Abstract
<title>Abstract</title> <p>Objective Intraoperative brain shift impairs preoperative MRI-based navigation, while intraoperative MRI (iMRI) is costly. This study explored a deformation field prediction method using 3D intraoperative ultrasound (3D iUS) and optical navigation to generate virtual MRI (vMRI) via non-rigid registration, providing a low-cost, high-precision guidance scheme for glioma resection. Methods An optical-ultrasound navigation platform unified 3D iUS and preoperative MRI coordinates. In vitro brain shift phantoms and ex vivo cadaveric specimens simulated bidirectional tumor growth-resection deformation. A B-spline free-form deformation algorithm computed tissue displacement fields from 3D iUS to generate vMRI. Target registration error (TRE) between vMRI and iMRI was calculated using implanted markers and anatomical landmarks. Algorithm performance was assessed by accuracy, real-time capability, and robustness. Results After deformation correction, mean TRE was 2.94 mm in vitro and 2.28 mm ex vivo, both within the clinically acceptable range (2–3 mm). Image overlay confirmed good alignment of key structures. The algorithm ran stably with a mean single-registration time of approximately 2.5 minutes, demonstrating acceptable near-real-time performance. Conclusion The "3D iUS + optical navigation + deformation field" pipeline is feasible in vitro and ex vivo, achieving clinically acceptable registration accuracy. These findings provide a methodological foundation for future in vivo validation and iMRI-free dynamic image-guided navigation.</p>