An Open Annotated Dataset and Baseline Machine Learning Model for Segmentation of Vertebrae with Metastatic Bone Lesions from CT

Sep 21, 2026

Uploaded by Sage

Automatic analysis of pathologic vertebrae from computed tomography (CT) scans can improve diagnostic management for patients with metastatic spine disease. The work releases a first publicly available annotated dataset of cancerous CT spines, enabling development of AI for vertebra segmentation and classification. The dataset includes 55 CT scans across primary cancer types, with manual segmentations and contours, vertebral level labels, lesion-type categories, and patient demographics. A nnU-Net-based baseline model is publicly provided to support downstream prediction of mechanical response, fracture risk, and spinal deformity.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages8

Reading Time20 min

AccessFree to download

Abstract

Background & Summary

Motivation and clinical impact

Imaging limitations and need for quantitative assessment

Related deep learning progress and dataset gaps

Contributions of the new dataset

medRxiv preprint doi: [https://doi.org/10.1101/2024.10.14.24314447](https://doi.org/10.1101/2024.10.14.24314447); this version posted October 15, 2024. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license . An Open Annotated Dataset and Baseline Machine Learning Model for Segmentation of Vertebrae with Metastatic Bone Lesions from CT Nazim Haouchine2,3, David B. Hackney1,3, Steve D. Pieper4, William M. Wells III2,3, Malika Sanhinova1,3, Tracy A. Balboni2,5, Alexander Spektor2,5, Mai A. Huynh2,5, David E. Kozono2, Patrick Doyle2,5, and Ron N. Alkalay1,3 1. Beth Israel Deaconess Medical Center, Boston, MA 2. Brigham and Women's Hospital, Boston, MA 3. Harvard Medical School, Boston, MA 4. Isomics, Inc., Cambridge, MA 5. Dana-Farber Cancer Institute., Boston, MA Corresponding author(s): Ron N. Alkalay ([rn_alkalay@bidmc.harvard.edu](rn_alkalay@bidmc.harvard.edu)) Abstract Automatic analysis of pathologic vertebrae from computed tomography (CT) scans could significantly improve the diagnostic management of patients with metastatic spine disease. We provide the first publicly available annotated imaging dataset of cancerous CT spines to help develop artificial intelligence frameworks for automatic vertebrae segmentation and classification. This collection contains a dataset of 55 CT scans collected on patients with various types of primary cancers at two different institutions. In addition to raw images, data include manual segmentations and contours, vertebral level labeling, vertebral lesion-type classifications, and patient demographic details. Our automated segmentation model uses nnU-Net, a freely available open-source framework for deep learning in healthcare imaging, and is made publicly available. This data will facilitate the development and validation of models for predicting the mechanical response to loading and the resulting risk of fractures and spinal deformity. Background & Summary In 2022, it is estimated that there were more than 1.9 million new cancer cases in the United States (1) . After the liver and the lungs, bones are the third most frequent target site of metastases (1), with the spine being the most frequent site of bone metastases (2, 3) . Cancer patient autopsies (4) showed 73% of breast, 68% of prostate, and 35% of renal and lung subjects had vertebral bone metastases. Up to 50% of spinal metastases receive some form of treatment, and 5-10% receive surgical management (5, 6, 7) . Patients are at high risk for skeletal adverse events, including bone pain, fracture, hypercalcemia, and neurological dysfunction secondary to the spinal cord or cauda equina compression (4, 9, 10, 11, 12), that may require palliative radiotherapy, surgery, or both. Skeletal adverse events profoundly and negatively impact patients' lives (13), causing disability and often requiring high doses of narcotics, causing reduced quality of life (14) . They also increase psychological distress, decrease social functioning (15), and significantly impair normal function (16) . Although the impact of spinal metastasis on the vertebral column anatomical integrity is readily imaged with MRI and CT (17, 18, 19, 20, 21, 22, 23), in the absence of quantitative assessment (11, 24, 25, 26), standard-of-care imaging modalities (19, 27, 28, 29) do not provide precise estimates of the structural competence of the spine (17, 18, 19, 20, 21, 22, 23) . This limited ability has a profound, negative impact on these patients' clinical care. Clinicians are forced to decide without objective quantitative criteria whether surgery or vertebral augmentation should be offered or withheld in attempts to treat to reduce the risk of fracture or to control pain. This can be particularly tragic when treatment could have been offered prior to the devel

FAQs

What dataset is being provided for metastatic spine disease research?

A publicly available annotated imaging dataset of cancerous CT spines, including raw CT images and detailed annotations for vertebrae and vertebral lesions.

What annotations and labels are included in the dataset?

The collection includes manual segmentations and contours, vertebral level labeling, vertebral lesion-type classifications, and patient demographic details.

What baseline machine learning model is included with the release?

An automated segmentation model based on nnU-Net, an open-source deep learning framework for healthcare imaging, is made publicly available.

Sep 21, 2026

Uploaded by Sage