Explainable AI in breast cancer ultrasound imaging: current developments and challenges

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Breast cancer is a leading cause of mortality globally, necessitating accurate and early diagnosis.
• Ultrasound imaging is a safe, accessible, and cost-effective tool for breast cancer diagnosis, especially in resource-limited settings.
• Deep learning models show promise in automating breast cancer detection and classification from ultrasound images, but their 'black-box' nature hinders clinical adoption.

Purpose of the Study:

• To review current developments in Explainable Artificial Intelligence (XAI) applied to breast cancer ultrasound imaging.
• To highlight XAI techniques that improve the interpretability and reliability of deep learning models in this domain.
• To identify obstacles and future research directions for integrating XAI into clinical practice.

Main Methods:

• Saliency-based approaches
• Model-agnostic methods
• Attention mechanisms

Main Results:

• XAI techniques offer insights into deep learning model decisions for breast cancer ultrasound analysis.
• Current XAI methods aim to address the 'black-box' problem in AI-driven medical diagnostics.
• The review summarizes advancements in making AI models more transparent and trustworthy.

Conclusions:

• XAI is crucial for bridging the gap between AI development and clinical implementation in breast cancer diagnostics.
• Standardized evaluation metrics, clinical validation, and handling noisy imaging data are key challenges for XAI in this field.
• Future research should focus on robust XAI solutions for reliable clinical application of AI in breast cancer ultrasound analysis.