Hepatocellular Carcinoma Automatic Diagnosis within CEUS and B-Mode Ultrasound Images Using Advanced Machine Learning...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Hepatocellular Carcinoma (HCC) is the most prevalent malignant liver tumor, typically developing in cirrhotic liver tissue.
• Current gold-standard diagnosis via needle biopsy is invasive and carries significant risks.
• There is a critical need for non-invasive diagnostic methods for HCC.

Purpose of the Study:

• To develop and assess non-invasive, computerized techniques for HCC diagnosis using ultrasound imaging.
• To evaluate the efficacy of machine learning, specifically Convolutional Neural Networks (CNNs), for automatic HCC detection in B-mode and Contrast-Enhanced Ultrasound (CEUS) images.

Main Methods:

• Utilized Convolutional Neural Networks (CNNs) for automatic HCC diagnosis on B-mode and CEUS images.
• Explored various data fusion techniques, including feature-level and decision-level fusion of multimodal ultrasound data.
• Investigated Kernel Principal Component Analysis (KPCA) for dimensionality reduction and compared CNN performance with traditional methods.

Main Results:

• Achieved diagnostic accuracy exceeding 97% using the developed AI methodology.
• Demonstrated superior performance of combined B-mode and CEUS image analysis compared to using individual modalities.
• The proposed CNN-based approach outperformed conventional classification methods and state-of-the-art techniques.

Conclusions:

• AI-driven analysis of ultrasound images, particularly when combining B-mode and CEUS data, offers a highly accurate non-invasive method for HCC diagnosis.
• This advanced machine learning approach provides a promising and safer alternative to invasive liver biopsy procedures.
• The developed methodology significantly advances the potential for early and reliable detection of Hepatocellular Carcinoma.