Comprehensive Review of AI Integration in Nuclear Medicine – Current Techniques, Limitations, and Future Innovations

Abstract

The integration of Artificial Intelligence (AI) methods in Positron Emission Tomography (PET) and Single-Photon Emission Computed Tomography (SPECT) presents unique challenges and opportunities in healthcare. Despite the slower adoption of AI in medical fields compared to other domains, recent advancements have showcased its potential to revolutionize diagnostic precision and therapeutic innovation. This literature review explores the integration of AI in nuclear imaging, focusing on the applications in photon detection, image reconstruction, and post-processing, as well as in further image analysis where segmentation and radiomics play an important role. Specific examples making use of different Machine Learning, and Deep Learning techniques such as Convolutional Neural Networks (CNNs) and Generative Adversarial Networks (GANs) have been reviewed, demonstrating their ability to either outperform the conventional methods in extracting information from images or to automatize those that are tedious and time-consuming for clinicians. Despite the great results obtained in research, many imitations keep these methods still a step behind in their commercialization. This review aims to provide insights into the current AI applications in nuclear imaging that address challenges such as data complexity, standardization, and lack of explainability, along with the expectations of future directions for research and clinical implementation.