ProtoMGNet: Memory Guided Network With Instance Prototype Enhancement for Echocardiography Segmentation.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Cardiology

Background:

• Accurate segmentation of cardiac structures in echocardiography is vital for diagnosing cardiovascular diseases.
• Current methods struggle with imaging limitations like low signal-to-noise ratios and speckle noise.
• Cardiologists excel by integrating prior knowledge with visual cues, a mechanism inspiring new AI approaches.

Purpose of the Study:

• To develop an AI model, ProtoMGNet, that mimics human cognitive memory reconstruction for improved cardiac segmentation.
• To leverage experiential memory to enhance the accuracy of segmenting cardiac structures in echocardiograms.

Main Methods:

• Proposed ProtoMGNet model incorporating Prototype Enhanced Memory Reconstructor (PEMR), Texture Feature Mixer (TFM), and Frequency Domain Edge Filter (FEF).
• PEMR dynamically updates memory units with dataset-level class distribution and uses predicted masks for weighted aggregation.
• TFM and FEF refine segmentation boundaries by mixing texture and edge features.

Main Results:

• ProtoMGNet demonstrated superior performance compared to 13 state-of-the-art methods on CAMUS and CardiacUDA datasets.
• The model effectively addresses challenges posed by low signal-to-noise ratios and viewpoint variations in echocardiography.
• Experimental results highlight the potential of ProtoMGNet as a clinical auxiliary tool for accurate cardiac segmentation.

Conclusions:

• The proposed ProtoMGNet effectively utilizes a cognitive memory reconstruction mechanism for enhanced cardiac segmentation.
• The model shows significant potential to assist clinicians in the early diagnosis of cardiovascular diseases.
• Further development could integrate this AI tool into clinical workflows for improved diagnostic accuracy.