Addressing disparities in transthyretin amyloid cardiomyopathy: A systematic review of artificial intelligence in the early identification to improve patient outcomes.
Authors
Affiliations (7)
Affiliations (7)
- Tulane University School of Medicine, New Orleans, Louisiana, USA. Electronic address: [email protected].
- Tulane University School of Medicine, New Orleans, Louisiana, USA. Electronic address: [email protected].
- Tulane University School of Medicine, New Orleans, Louisiana, USA. Electronic address: [email protected].
- Tulane University School of Medicine, New Orleans, Louisiana, USA. Electronic address: [email protected].
- Tulane University School of Medicine, New Orleans, Louisiana, USA. Electronic address: [email protected].
- Tulane University School of Medicine, New Orleans, Louisiana, USA. Electronic address: [email protected].
- Tulane University School of Medicine, New Orleans, Louisiana, USA. Electronic address: [email protected].
Abstract
Transthyretin amyloid cardiomyopathy (ATTR-CM) remains substantially underdiagnosed among Black patient populations. When applied to non-invasive diagnostic and screening techniques, artificial intelligence (AI) can aid in the early detection of subtle patterns indicative of transthyretin amyloid cardiomyopathy (ATTR-CM), potentially before clinical symptoms appear. This systematic review examines the current landscape of AI-based diagnostic tools for ATTR-CM. A comprehensive search was performed in PubMed, Embase, Cochrane Library, Scopus, and Web of Science. The search strategy targeted peer-reviewed, English-only articles published from January 1, 2015, to May 27, 2025, focusing on the application of artificial intelligence to electrocardiograms (ECG), echocardiograms, and cardiac magnetic resonance imaging (CMR) on ATTR-CM specifically. We include the studies that have diagnostic performance measures as their primary outcomes. 713 studies were identified from databases, and 27 were included. Studies were categorized into three groups based on their primary AI-applied non-invasive diagnostic tools: ECG (n = 16), echocardiogram (n = 8), and CMR (n = 3), either as solo or combined interventions. Performance of AI-enhanced ECG yielded Area Under the Curves (AUCs) ranging from 0.82 to 0.97. AI-integrated echocardiography achieved AUCs between 0.87 and 0.97, while AI-CMR models resulted in AUCs up to 0.92. Multimodal approaches that integrated data from multiple sources also reported AUCs ranging up to 0.97. The integration of AI into ECG, echocardiograms, and CMR have the potential to significantly improve the early detection and classification of ATTR-CM, preventing the progression to end-stage heart failure. Accessible AI-based screening tools could address health disparities by enabling earlier identification and treatment initiation for Black populations who face disproportionate disease burden of ATTR-CM and diagnostic delays.