The Role of [18F]FDG PET-Based Radiomics and Machine Learning for the Evaluation of Cardiac Sarcoidosis: A Narrative Literature Review

Abstract

Background/Objectives: Cardiac sarcoidosis (CS) is an inflammatory cardiomyopathy with a strong clinical impact on patients affected by the disease and a challenging diagnosis. Methods: This comprehensive narrative review evaluates the role of [18F]fluorodesoxyglucose ([18F]FDG) positron emission tomography (PET)-based radiomics and machine learning (ML) analyses in the assessment of CS. Results: The value of [18F]FDG PET-based radiomics and ML has been investigated for the clinical settings of diagnosis and prognosis of patients affected by CS. Even though different radiomics features and ML models have proved their clinical role in these settings in different cohorts, the clear superiority and added value of one of them across different studies has not been demonstrated. In particular, textural analysis and ML showed high diagnostic value for the diagnosis of CS in some papers, but had controversial results in other works, and may potentially provide prognostic information and predict adverse clinical events. When comparing these analyses with the classic semiquantitative evaluation, a conclusion about which method best suits the final objective cannot be drawn with the available references. Different methodological issues are present when comparing different papers, such as image segmentation and feature extraction differences that are more evident. Furthermore, the intrinsic limitations of radiomics analysis and ML need to be overcome with future research developed in multicentric settings with protocol harmonization. Conclusions: [18F]FDG PET-based radiomics and ML show preliminary promising results for CS evaluation, but remain investigational tools since the current evidence is insufficient for clinical adoption due to methodological heterogeneity, small sample sizes, and lack of standardization.