Attention-Based Point Cloud Completion for CT-Equivalent Kidney Volumetry from Tracked Freehand Ultrasound

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Radiology

Background:

• Accurate kidney volume measurement is vital for managing conditions like polycystic kidney disease and monitoring kidney transplants.
• Computed tomography (CT) provides accurate kidney volumes but involves ionizing radiation.
• Conventional ultrasound methods for kidney volumetry have significant error rates (20-30%).

Purpose of the Study:

• To validate tracked freehand ultrasound with attention-based point cloud completion for kidney volumetry.
• To determine if this novel ultrasound method is equivalent to CT within a ±2% margin.
• To assess the accuracy of different scanning maneuvers for kidney volume measurement.

Main Methods:

• Sixty healthy volunteers underwent both electromagnetic-tracked freehand ultrasound and same-day CT scans.
• Sparse ultrasound point clouds were completed using an attention-based transformer network (PointAttN) and benchmark architectures.
• Kidney volumes were compared against CT using equivalence testing on per-kidney relative differences within a ±2% margin.

Main Results:

• Tracked freehand ultrasound in multiaxis merged mode achieved a mean absolute error of 3.94 mL (3.0% relative error).
• This method passed statistical equivalence testing (TOST) against CT within the predefined ±2% margin.
• Single-axis scanning failed equivalence testing, showing 2-4 times higher errors compared to multiaxis scanning.

Conclusions:

• Tracked freehand ultrasound with attention-based point cloud completion demonstrates accuracy comparable to CT for kidney volume measurement in healthy volunteers.
• The study validates the necessity of multiaxis scanning for achieving clinical accuracy in kidney volumetry.
• Further validation in patients with renal pathology and comparison with radiologist assessments are required before clinical deployment.