Performance evaluation of computer-aided automated master frame selection techniques for fetal echocardiography

Area of Science:

• Medical Imaging
• Cardiology
• Fetal Medicine

Background:

• Fetal echocardiography is crucial for assessing fetal heart development and detecting congenital heart disease (CHD).
• Current methods rely on manual selection of the diastole frame for cardiac parameter measurement, which is subject to sonographer expertise and prone to errors.
• Automated techniques are needed to enhance the reliability and accuracy of fetal echocardiography analysis.

Purpose of the Study:

• To propose and evaluate automated techniques for selecting the optimal 'Master Frame' from fetal echocardiography cine loops for cardiac parameter measurement.
• To overcome the limitations of manual diastole frame selection, reducing intra- and inter-observer variability.
• To validate the proposed automated methods against expert annotations using established fidelity metrics.

Main Methods:

• Three automated methods were developed: Frame Similarity Measures (FSM), Averaging Mid-Frames (AMF), and Averaging All Frames (AAF).
• FSM utilizes correlation, SSIM, PSNR, and MSE to identify the cardiac cycle and generate a master frame by superimposing frames.
• The performance of each method was assessed using Dice coefficient, Jaccard ratio, Hausdorff distance, SSIM, MAE, and Pratt figure of merit, comparing automated master frames with expert-annotated diastole frames.

Main Results:

• The FSM-based method demonstrated a close match with manually selected diastole frames, indicating high accuracy and automatic cardiac cycle detection.
• AMF produced master frames similar to diastole frames but resulted in reduced chamber sizes, potentially leading to inaccurate measurements.
• AAF did not yield master frames identical to the clinical diastole frames.

Conclusions:

• The FSM-based master frame selection is a suitable automated technique for clinical integration in fetal echocardiography.
• This automated approach eliminates manual intervention, enhancing consistency and reducing errors in cardiac chamber measurements.
• Fidelity metric assessments confirm the FSM method's suitability for automated fetal chamber recognition and CHD detection.