Ultrasound fusion image error correction using subject-specific liver motion model and automatic image registration

Area of Science:

• Medical Imaging
• Image-guided therapy
• Computational anatomy

Background:

• Ultrasound fusion imaging aids in diagnosing and treating liver conditions like hepatocellular carcinoma.
• Respiratory liver motion causes misalignment (fusion error) in multimodal images, compromising clinical effectiveness.
• Accurate image registration is crucial for reliable ultrasound fusion imaging.

Purpose of the Study:

• To develop a subject-specific liver motion model and automatic registration method to correct fusion errors.
• To enhance the accuracy and practicality of ultrasound fusion imaging in clinical settings.
• To mitigate the impact of respiratory motion on image fusion quality.

Main Methods:

• Developed an online, subject-specific liver motion model for 2D ultrasound and 3D magnetic resonance (MR) images.
• Implemented an automatic registration method to compensate for respiratory liver motion during fusion.
• Validated the approach using a liver phantom and five human subjects.

Main Results:

• Phantom study: Fusion error reduced from 13.90±2.38mm to 0.63±0.53mm; rotation error decreased from 7.06±0.21° to 1.18±0.66°.
• Clinical study: Fusion error reduced from 12.90±9.58mm to 1.96±0.33mm.
• The combined motion model and registration method significantly improved fusion accuracy.

Conclusions:

• The proposed method effectively corrects respiration-induced fusion errors, enhancing ultrasound fusion imaging quality.
• This approach reduces reliance on initial image registration accuracy.
• The method improves the clinical practicability of ultrasound fusion imaging for liver applications.