Preoperative MRI-Guided Freehand Ultrasound Volume Reconstruction.

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Prostate Cancer Research

Background:

• Transrectal ultrasound (TRUS) is crucial for guiding prostate biopsies, but lacks sensitivity for cancer detection.
• Preoperative magnetic resonance imaging (MRI) offers superior soft tissue contrast, essential for accurate diagnosis.
• Accurate 3D reconstruction of freehand TRUS is vital for MRI-ultrasound fusion in interventions.

Purpose of the Study:

• To develop a novel framework for direct 3D ultrasound volume reconstruction from 2D frames using preoperative MRI.
• To overcome limitations of sequential estimation methods, such as accumulated errors and drift.
• To enhance the accuracy and interpretability of MRI-ultrasound fusion for prostate interventions.

Main Methods:

• Leveraging preoperative MRI and supervised contrastive learning for 3D ultrasound reconstruction.
• Aligning ultrasound images with MRI slices based on anatomical similarity to bypass sequential estimation.
• Training and validation on a large clinical dataset of over 500 prostate biopsy cases.

Main Results:

• Demonstrated over 50% improvement in drifting errors compared to existing methods.
• Achieved enhanced precision and interpretability in 3D ultrasound reconstruction.
• Successfully bypassed sequential estimation, avoiding drift and improving tracking accuracy.

Conclusions:

• The proposed framework enables reliable MRI-ultrasound fusion by enhancing 3D ultrasound reconstruction accuracy.
• The method holds significant potential for improving the diagnostic accuracy of prostate cancer interventions.
• Direct 3D reconstruction from 2D frames offers a more robust alternative to traditional methods.