NA-MIC Project Weeks

Back to Projects List

Nephrostomy tutor low cost training system for percutaneous nephrostomy

Key Investigators

• Rebecca Hisey (Queen's University, Canada)
• Gabriella d'Albenzio (Queen's University, Canada)
• Kyle Sunderland (Queen's University, Canada)
• Mamadou Camara (ESP, Senegal)
• Gabor Fichtinger (Queen's University, Canada)

Project Description

Working to integrate recent developments on AI-based volume reconstruction from US into the existing Nephrostomy tutor system to improve guidance for trainees.

Objective

1. Integrate AI-based volume reconstruction from US segmentations
2. Test updated system on a low-cost phantom

Approach and Plan

1. AI Model Integration

• Review & Select AI Model: Identify and evaluate recent AI-based models for US segmentation that meet the needs of the Nephrostomy tutor system.
• Model Integration: Develop an interface to integrate the selected AI model with the existing tutor system, ensuring seamless data flow from ultrasound segmentations to the volume reconstruction process.

1. System Update and Refinement

• System Architecture Modification: Modify the Nephrostomy tutor system architecture to incorporate real-time volume reconstruction from the AI model.
• User Interface Adjustments: Update the user interface to display reconstructed volumes alongside the current guidance information for trainees.

Progress and Next Steps

1. Low-cost phantom created.
2. Integrated real-time UNet predictions into Nephrostomy tutor
3. Updated visualizations to show predictions
4. Initial ultrasound segmentations in progress.

Next steps:

1. Collect and segment more ultrasound kidney scans
2. Train an effective model

Illustrations

https://github.com/user-attachments/assets/74945fc2-c866-4739-8a3b-3a295c87484c

Background and References

Feasibility of video‐based skill assessment for percutaneous nephrostomy training in Senegal