What is the clinical evidence from this study?

Study design: Cohort. Population: Type 2 Diabetes Mellitus (n=2438). Intervention: ECG-DiaNet (ECG features + Clinical Risk Factors) vs. Clinical Risk Factors (CRF)-only model. Primary outcome: AUROC for incident T2DM prediction (95% CI 0.8427-0.8451, p=<0.001).

July 25, 2025Scientific Reports3 citationsOpen Access

ECG features improve multimodal deep learning prediction of incident T2DM in a Middle Eastern cohort

Q: What are the key findings of this study?

The ECG-DiaNet model, integrating ECG features with clinical risk factors, significantly improved incident Type 2 diabetes prediction compared to clinical risk factors alone (AUROC 0.845 vs 0.8217).

FMFarida MohsenHamad bin Khalifa University ASAli SafaLebanese University ZSZubair ShahUniversity of Kansas Medical Center

Key Result

The ECG-DiaNet model, integrating ECG features with clinical risk factors, significantly improved incident Type 2 diabetes prediction compared to clinical risk factors alone (AUROC 0.845 vs 0.8217).

Study Design

Type

Cohort (n=2,438)

Structured PICO

Does integrating ECG features with clinical risk factors using a deep learning model improve the prediction of incident T2DM compared to clinical risk factors alone?

Population

Middle Eastern cohort from the Qatar Biobank (QBB), including a development cohort (n=2043) and a longitudinal test cohort (n=395).

Intervention

ECG-DiaNet, a multimodal deep learning model integrating electrocardiogram (ECG) features with established clinical risk factors (CRFs).

Comparator

Unimodal models based solely on ECG or clinical risk factors (CRFs).

Outcome

Prediction of incident Type 2 Diabetes Mellitus (T2DM) onset, measured by area under the receiver operating characteristic curve (AUROC).surrogate

A multimodal deep learning model combining ECG features and clinical risk factors significantly improves the prediction of incident Type 2 Diabetes Mellitus.

Main Result

Absolute Event Rate: 0.845% vs 0.8217%

p-value: p=<0.001

Limitations

Exclusive use of Qatar Biobank data limits clinical generalizability and lacks external validation.
Single-region design may not extend to other populations, age groups, or ethnicities.
The longitudinal test cohort was relatively small (n = 395).
Retrospective design could introduce selection and information biases.
Model relies solely on six ECG-derived features and a limited set of clinical risk factors.

Abstract

Type 2 Diabetes Mellitus (T2DM) remains a significant global health challenge, underscoring the need for early and accurate risk prediction tools to enable timely interventions. This study introduces ECG-DiaNet, a multimodal deep learning model that integrates electrocardiogram (ECG) features with established clinical risk factors (CRFs) to improve the prediction of T2DM onset. Using data from the Qatar Biobank (QBB), we compared ECG-DiaNet against unimodal models based solely on ECG or CRFs. A development cohort (n = 2043) was utilized for model training and internal validation, while a separate longitudinal cohort (n = 395) with a median five-year follow-up served as the test set. ECG-DiaNet demonstrated superior predictive performance, achieving a higher area under the receiver operating characteristic curve (AUROC) compared to the CRF-only model (0.845vs.0.8217), which was statistically significant based on the DeLong test (p < 0.001), thus highlighting the added predictive value of incorporating ECG signals. Reclassification metrics reinforced these improvements, with a significant Net Reclassification Improvement (NRI = 0.0153,p < 0.001) and Integrated Discrimination Improvement (IDI = 0.0482,p = 0.0099), confirming the enhanced risk stratification. Furthermore, stratifying participants into Low-, Medium-, and High-risk categories revealed that ECG-DiaNet achieved a higher positive predictive value (PPV) in the high-risk group compared to CRF-only models. These findings, together with the non-invasive nature and wide accessibility of ECG technology, suggest the potential of ECG-DiaNet for clinical implementation. However, further validation using larger and more diverse datasets is needed to improve generalizability.

问 AI

Bookmark

View Full Paper