Objective:

To evaluate which subset of the standard 12 leads should be recorded by portable reduced-lead ECG devices for optimal reconstruction and diagnostic accuracy.

Approach:

• Benchmarking: Evaluated all 4,094 C(12, N) lead subsets for N = 1–11 using four reconstruction paradigms: linear regression, ridge regression, a lightweight 1-D convolutional network, and a Transformer encoder–decoder.
• Performance Assessment: Assessed performance along three axes: reconstruction fidelity, downstream diagnostic accuracy, and acquisition efficiency, using a Composite Lead Score.
• Validation: Conducted external validation on two datasets (CPSC2018 and Chapman-Shaoxing) to confirm the robustness of recommended configurations.

Key Findings:

• The consensus efficiency–accuracy knee is at N = 4, achieving a mean macro-F1 of 0.631.
• Recommended configurations include V6 for pre-hospital triage, I + II + AVR + AVF for community screening, and a 7-lead set for home monitoring.
• All recommended configurations retain ≥ 83% of within-PTB-XL three-class F1 on external cohorts.

Interpretation:

The study presents configurations for reduced-lead ECG devices based on systematic lead selection for cardiac monitoring.

Limitations:

• The study may not account for all clinical scenarios or patient populations.
• Configurations derived may require cohort-specific re-derivation for optimal performance.

Conclusion:

The findings support the development of reduced-lead ECG devices with specific lead configurations for various clinical settings.