Advancements in artificial intelligence have enabled the extraction of valuable diagnostic information from imaging modalities traditionally used for a single purpose. In the context of myocardial perfusion imaging (MPI), computed tomography attenuation correction (CTAC) scans are primarily used for attenuation correction, but they also contain additional anatomical and pathological data that remain underutilised. A holistic AI-based approach aims to integrate this underexplored information to enhance mortality prediction in patients undergoing hybrid single-photon emission computed tomography/computed tomography (SPECT/CT) MPI.
Must Read: Assessing Myocardial Fibrosis at 3-Tesla versus 5-Tesla MRI
A recent study has investigated the potential of AI to extract and leverage multimodal imaging features for a more accurate and comprehensive assessment of patient risk. By utilising AI models to analyse the wealth of information embedded within CTAC scans, researchers seek to refine risk stratification methodologies and improve patient outcomes.
Patient Characteristics and Imaging Parameters
The study analysed 10,480 patients from four different medical institutions, extracting 15 radiomic features from 33 different structures visible in CTAC scans. These features were assessed in combination with MPI data, as well as additional AI-driven metrics such as coronary artery calcium (CAC) and epicardial adipose tissue (EAT) volume. Patients with normal myocardial perfusion had a lower mortality rate compared to those with abnormal perfusion, reinforcing the importance of comprehensive cardiac assessments.
A range of clinical and demographic factors were also evaluated, with findings indicating that age, sex, hypertension, diabetes and dyslipidaemia played significant roles in influencing mortality risk. Notably, male patients exhibited a higher risk profile, particularly those with elevated CAC and EAT volumes. The study also highlighted substantial variability in imaging parameters across different patient groups, suggesting that a standardised AI-based analysis could enhance the consistency and accuracy of risk assessments. The ability to integrate CTAC, MPI and clinical data into a unified AI-driven model demonstrated promising potential for improving predictive accuracy in clinical settings.
AI-Driven Model Performance and Risk Stratification
The AI hybrid model, which incorporated CTAC, MPI and clinical features, demonstrated superior predictive performance for all-cause mortality compared to models relying solely on traditional imaging or clinical assessment. The area under the receiver-operating characteristic curve (AUC) for the AI hybrid model was 0.80, significantly outperforming both CAC and perfusion-based models, which individually achieved AUCs of 0.64 and 0.62, respectively.
One of the most significant findings of the study was the high predictive value of lung-related features. The lungs emerged as a critical contributor to mortality prediction, emphasising the importance of extracardiac imaging in patient assessments. This suggests that conventional cardiac evaluations may overlook crucial indicators of mortality risk, which could otherwise be captured through AI-enhanced multimodal analysis. The AI-driven CTAC model alone achieved an AUC of 0.78, further highlighting the value of extracardiac information in risk stratification.
The model’s robust performance was maintained across various acquisition protocols and patient demographics, reinforcing its potential for widespread clinical application. Notably, the AI hybrid model demonstrated consistent effectiveness across different medical sites, suggesting that it could be integrated into a broad range of clinical environments without requiring substantial modifications to existing imaging protocols.
Clinical Implications and Future Applications
The study’s findings suggest that integrating AI-driven analysis into hybrid MPI workflows could significantly enhance clinical decision-making. By leveraging AI to identify key imaging markers, clinicians can improve risk assessment for patients with coronary artery disease (CAD) and other cardiovascular conditions. Given the study’s demonstration of the significance of extracardiac findings, particularly lung abnormalities, it becomes evident that broader imaging evaluations should be considered in routine assessments.
The potential for AI to automate and standardise diagnostic processes represents a significant step forward in nuclear cardiology. With the ability to integrate AI-driven models into existing workflows, healthcare providers may achieve more precise risk stratification and earlier intervention, ultimately improving patient outcomes. As AI technology continues to evolve, further research should explore real-world applications of these models across diverse clinical settings and patient populations. Future studies should also assess the impact of AI-driven diagnostics on clinical workflows, investigating how automation can enhance efficiency while maintaining high diagnostic accuracy.
This study highlights the significant role AI can play in augmenting hybrid MPI through the extraction of multimodal imaging features. By incorporating AI-based analysis of CTAC scans, clinicians can gain deeper insights into patient risk factors beyond conventional MPI findings. The AI hybrid model demonstrated superior performance in predicting all-cause mortality, reinforcing the potential of AI to enhance diagnostic precision and patient care.
Future research should focus on validating these findings in broader patient populations and integrating AI models into clinical practice to optimise risk stratification and treatment planning. The study’s demonstration of the value of extracardiac imaging underscores the need for a more comprehensive approach to patient assessments. Through continued advancements in AI and imaging technologies, medical professionals may be able to further refine predictive models and improve early detection of high-risk conditions, ultimately contributing to better healthcare outcomes for patients worldwide.
Source: npj Digital Medicine
Image Credit: iStock
