Opportunistic osteoporosis screening in intraoperative CT can accurately identify patients with low volumetric bone mineral density and osteoporosis during spine surgery

Sebastian Rühling, Moritz Roman Hernández Petzsche, Maximilian T. Löffler, Nico Sollmann, Thomas Baum, Jannis Bodden, Julian Schwarting, Nicole Lange, Amir Kaywan Aftahy, Maria Wostrack, Claus Zimmer, Jan S. KirschkeEuropean 2025 · DOI

Abstract

Opportunistic osteoporosis screening with the presented approach is feasible and demonstrates high accuracy in reference to preoperative MDCT scans. This could enable the identification of patients with low bone mass during surgery, allowing surgeons to take measures (e.g., adapted techniques) that prevent postoperative complications and improve patient outcomes.

Introduction: The Challenge of Bone Quality in Spine Surgery

In spine surgery, bone quality is a critical determinant of success. Undiagnosed osteoporosis significantly increases the risk of hardware failure, such as screw loosening or pull-out, which often necessitates painful and complex revision surgeries.

While Dual-energy X-ray Absorptiometry (DXA) remains the "gold standard" for bone density measurement, it faces practical limitations: it is frequently omitted from preoperative workups, results may be unavailable by the surgery date, and degenerative spinal changes can skew its accuracy. Preoperative CT scans offer an alternative through "opportunistic screening," but these scans are not always accessible or may not cover the specific surgical levels.

This study investigates whether intraoperative CT (iCT)—scans already routinely performed in the operating room for robotic or navigated guidance—can be leveraged as a reliable, real-time diagnostic tool for assessing bone mineral density (vBMD).

Intraoperative CT (iCT) setup in the operating room showing the scanner and patient positioning.

Methods: AI-Driven Bone Density Analysis

The researchers evaluated a retrospective cohort of 105 patients, comparing vBMD measurements from preoperative reference scans to those obtained during surgery. The core of this approach is a Convolutional Neural Network (CNN)-based framework designed to automate the diagnostic workflow:

  1. Automated Segmentation: The AI identifies and labels individual vertebrae (T1 through L5) and creates precise "masks" for analysis.
  2. Asynchronous Calibration: Standard CT scans measure Hounsfield Units (HU), which vary across different scanners. The framework utilizes asynchronous calibration to convert these values into standardized vBMD (mg/cm³) without requiring a physical calibration phantom during the scan.
  3. Contrast-Phase Correction: To account for intravenous contrast in preoperative scans—which can artificially inflate density readings—the CNN detects the contrast phase and applies linear regression to normalize the data.
Automated vertebral segmentation and masking (A and B) used by the CNN framework to isolate bone for density measurement.

Datasets and Study Population

The study analyzed data from 105 adult patients (mean age 73) undergoing surgery for trauma, tumors, or degenerative diseases.

  • Preoperative Scans: Collected from five different MDCT scanner models (Philips and Siemens).
  • Intraoperative Scans: Performed on a Philips Brilliance Big Bore scanner located within the operating suite.
  • Scope: A total of 1,110 vertebrae across the thoracic and lumbar spine were evaluated.

Key Insights: Real-Time Clinical Utility

  • "Free" Diagnostic Data: By repurposing existing iCT data, surgeons obtain a comprehensive bone quality report without additional radiation exposure or financial cost to the patient.
  • Intraoperative Decision Making: Real-time identification of low bone density allows surgeons to immediately adapt their strategy, such as employing cement-augmented screws or extending instrumentation to better distribute mechanical loads.
  • Reliability in the OR: Despite the complexities of the operating room environment—including unique patient positioning and the presence of surgical reference frames—the AI-driven measurements showed high consistency with controlled preoperative scans.
  • Addressing the Diagnostic Gap: Automated screening ensures that patients, particularly males who are less frequently screened for osteoporosis, do not miss out on critical bone health assessments.

Results: High Accuracy and Diagnostic Power

The intraoperative measurements demonstrated exceptional agreement with preoperative references:

  • Correlation: Strong associations were observed in both the thoracic ($R^2 = 0.94$) and lumbar spine ($R^2 = 0.96$).
  • Minimal Bias: The mean bias was remarkably low (-1.3 mg/cm³ for thoracic; -3.0 mg/cm³ for lumbar).
  • Diagnostic Accuracy: For identifying osteoporosis (vBMD < 80 mg/cm³), the intraoperative approach achieved 100% sensitivity and 95.3% specificity. Crucially, no patient was misclassified across the threshold from healthy to osteoporotic.

Limitations and Future Directions

While highly promising, the authors note several areas for further development:

  • Refining Automation: Currently, radiologists must manually exclude vertebrae with fractures or severe metal artifacts. Future iterations of the CNN could automate this filtering process.
  • Multi-Center Validation: As intraoperative scans were limited to one scanner model, broader studies are needed to confirm these results across different manufacturers.
  • Clinical Impact: Future prospective studies will be essential to demonstrate that intraoperative tactical changes based on this data directly lead to improved patient outcomes and reduced complication rates.

Summary updated

Abstract

Opportunistic osteoporosis screening with the presented approach is feasible and demonstrates high accuracy in reference to preoperative MDCT scans. This could enable the identification of patients with low bone mass during surgery, allowing surgeons to take measures (e.g., adapted techniques) that prevent postoperative complications and improve patient outcomes.
BibTeX
@inproceedings{ruhling2025opportunistic,
  author = {Sebastian Rühling and Moritz Roman Hernández Petzsche and Maximilian T. Löffler and Nico Sollmann and Thomas Baum and Jannis Bodden and Julian Schwarting and Nicole Lange and Amir Kaywan Aftahy and Maria Wostrack and Claus Zimmer and Jan S. Kirschke},
  title = {{Opportunistic osteoporosis screening in intraoperative CT can accurately identify patients with low volumetric bone mineral density and osteoporosis during spine surgery}},
  booktitle = {European},
  year = {2025},
  doi = {10.1007/s00586-025-08697-0},
}
Figure