Empowering PET: harnessing deep learning for improved clinical insight

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Artificial Intelligence
• Deep Learning

Background:

• Artificial intelligence (AI) is revolutionizing medical fields, including nuclear medicine and radiology.
• AI is poised to support healthcare professionals, improving diagnosis, treatment response assessment, and patient management.
• Positron emission tomography (PET) imaging is a key area benefiting from AI advancements.

Purpose of the Study:

• To review the transformative impact of artificial intelligence (AI) on positron emission tomography (PET) imaging.
• To provide an overview of AI applications in nuclear medicine and deep learning (DL) in cancer diagnosis and therapy using PET.
• To discuss the integration of AI into clinical practice and future directions in molecular imaging.

Main Methods:

• Review of AI applications in PET image generation: acquisition, reconstruction, restoration, and motion correction.
• Exploration of AI integration in clinical practice: segmentation, detection, classification, quantification, treatment planning, dosimetry, and radiomics/radiogenomics.
• Discussion of limitations, ethical considerations, and future challenges of AI in molecular imaging.

Main Results:

• AI enhances the entire PET imaging pipeline, from image generation to clinical applications.
• AI supports various clinical tasks in PET diagnosis, prognosis, and treatment assessment.
• AI is crucial for interpreting complex relationships between imaging and multiomics data.

Conclusions:

• AI significantly transforms PET imaging, leading to tangible improvements in patient treatment and response assessment.
• AI integration in nuclear medicine offers substantial potential for enhanced diagnostic and therapeutic capabilities.
• Addressing challenges like data access, interoperability, and the 'black-box' problem is vital for AI adoption in molecular imaging.