Computational shape analysis of three pelvic structures for sex estimation: de-identified CT-derived boundary coordinates and analysis code

<p>This archive contains the complete analysis code and the de-identified datasupporting a study of sex estimation from three pelvic structures &mdash; the<strong>pelvic inlet (PI)</strong>, the <strong>subpubic angle (SPA)</strong>, and the<strong>greater sciatic notch (GSN)</strong> &mdash; using computational shape analysisof boundary coordinates derived from computed tomography.</p> <p>The pipeline extracts millimetre-scale boundary coordinates from images of the threeisolated structures, computes a large set of size and shape descriptors (classicalmorphometrics, ellipse and partial-conic fitting, polygon approximation, elliptic Fourieranalysis, curvature profiling, and Procrustes-based density heatmaps), performs statisticalcomparison and sex classification, and reports the results in an evaluativelikelihood-ratio framework.</p> <h3>Dataset summary</h3><table border="1" cellpadding="4" cellspacing="0"> <tr><td>Specimens</td><td>351 (164 male, 187 female)</td></tr> <tr><td>Structures per specimen</td><td>3 (pelvic inlet, subpubic angle, greater sciatic notch)</td></tr> <tr><td>Coordinate files</td><td>1053</td></tr> <tr><td>Units</td><td>millimetres</td></tr> <tr><td>Point density</td><td>0.5 points per mm of boundary length</td></tr> <tr><td>Points per contour (min / median / max)</td><td>PI 154 / 198 / 255 &middot; SPA 37 / 52 / 77 &middot; GSN 25 / 35 / 47</td></tr></table> <h3>Contents</h3><ul> <li><code>code/</code> &mdash; six Python scripts covering the full workflow: building the pixel-to-millimetre scale table, extracting boundary coordinates from images, arranging the de-identified data for analysis, the main feature-extraction and classification pipeline, the score-based likelihood-ratio analysis, and an optional stand-alone re-run of the hold-out validation.</li> <li><code>data/coordinates/</code> &mdash; one CSV per specimen per structure (<code>&lt;anon_id&gt;_&lt;STRUCTURE&gt;.csv</code>) with columns <code>i</code>, <code>u_mm</code>, <code>v_mm</code>.</li> <li><code>data/sex_labels.csv</code> &mdash; anonymous specimen code and sex.</li> <li><code>README.md</code>, <code>data/README.md</code> &mdash; reproduction instructions and a data dictionary.</li></ul> <h3>De-identification</h3><p>Each specimen is identified only by an anonymous code (e.g. <code>S001</code>) that encodesneither the original patient identity, their sex, nor their acquisition order; specimen orderwas randomised before codes were assigned. No patient-identifying information is included, andthe mapping between codes and original records is not part of this archive.</p> <h3>Licensing</h3><p>Code in <code>code/</code> is released under the MIT License; data in <code>data/</code>are released under CC BY 4.0. See the <code>LICENSE</code> file for the full terms.</p>