Artificial intelligence in nuclear cardiology: Technical perspectives, strategic directions, and recommendations from an IAEA expert working group
: Wiefels, Christiane; Juárez-Orozco, Luis Eduardo; Craviolatti, Pietro Selemo; Diahiliev, Oleksandr; Eskander, Amir; Giubbini, Raffaele; Milan, Elisa; Zhou, Weihua; Karthikeyan, Ganesan; Jimenez-Heffernan, Amelia; Peix, Amalia; Apostol, Angelin; Brink, Anita; Dondi, Maurizio; Paez, Diana
Publisher: Elsevier
: 2025
Seminars in Nuclear Medicine
: 56
: 1
: 119
: 131
: 0001-2998
: 1558-4623
DOI: https://doi.org/10.1053/j.semnuclmed.2025.11.011
: https://doi.org/10.1053/j.semnuclmed.2025.11.011
: https://research.utu.fi/converis/portal/detail/Publication/508350955
Artificial intelligence (AI) is increasingly permeating nuclear cardiology and offers the possibility to enhance diagnostic accuracy, prognostic stratification, and operational efficiency. AI is demonstrating applicability across the imaging workflow—from individualized patient selection and adaptive image reconstruction to denoising of low-dose datasets, automated attenuation and motion correction, calcium scoring, and the integration of imaging with clinical and functional variables for enhanced diagnosis and comprehensive risk assessment.But the translational trajectory of AI in nuclear cardiology is challenged by the lag in fundamental AI knowledge among researchers and clinicians, the quality of the target data regarding heterogeneity in acquisition protocols, scanner platforms, and patient populations, and by infrastructural disparities that constrain the generation of large, representative datasets needed for training and validation, particularly in low-resource settings. Additionally, necessary regulatory and legal frameworks remain in early stages of harmonization.This white paper, developed by an International Atomic Energy Agency (IAEA) working group, provides a succinct overview of the technical basis, areas of deployment, clinical value and unmet challenges of AI in nuclear cardiology. It makes punctual suggestions to aid maturation in this area while maintaining a sober interaction with the overwhelming nature of the field. These include promoting standardized acquisition and reporting practices, establishing globally representative reference datasets, promoting imaging multimodality frameworks and developing AI-proficient clinical and technical personnel. Under these conditions, AI may meaningfully enhance the diagnostic and prognostic value of nuclear cardiology while supporting equitable implementation and preserving clinical accountability.
UTU Research Portal