What is the clinical evidence from this study?

Study design: Observational. Population: Transthyretin-associated cardiac amyloidosis (ATTR-CM) (n=124). Intervention: AI-driven multidimensional patient-specific model vs. ECG and echocardiographic data only. Primary outcome: Diagnostic performance (AUC) for detecting ATTR-CM.

What question did this study set out to answer?

This study aims to enhance early diagnosis of transthyretin cardiac amyloidosis by integrating multimodal patient data.

June 25, 2026Open Access

Model-guided medicine for early diagnosis of transthyretin-associated cardiac amyloidosis using multimodal data integration and standardized interoperable models (the CRONOS-ATTR study)

Key Result

An AI-driven multidimensional patient-specific model integrating clinical, electrocardiographic, and echocardiographic data achieved an AUC of 0.84 for the detection of transthyretin cardiac amyloidosis.

Key Points

This study aims to enhance early diagnosis of transthyretin cardiac amyloidosis by integrating multimodal patient data.
Utilized AI algorithms and human intelligence to create a patient-specific model from diverse clinical data sources.
Trained a machine learning model on data from 124 patients to evaluate diagnostic performance.
Achieved a diagnostic performance AUC of 0.84 with high sensitivity and precision.
Utilized SHAP values to provide interpretable outputs for clinical understanding.

Study Design

Type

Observational (n=124)

Multicenter

Structured PICO

Does an AI-driven multimodal data integration model improve the early detection of ATTR-CM in heart failure patients with suspected cardiac amyloidosis?

Population

124 heart failure patients with an interventricular septum ≥ 12 mm and suspected cardiac amyloidosis, evaluated for ATTR-CM using a multimodal AI-driven diagnostic model.

Exposure

Multimodal data integration (clinical, electrocardiographic, and echocardiographic) using an AI-driven, interpretable predictive machine learning model (XGBoost) within a model-guided medicine framework.

Comparator

Model restricted to ECG and echocardiographic data only (excluding clinical variables).

Outcome

Diagnostic performance (Area Under the Curve - AUC) for the early detection of ATTR-CM.surrogate

An explainable AI model integrating clinical, ECG, and echocardiographic data achieved strong diagnostic performance (AUC 0.84) for the early detection of transthyretin cardiac amyloidosis, significantly outperforming imaging data alone.

Main Result

Absolute Event Rate: 0.84% vs 0.56%

Limitations

Single-center setting using a manually curated case-control cohort
Excluded AL amyloidosis cases, limiting generalizability
Lack of external validation in larger and more diverse populations
Integration of AI tools into hospital infrastructure was not fully automated
Evaluated performance based on retrospective data with known diagnoses

Abstract

BACKGROUND: Early diagnosis of transthyretin cardiac amyloidosis (ATTR-CM) is essential for timely intervention but remains challenging due to its subtle and nonspecific clinical presentation. The CRONOS-ATTR study aimed to improve early detection of ATTR-CM by integrating multimodal data (clinical, electrocardiographic, and echocardiographic) within a model-guided medicine framework. METHODS: Using artificial intelligence (AI) algorithms from CardiolyseECGSoftware and Ligence Heart, along with human intelligence (multidimensional interpretable models), we standardized and harmonized heterogeneous data sources into a unified patient-specific model (PSM). RESULTS: A machine learning model based on XGBoost was trained on a cohort of 124 patients and achieved strong diagnostic performance (AUC 0.84), with high sensitivity and precision. The model provided interpretable outputs using SHAP values, facilitating clinical understanding and trust. This approach not only enabled accurate early detection of ATTR-CM but also demonstrated feasibility for integration into real-world clinical workflows. CONCLUSIONS: Our findings support the use of explainable AI to enhance screening strategies for cardiac amyloidosis and establish a foundation for scalable, automated tools that can be embedded within healthcare systems.

KI fragen

Bookmark

View Full Paper