--- a
+++ b/README.md
@@ -0,0 +1,23 @@
+<h1>About Dataset</h1>
+
+<p>
+This dataset contains metadata extracted from training DICOM image files used in the <strong>RSNA 2022 Cervical Spine Fracture Detection</strong> competition.
+</p>
+
+<h2>Files Included</h2>
+<ul>
+  <li><strong>meta-train:</strong> Original extracted metadata (use with caution)</li>
+  <li><strong>meta-train-clean:</strong> Cleaned version of <code>meta-train</code>, easier to use</li>
+  <li><strong>meta-segmentations:</strong> Metadata for images with segmentations, includes correct vertebra labels (C1–C7) extracted from segmentation masks</li>
+  <li><strong>meta-segmentation-clean:</strong> Cleaned version of <code>meta-segmentations</code></li>
+  <li><strong>meta-train-with-vertebrae:</strong> Metadata for all training images, includes 88% accurate <em>Random Forest</em> predictions of the vertebra in each image</li>
+  <li><strong>train-segmented:</strong> Metadata for all train images with 95% accurate <em>EffNetV2</em> vertebrae predictions (based on provided notebook)</li>
+  <li><strong>train-vert-fold4:</strong> Version of <code>train-segmented</code> cleaned and enriched via an image+tabular model; includes extra feature-engineered columns</li>
+  <li><strong>train-vert:</strong> Final ensembled predictions from <code>train-segmented</code> and <code>train-vert-fold4</code></li>
+</ul>
+
+<h2>Related Notebooks</h2>
+<ul>
+  <li><strong>RSNA Fracture Detection – In-depth EDA</strong></li>
+  <li><strong>Extracting Vertebrae C1, ..., C7</strong></li>
+</ul>