Artificial intelligence in diagnosing cardiac amyloidosis by standard echocardiography: opportunity or threat for sonographers? | ||
| Cardiovascular Biomedicine | ||
| Volume 6, Issue 1, September 2026, Pages 49-50 PDF (324.68 K) | ||
| Document Type: Letter to the editor. | ||
| DOI: 10.18502/cbj.v6i1.22789 | ||
| Authors | ||
| Razieh Parizad* ; Mohammadreza Taban Sadeghi | ||
| Cardiovascular Research Center, Tabriz University of Medical Sciences, Tabriz, Iran | ||
| Abstract | ||
| Dear Editor, Cardiac amyloidosis (CA) remains an underdiagnosed cause of heart failure with preserved ejection fraction (HFpEF) (1). Standard transthoracic echocardiography (TTE) continues to serve as the first-line screening tool; however, classic red flags, including increased left ventricular (LV) wall thickness with normal or low voltage, reduced global longitudinal strain (GLS), and the relative apical sparing pattern, are frequently overlooked (2). The rapid integration of artificial intelligence (AI) into echocardiography laboratories raises a critical question: does AI represent an opportunity or a threat for sonographers? AI as an Opportunity AI algorithms can detect subtle CA patterns that may escape human visual assessment. A deep learning (DL) model using a single apical four-chamber video clip achieved an area under the receiver operating characteristic (ROC) curve (AUROC) of 0.93, 85% sensitivity, and 93% specificity during external validation across 18 global sites (3). Another model discriminated CA from hypertensive heart disease (area under the curve [AUC] 0.92), hypertrophic cardiomyopathy (AUC 0.91), and aortic stenosis (AUC 0.93) (4). A systematic review confirmed that AI models perform comparably to or better than professional cardiologists (5). For sonographers, AI functions as a real-time "second pair of expert eyes," transforming them from passive data collectors into active clinical screeners. AI as a Threat There are three principal threats. First, overreliance on AI may lead to false-positive findings. For instance, the apical sparing pattern has only 72% sensitivity and 66% specificity for CA, and 32% of non-CA controls exhibit the same pattern (6). Second, overreliance on automated measurements might weaken pattern recognition skills and potentially lead to a gradual loss of diagnostic proficiency. Third, professional autonomy is likely to weaken if AI-driven automated reports bypass sonographer input entirely, thereby reducing their role primarily to image acquisition. Furthermore, implementation barriers, including cost, interoperability with electronic health records (EHRs), regulatory challenges, and liability concerns, remain unresolved (7). Conclusion AI will not replace sonographers who understand CA pathology; however, it may replace those who refuse to adopt AI-assisted diagnosis. Therefore, AI literacy should be a fundamental part of sonography training. Sonography training should include basic principles of machine learning, recognition of AI-generated false positives, and integration of AI alerts into the clinical workflow. Table 1 presents the key opportunities and threats with supporting references. | ||
| Keywords | ||
| Cardiac amyloidosis; Artificial intelligence; Echocardiography; Sonographers; Deep learning | ||
| References | ||
|
| ||
|
Statistics Article View: 14 PDF Download: 13 |
||