Machine Learning in Ultrasound Computer-Aided Diagnostic Systems: A Survey

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Diagnostic Systems

Background:

• Ultrasound imaging is a widely used, safe, and cost-effective diagnostic tool.
• Interpreting ultrasound images can be challenging, necessitating advanced diagnostic support.
• Computer-aided diagnosis (CAD) systems aim to assist clinicians and reduce diagnostic workload.

Purpose of the Study:

• To review and categorize recent research on machine learning-based ultrasound CAD systems.
• To differentiate between traditional CAD systems using engineered features and deep learning CAD systems.
• To provide an overview of features, classifiers, and applications in both traditional and deep learning ultrasound CAD.

Main Methods:

• Literature review and categorization of ultrasound CAD systems.
• Analysis of traditional CAD systems based on handcrafted features and classifiers.
• Summary of recent deep learning applications in ultrasound CAD, including image classification and segmentation.

Main Results:

• Ultrasound CAD systems are broadly classified into traditional (manual features) and deep learning (automated features) approaches.
• Traditional systems rely on engineered features and specific classifiers for disease detection.
• Deep learning models show significant potential for performance improvement in ultrasound image analysis.

Conclusions:

• Machine learning, particularly deep learning, offers substantial advancements for ultrasound CAD systems.
• Understanding the evolution from traditional to deep learning methods is crucial for future research.
• This review serves as a valuable resource for researchers in the field of ultrasound CAD.