Potential of Artificial Intelligence Cardiovascular PET imaging

Introduction: 1) Review standard AI algorithms to improve the imaging process. 2) Review the current state of PET imaging on cardiovascular disease and vascular abnormalities. 3) Cover the potential applications of artificial intelligence (AI) on PET imaging in this disease scope.

Methods: Artificial Intelligence (AI) has been a catalyst for innovation in the technology space and has also integrated into medical imaging. Employing AI can prove beneficial for the entire imaging process. From taking patient history to explaining imaging to patients to post-processing and diagnosis, AI algorithms can reduce the mundane tasks of nuclear medicine technologists. Moreover, the rapid increase of COVID-19 cases and its urgency for streamlined hospital care has accelerated the need for more efficient imaging techniques, which is why AI would be a useful tool. Several AI and machine learning (ML) algorithms can be beneficial for PET instrumentation. AI can be classified into supervised and unsupervised learning. Artificial Neural Networks (ANNs), Convolutional Neural Networks (CNNs), and Convolutional Autoencoders are also popular algorithms that can be implemented in imaging. In the scope of cardiology, Positron Emission Tomography (PET) is typically combined with Computed Tomography (CT) scans in detecting cardiovascular disease. This includes atherosclerosis, caused by plaque formation in arterial walls, and is one of the leading causes of cardiovascular disease.

Results: Currently, Positron Emission Tomography (PET) scans can assist in detecting CVD through atherosclerosis. FDG PET imaging is a good marker for detecting atherosclerotic plaque because it has macrophages that match with the disease. Research from our lab has stated the potential for total-body PET for atherosclerosis to be promising. AI can further enhance atherosclerosis imaging in attenuation correction imaging. While CT attenuation correction scans are typically created for PET, emerging AI techniques and CNNs can transform magnetic resonance images to forged CT images, which are useful for attenuation correction for PET scans. Another way artificial intelligence can assist in cardiac PET imaging is through motion correction. Currently, patient breathing may affect the quality of the scans, which may make it harder to diagnose. Previous studies have developed an algorithm for irregular breathing in abdominal PET imaging, which can be converted for cardiac imaging as well. This can be especially useful in the pandemic era and apply respiratory corrections to minimize the need for re-scanning or provide a higher dosage. Finally, AI can improve the patient experience by reducing time for quality control, radiotracer injection, image acquisition, and post-processing. This can allow more time for catching abnormalities and tumors early on.

Conclusions: Cardiovascular Disease is the number one cause of death globally, and AI has tremendous potential to improve image quality, reduce waiting and stress time in the imaging department, and increase diagnostic accuracy for disease. Because these models can teach themselves, training AI algorithms will likely prove to be a beneficial approach. It is not just limited to cardiac and vascular PET scans, but also other malignancies.