A Heart Segmentation Algorithm Based on Dynamic Ultrasound

Area of Science:

• Cardiology
• Medical Imaging
• Signal Processing

Background:

• The heart is vital for blood circulation to organs and tissues.
• Cardiac ultrasound imaging is a primary noninvasive method for assessing heart dynamics.
• Understanding heart movement patterns is crucial for clinical diagnosis.

Purpose of the Study:

• To develop an advanced algorithm for accurate cardiac segmentation from ultrasound images.
• To improve the analysis of heart dynamics using signal processing techniques.
• To enhance diagnostic capabilities for medical professionals.

Main Methods:

• Analysis of cardiac ultrasound image characteristics from medical and signal processing perspectives.
• Decomposition of image signals into high- and low-frequency components.
• Application of an attention model for focusing on the heart region.
• Establishment of a multidimensional network carrying model for segmentation.

Main Results:

• The proposed algorithm achieves an Average Overlap Measure (AOM) of 92% for cardiac segmentation.
• The method effectively highlights dimensional information through signal decomposition.
• The attention model successfully focuses on critical heart regions.

Conclusions:

• The developed algorithm demonstrates significant advancement in cardiac segmentation accuracy.
• This technique can serve as a valuable tool to assist doctors in making accurate diagnoses.
• Further research can build upon this model for enhanced cardiovascular diagnostics.