Spatial Calibration for Swept Synthetic Aperture Imaging Using Differentiable Beamforming

Ultrasound imaging is valued for its affordability, non-invasiveness, and ease of use but is fundamentally limited in lateral resolution by diffraction, especially at depth. Previously, we addressed this limitation using freehand swept synthetic aperture (SSA) imaging with optical tracking (OT-SSA), which relies on a calibration matrix to transform optical tracking data into the transducer’s coordinate system for accurate beamforming. Here, we focus on optimizing the calibration matrix to enhance image quality. We propose a differentiable beamforming approach utilizing PyTorch’s automatic differentiation to refine the calibration matrix parameters. Our optimization process improves the definition of targets throughout the field of view and demonstrates repeatability for relevant degrees of freedom.