What is the clinical evidence from this study?

Study design: Cohort. Population: Heart Failure (n=343). Intervention: Digital Twin Framework vs. Traditional risk assessment based on clinical variables. Primary outcome: Composite endpoint including all-cause mortality, LVAD implantation, and heart transplantation (HR 2.72, 95% CI 1.53-5.06, p=0.001).

What does this research mean for the field?

Digital twin-guided AI models increased the predictive risk for the primary composite endpoint by 2.72 times in heart failure patients compared to traditional methods. Novelty: ClaimNovelty.NOVEL_FINDING. Consensus alignment: ConsensusAlignment.NEUTRAL.

February 17, 2025Open Access

Identification of digital twins to guide interpretable AI for diagnosis and prognosis in heart failure

FGFeng GuCentral South University AMAndreas MeyerUniversity of Zurich FJFilip JežekCharles University

Key Result

Digital twin-guided AI models increased the predictive risk for the primary composite endpoint by 2.72 times in patients compared to traditional methods, with a 5-year primary composite endpoint rate of 31.2% in the cohort.

Study Design

Type

Cohort (n=343)

Multicenter

Structured PICO

Does incorporating digital twin-derived features into prognostic AI models improve performance and interpretability for predicting cardiovascular death risk in heart failure patients compared to using only clinical variables?

Population

343 heart failure patients (215 with HFrEF, 128 with HFpEF defined as baseline LVEF ≥50% including 6 with pulmonary arterial hypertension and 32 with hypertrophic cardiomyopathy) from a retrospective Electronic Health Records review at the University of Michigan Health System.

Intervention

Prognostic AI models incorporating digital twin-derived features (mechanistic computational models of the cardiovascular system comprising optimized measurable and unmeasurable parameters).

Comparator

Prognostic AI models using only clinical variables.

Outcome

Prognostic performance for cardiovascular death risk.

Digital twins derived from mechanistic computational models can capture patient-specific cardiovascular physiology and improve the performance and interpretability of AI prognostic models for heart failure.

Main Result

Effect estimate: HR 2.72 (95% CI 1.53-5.06)

p-value: p=0.001

Limitations

Lack of an external validation set
Imbalance between the training and testing cohorts
Heterogeneity in patient backgrounds between cohorts

Abstract

Heart failure (HF) is a highly heterogeneous condition, and current methods struggle to synthesize extensive clinical data for personalized care. Using data from 343 HF patients, we developed mechanistic computational models of the cardiovascular system to create digital twins. These twins, consisting of optimized measurable and unmeasurable parameters alongside simulations of cardiovascular function, provided comprehensive representations of individual disease states. Unsupervised machine learning applied to digital twin-derived features identified interpretable phenogroups and mechanistic drivers of cardiovascular death risk. Incorporating these features into prognostic AI models improved performance, transferability, and interpretability compared to models using only clinical variables. This framework demonstrates potential to enhance prognosis and guide therapy, paving the way for more precise, individualized HF management.

Ask AI

Helpful

Bookmark

View Full Paper