Objective Accurate cuffless blood pressure (BP) estimation could enable more accessible and continuous cardiovascular monitoring using wearable devices. Modern wearable sensors increasingly carry multiple LEDs and PDs, each probing tissue at a different depth and carrying complementary hemodynamic information.
This competition aims to promote the development of generalizable AI algorithms for snapshot cuffless BP estimation from multi-point multi-wavelength wearable PPG and ECG. Through a standardized, high-quality dataset and a reproducible competition framework, the challenge seeks to establish a transparent benchmark, encourage methodological innovation in multi-channel sensor fusion, and help bridge the gap between algorithm development and real-world wearable cardiovascular monitoring.
Evaluation Submissions are evaluated by a sample-level weighted Mean Absolute Error (MAE) across all test samples N:
MAE = (1/N) · Σᵢ₌₁ᴺ (0.6 · |SBP_true,i - SBP_est,i| + 0.4 · |DBP_true,i - DBP_est,i|)
Lower MAE indicates better estimation performance. For the leaderboard, the MAE is transformed via:
Leaderboard Score = 1 / (1 + MAE)
This transformation maps the score to the range (0, 1], while preserving the ranking order (lower raw score → higher leaderboard score). The Public Leaderboard is provided for interim feedback during the competition and may help you gauge your progress. The final rankings, however, are determined solely by the Private Leaderboard. All records are assigned exclusively to either the Public or Private Leaderboard.