Enhanced optical flow field of left ventricular motion using quasi-Gaussian DCT filter

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Computational Cardiology

Background:

• Accurate left ventricular motion estimation is crucial for diagnosing cardiac abnormalities.
• Optical flow techniques are widely used for motion quantification but struggle with cardiac motion complexity and speckle noise.
• Existing methods often yield non-smooth vector fields, limiting diagnostic accuracy.

Purpose of the Study:

• To introduce a novel quasi-Gaussian discrete cosine transform (QGDCT)-based filter for enhancing optical flow fields.
• To improve myocardial motion estimation by reducing speckle noise and enhancing flow field smoothness.
• To validate the QGDCT filter's performance on both synthetic and clinical echocardiography data.

Main Methods:

• A QGDCT filter was developed, utilizing a customized quasi discrete Gaussian filter with coefficients derived from a 2D DCT.
• The QGDCT filter was applied both before and after optical flow computation to address speckle noise and improve smoothness.
• The algorithm was validated using synthetic echocardiography data simulating myocardial motion and subsequently on clinical echocardiography images.

Main Results:

• Quantitative analysis using magnitude error, angular error, and standard error of measurement demonstrated the filter's effectiveness.
• The QGDCT filter successfully reduced speckle noise and improved the smoothness of the optical flow vector field.
• Motion estimation results obtained using the QGDCT filter showed strong agreement with manual interpretations by physicians.

Conclusions:

• The proposed QGDCT-based filter is an effective tool for enhancing optical flow-based myocardial motion estimation.
• This technique offers a significant improvement over traditional optical flow methods by addressing noise and smoothness issues.
• The QGDCT filter shows promise for improving the accuracy and reliability of cardiac abnormality diagnosis through enhanced motion analysis.