Research Outline:
We aim to validate the accuracy and generalizability of our AI-based heart failure detection model, developed in Japan, using international datasets from the UK Biobank. Key research questions include:
1. How well does the AI model detect heart failure across different populations, ethnicities, and demographics?
2. What variations exist in sensitivity, specificity, and overall performance between UK Biobank and Japanese datasets?
3. How do demographic and ethnic differences impact the model’s predictive capability?
Objectives
– Evaluate model performance using UK Biobank ECG and clinical data.
– Identify potential biases or discrepancies in detection accuracy across demographic, ethnic, and clinical subgroups.
– Optimize the AI model for improved generalizability in diverse populations.
Scientific Rationale
Heart failure (HF) is a progressive, high-mortality condition with significant healthcare costs. Early diagnosis is challenging due to reliance on subjective symptoms and costly diagnostic methods, leading many patients to delay medical attention. Misdiagnosis is not rare, as HF symptoms overlap with other conditions, and access to advanced imaging remains limited.
Simplex Quantum has developed an AI-based HF detection model using ECG data to enable rapid, non-invasive, and cost-effective screening. While validated on Japanese datasets, external validation with international data is essential to ensure robustness.
Using UK Biobank data, this study will assess the model’s generalizability across diverse populations, identify biases, and refine its predictive accuracy. This research contributes to developing globally reliable diagnostic tools, ultimately improving early HF detection, patient outcomes, and healthcare efficiency.