumbar spine in MRI scans. RSNA collected de-identified magnetic resonance imaging (MR) scans of the lumbar spine from twelve sites, totaling over 2,600 imaging studies.

Description

The dataset is contained in a Zip archive that includes both DICOM image files (.dcm) and a tabular annotation file (.csv).

License

You may access and use these de-identified imaging datasets and annotations (β€œthe data”) for non-commercial purposes only, including academic research and education, as long as you agree to abide by the following provisions: Not to make any attempt to identify or contact any individual(s) who may be the subjects of the data.

Tutorial

Files

train.csv
Labels for the training set. Each row corresponds to a study and contains severity labels for multiple conditions and spine levels.

train_label_coordinates.csv
Coordinates and metadata for annotated labels in the training set.

sample_submission.csv
A template for submission, with one row per prediction.

[train/test]_images/[study_id]/[series_id]/[instance_number].dcm
The DICOM images. Each study may include multiple series, and each series is a stack of images ordered by instance number.

[train/test]_series_descriptions.csv
Series-level metadata for each study, including descriptive scan names and orientations.

Data Fields

train.csv

  • study_id – ID of the study. Each study may contain multiple image series.
  • [condition]_[level] – Target labels for each condition and spine level (e.g., spinal_canal_stenosis_l1_l2). The values represent severity: normal_mild, moderate, or severe. Some entries may be missing.

train_label_coordinates.csv

  • study_id – Study ID associated with the image.
  • series_id – The specific series ID within the study.
  • instance_number – The image’s index within the 3D image stack.
  • condition – One of the three conditions:
    • spinal_canal_stenosis
    • neural_foraminal_narrowing (left/right)
    • subarticular_stenosis (left/right)
  • level – The vertebral level involved (e.g., l3_l4).
  • x, y – Pixel coordinates representing the center of the labeled area.

sample_submission.csv

  • row_id – A composite ID of the format [study_id]_[condition]_[level] (e.g., 12345_spinal_canal_stenosis_l3_l4).
  • normal_mild, moderate, severe – The prediction columns, containing probabilities for each severity level.

[train/test]_series_descriptions.csv

  • study_id – The ID of the study.
  • series_id – Unique ID of the image series.
  • series_description – A string describing the scan orientation or protocol.

Download

Medical Imaging Resource for AI (MIRA)