Artificial intelligence in bone metastasis analysis: Current advancements, opportunities and challenges

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Oncology
• Radiology Workflow Optimization

Background:

• Artificial intelligence (AI) is revolutionizing medical imaging, especially for detecting bone metastases (BM), a critical issue in advanced cancers.
• Machine learning and deep learning offer advanced capabilities for BM detection, recognition, and segmentation, but face challenges like data scarcity and interpretability.

Purpose of the Study:

• To review artificial intelligence (AI) methods and applications for bone metastasis (BM) analysis across various medical imaging modalities.
• To examine the impact of datasets on AI development for BM analysis.

Main Methods:

• A PRISMA-guided review of AI techniques for bone metastasis analysis in CT, MRI, PET, SPECT, and bone scintigraphy.
• Inclusion of traditional machine learning and deep learning models, including Convolutional Neural Networks (CNNs) and transformers.
• Assessment of existing datasets and their influence on AI development in this field.

Main Results:

• AI models, particularly CNNs and transformers, demonstrate high performance in BM analysis across different imaging modalities and tasks.
• Persistent limitations include data imbalance, risk of overfitting, and a need for enhanced model transparency.
• Clinical translation is hindered by regulatory and validation challenges.

Conclusions:

• AI has significant potential to enhance bone metastasis (BM) diagnosis and improve radiology workflows in oncology.
• Addressing data diversity, model explainability, and conducting large-scale validation are essential for clinical adoption and trust.