Continuous, noninvasive blood pressure (BP) monitoring is an ongoing unmet clinical need, particularly in intensive care unit (ICU) settings where hemodynamically unstable patients require high-frequency measurements. The current standard for continuous BP monitoring is invasive arterial catheterization, which provides beat-to-beat data but carries procedural risks and constrains patient mobility. The evaluated work addresses this gap by testing a wearable, wireless, cuffless alternative intended to enable untethered continuous BP assessment with minimal disruption to activities of daily living.
The device tested is the MOSAIC system, a multimodal, multi-nodal wearable sensor array. MOSAIC comprises multiple wireless sensor nodes placed at different body locations to capture biosignals used for BP estimation. The preprint emphasizes the form factor and deployment approach as differentiators from existing continuous BP sensors, highlighting an untethered configuration that aims to reduce impact on patient mobility.
The clinical evaluation was conducted in a cohort of ICU patients. Invasive arterial-line blood pressure measurements were used as the ground truth comparator for all analyses. Ethical approval was obtained from the Johns Hopkins University Institutional Review Board under IRB00384821. The authors state that patient/participant consent and appropriate institutional forms were archived and that reporting guidelines were followed. The article is a preprint and has not undergone peer review; it should therefore not be used to guide clinical practice.
Biosignals collected from the MOSAIC sensor nodes were processed to derive features for continuous BP estimation. The primary modelling approach reported is a Light gradient boosted machine (LGBM) regression model trained to predict beat-to-beat systolic and diastolic BP. Model performance was assessed against arterial-line-derived BP readings. The LGBM model was compared to linear and deep-learning regression baselines in the analyses described.
Using the LGBM regression model, the MOSAIC-derived predictions demonstrated close beat-to-beat agreement with invasive arterial-line measurements. Reported performance metrics include mean absolute error (MAE) and average ratio variability (ARV):
These results are presented in the preprint as evidence of strong agreement between MOSAIC predictions and arterial-line BP values.
The authors report that the LGBM regression achieved better beat-to-beat agreement and variability metrics compared with reported linear and deep-learning regression baselines. Specific numerical comparisons for the baseline models beyond the LGBM metrics are not provided in the abstract and were not reported in the source material supplied.
The findings are presented as supporting the feasibility of wearable, wireless, and cuffless BP monitoring in high-acuity clinical settings. MOSAIC’s multi-nodal, untethered design aims to reduce the procedural risks and mobility limitations associated with invasive arterial catheterization while delivering continuous beat-to-beat BP estimates aligned with arterial-line measurements. As the report is a preprint, further validation and peer review are needed before clinical implementation.
Data from the MOSAIC sensors are subject to Johns Hopkins University privacy restrictions; sharing requires a Data Use Agreement with the University. The study declares funding from the National Center for Advancing Translational Sciences (NCATS) under grant 1UM1TR004926. The authors declared no competing interest. Ethical approvals and patient consent were obtained and documented as described above.
The source material is a medRxiv preprint and explicitly notes it has not been certified by peer review. Details such as cohort size, patient selection criteria, full modelling methods, training/validation split, and comparative baseline numeric results beyond those reported for the LGBM model were not provided in the abstract or accompanying sections included in the supplied source text. For these methodological and dataset specifics, the full preprint should be consulted.
In ICU patients, the MOSAIC multimodal, multi-nodal wearable sensor array combined with an LGBM regression model produced beat-to-beat BP estimates showing MAE 5.66 ± 5.94 mmHg (SBP) and 2.45 ± 2.87 mmHg (DBP) against invasive arterial-line measurements. Reported ARV values were 0.527 ± 0.185 (SBP) and 0.489 ± 0.170 (DBP). The evaluation supports the feasibility of a wireless, cuffless approach for continuous BP monitoring in high-acuity settings, while emphasizing that the work is preliminary and presented as a preprint.