Classification of Microcalcification Clusters Using Bilateral Features Based on Graph Convolutional Network.

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Breast cancer remains a leading cause of mortality in women globally.
• Clustered microcalcifications are a key indicator of breast cancer, necessitating accurate classification.
• Effective classification of microcalcifications is vital for early breast cancer detection and management.

Purpose of the Study:

• To develop and evaluate a novel computational model for classifying clustered microcalcifications in mammograms.
• To improve the accuracy of breast cancer diagnosis by enhancing the analysis of microcalcification patterns.
• To integrate image-derived features with spatial distribution characteristics for a comprehensive classification approach.

Main Methods:

• Utilized a discriminant model based on image convolution to extract image features.
• Employed a graph convolutional network (GCN) on a topological graph to analyze spatial distribution.
• Developed a fusion model combining convolutional and graph convolutional networks for complementary information integration.

Main Results:

• The fusion model demonstrated superior performance in classifying clustered microcalcifications.
• The combined approach outperformed individual image-based and spatial-based classification models.
• The study highlights the benefit of integrating diverse data representations for improved diagnostic accuracy.

Conclusions:

• A fusion model integrating image and spatial information offers a powerful approach for clustered microcalcification classification.
• This method shows significant potential for improving breast cancer diagnosis accuracy.
• Further research can explore this fusion technique for other medical image analysis tasks.