Large models in medical imaging: Advances and prospects

Area of Science:

• Artificial Intelligence in Healthcare
• Medical Imaging Analysis
• Computational Pathology

Background:

• Large models show promise for revolutionizing medical imaging.
• These models offer advanced capabilities for processing complex medical data.
• Current applications span various imaging modalities and clinical tasks.

Purpose of the Study:

• To review concepts, methods, and progress of large models in medical imaging.
• To highlight the potential of large models in precision medicine.
• To identify challenges and future research directions for clinical implementation.

Main Methods:

• Integration of multimodal data using large model technologies.
• Training large models with medical datasets via self-supervised pretraining and fine-tuning.
• Leveraging cross-modal representation learning for enhanced interpretation.

Main Results:

• Large models can improve diagnostic accuracy and workflow efficiency.
• Potential applications include image segmentation, disease diagnosis, and personalized treatment.
• Successful implementation requires addressing data scarcity and integration challenges.

Conclusions:

• Large models offer significant potential to advance medical imaging and precision medicine.
• Overcoming challenges related to data, safety, and integration is crucial for clinical deployment.
• Future research should focus on developing efficient, interpretable, and generalizable models.