ARTIFICIAL INTELLIGENCE FOR SURGICAL PLANNING IN SPINAL PROCEDURES: PROGRESS OF THE AITPACOL PROJECT
Funded by the European Union – NextGenerationEU
With support from Spain’s Recovery, Transformation and Resilience Plan and the Community of Madrid
Digital Anatomics is developing new artificial intelligence (AI) tools to automate and improve spinal surgery planning, under the framework of the AITPACOL project. This initiative has been selected in the call for “Use Cases of Artificial Intelligence Applied to Industry”, promoted by the Recovery, Transformation and Resilience Plan (PRTR), and funded by the European Union – NextGenerationEU, with additional support from the Community of Madrid.
AITPACOL addresses a clinical need: improving precision, reducing surgical planning times, and enabling large-scale production of personalized medical devices such as spinal surgical guides.
CURRENT CHALLENGES IN PERSONALIZED SURGICAL PLANNING
Today, the creation of patient-specific devices for spinal surgery requires highly specialized manual planning, which may take several hours per case. This process relies on expert analysis of CT scans, introducing variability and limiting scalability.
The exclusive use of CT imaging also implies high radiation exposure, which may not be ideal for pediatric patients or repeated procedures. While X-rays are more accessible, they generally cannot be used directly for surgical planning due to a lack of adequate segmentation and modeling tools.
AI-BASED TECHNOLOGICAL DEVELOPMENTS
AITPACOL brings together five AI systems within a single cloud-based planning platform:
1. Automatic vertebra segmentation from CT scans:
A neural network automatically identifies and segments vertebrae from CT images, generating a 3D environment in seconds where surgery can be planned.
2. Automatic screw trajectory prediction:
A system that, based on the CT scan, suggests the optimal implant positions. These proposals are then reviewed by a technician, accelerating and facilitating the planning process.
3. Conversion of CT into synthetic X-ray images:
Enables the training of systems that operate directly on standard radiographs, which are more cost-effective and accessible, while also reducing patient radiation exposure.
4. Pedicle segmentation in radiographs:
A neural network that helps extract key measurements to select the correct screw size, which is especially important in pediatric surgeries.
5. 3D models from radiographs:
An algorithm capable of generating three-dimensional models without requiring a CT scan, enabling surgical planning with lower radiation exposure.
AUTOMATION, SCALABILITY AND PRECISION
The AITPACOL project is advancing toward a level of automation that minimizes the need for manual intervention and accelerates the design of personalized surgical tools. This enables more consistent outcomes and scalable production of custom-made devices with reduced time and cost.
Additionally, the project seeks to bridge CT and X-ray image modalities, expanding the clinical applicability of AI tools across centers with different levels of technological infrastructure. This interoperability enhances planning flexibility and accessibility.
AITPACOL also focuses on optimizing existing algorithms to improve speed and accuracy, and on extending their use to complex clinical scenarios, including pediatric patients or cases where only X-ray imaging is available.
AITPACOL is currently under development and validation. Its goal is to explore how advanced AI can support surgical workflows, without replacing clinical judgment or human expertise. The solutions described are still being tested and are not yet in clinical use.