Can AI Enhance Ultrasound Detection of Cardiac Amyloidosis? What a Multicenter Study Reveals

Based off a transthoracic apical four-chamber view, an emerging echocardiography-based artificial intelligence (AI) screening software may offer enhanced detection of cardiac amyloidosis (CA) in comparison to the transthyretin cardiac amyloidosis score (TCAS) and the increased wall thickness (IWT) score.

For a new multicenter study, recently published in the European Heart Journal, researchers evaluated the AI screening software (EchoGo Amyloidosis, Ultromics) for the detection of CA. The external validation cohort, including 597 people with CA and 2,122 controls, was derived from 18 facilities, according to the study.

In external validation testing, the study authors found that the AI screening software had an 85 percent sensitivity and a 93 percent specificity for detecting CA. The researchers also noted a 95.6 percent negative predictive value (NPV) and a 78 percent positive predictive value (PPV).

“The AI model presented in this manuscript has the potential to improve both the accuracy and efficiency of CA detection compared with traditional TTE-based methods. Importantly, the model had sufficiently high PPV and NPV to demonstrate clinical utility, offering the potential to augment the frontline screening role that echocardiography plays in the evaluation of suspected CA,” wrote lead study author Jeremy A. Slivnick, M.D., FACC, an assistant professor of medicine in the Section of Cardiovascular Medicine and the Department of Internal Medicine at the University of Chicago, and colleagues.

The researchers also noted that the area under the receiver operating characteristic curve (AUROC) for the AI software (93 percent) was 20 percent higher than that of the TCAS (73 percent) and 13 percent higher than the IWT score (80 percent).

“Our AI model also outperformed the accuracy of both the TCAS and IWT score in a subset of older adults with HFpEF (heart failure with preserved ejection fraction) and increased left ventricular wall thickness. These multiparametric risk scores were previously found to be accurate at differentiating ATTR-CA from phenotypically similar conditions including HFpEF and left ventricular hypertrophy,” added Slivnick and colleagues.

The AI screening software also offered consistent sensitivity across different subtypes of CA, including light-chain CA (84 percent), wild-type transthyretin CA (85 percent) and hereditary transthyretin CA (86 percent), according to the researchers.

“The use of this rapid, fully automated AI model has the potential to improve the accuracy and efficacy of echocardiographic CA detection, thereby facilitating access to life-prolonging therapies,” emphasized Slivnick and colleagues.

(Editor’s note: For related content, see “FDA Clears AI-Powered Ultrasound Software for Cardiac Amyloidosis Detection,” “Emerging PET Imaging Agent Gets FDA’s Breakthrough Therapy Designation for Cardiac Amyloidosis” and “UltraSight’s AI-Powered Cardiac Ultrasound Guidance Gets FDA Nod.”)

In regard to study limitations, the authors acknowldged the retrospective nature of the research as well as the larger proportion of men and higher percentage of left ventricular wall thickness in the CA group. The researchers also conceded the majority of the cohort was derived from tertiary academic medical facilities.