Recurrent neural networks for generalization towards the vessel geometry in autonomous endovascular guidewire navigation in the aortic arch - Scorecard - MDSpire

Recurrent neural networks for generalization towards the vessel geometry in autonomous endovascular guidewire navigation in the aortic arch

  • By

  • Lennart Karstensen

  • Jacqueline Ritter

  • Johannes Hatzl

  • Floris Ernst

  • Jens Langejürgen

  • Christian Uhl

  • Franziska Mathis-Ullrich

  • May 28, 2023

  • 0 min

Share

Clinical Scorecard: Utilizing Recurrent Neural Networks to Enhance Generalization of Vessel Geometry in Autonomous Navigation of Endovascular Guidewires within the Aortic Arch

At a Glance

CategoryDetail
ConditionVascular diseases requiring endovascular interventions such as heart attack and stroke
Key MechanismsAutonomous navigation of endovascular guidewires using recurrent neural networks to generalize across varying aortic arch geometries without explicit vessel geometry feedback
Target PopulationPatients undergoing endovascular interventions targeting supraaortal arteries within the aortic arch
Care SettingInterventional radiology or endovascular surgical suites with fluoroscopy imaging

Key Highlights

  • Endovascular navigation currently relies on manual manipulation of guidewires and catheters under fluoroscopy with contrast agents to visualize vessels.
  • Automation via learning-based controllers can reduce surgeon radiation exposure and cognitive load, and improve access in remote areas.
  • Recurrent neural networks improve generalization of autonomous navigation across different aortic arch geometries without requiring vessel geometry input.

Guideline-Based Recommendations

Diagnosis

  • Use fluoroscopy imaging with contrast agents to visualize vascular anatomy during endovascular interventions.

Management

  • Manually manipulate guidewire and catheter to navigate to lesion sites, using skillful rotation and insertion.
  • Consider autonomous navigation systems to reduce radiation exposure and surgeon strain where available.
  • Utilize recurrent neural network-based controllers trained on variable vessel geometries to improve navigation success.

Monitoring & Follow-up

  • Continuously monitor guidewire tip position via fluoroscopy imaging in left anterior oblique projection.
  • Minimize contrast agent use to reduce patient risk while maintaining adequate vessel visualization.

Risks

  • Radiation exposure to operating surgeon during manual navigation.
  • Health risks associated with excessive contrast agent use.
  • Reduced navigation success when controllers are trained on limited vessel geometries.

Patient & Prescribing Data

Patients undergoing endovascular interventions involving navigation through the aortic arch to supraaortal arteries

Autonomous navigation systems using recurrent neural networks can potentially improve procedural success and safety by adapting to patient-specific vessel geometries without requiring contrast agent-based imaging.

Clinical Best Practices

  • Employ fluoroscopy imaging with minimal contrast agent to balance vessel visualization and patient safety.
  • Use guidewire leading technique with skillful rotation and insertion to probe target arteries.
  • Adopt autonomous navigation technologies to reduce radiation exposure and cognitive load on surgeons.
  • Train autonomous controllers on diverse vessel geometries to enhance generalizability and real-world applicability.

References

Original Source(s)

Related Content