跳到正文
DEV Community· News·· 15 天前AI 評分59

Google Research AI 系統以消費級智能手錶檢測高血壓徵兆

Google Research: an AI system detects signs of hypertension from a consumer smartwatch

AI 導讀

Google Research 公佈 AI-PPG-ACC-HTN,利用消費級智能手錶的 PPG 和加速度計訊號,在無袖帶情況下檢測高血壓徵兆。在一項涵蓋 196 名參與者的多中心前瞻研究中,初篩靈敏度為 65.8%,特異度為 90%,陽性預測值為 80.6%;同一組別的初次診室測量靈敏度為 55.3%。

正文

News

Infographic cover

Google Research: an AI system detects signs of hypertension from a consumer smartwatch

📅 November 7, 2026 • Google Research and Nature

A blog post and a Nature paper
Google Research published a blog post and an article in Nature on November 5, 2026 about a system that detects signs of hypertension from a consumer smartwatch without a cuff. The Nature paper is titled Opportunistically Detecting Signs of Hypertension on a Consumer Smartwatch. The system is called AI-PPG-ACC-HTN and uses two sensors that ordinary smartwatches already carry: photoplethysmography (PPG) and accelerometry (ACC).

How the system is built
The system works in two stages. First, a one-dimensional deep convolutional network is trained with self-supervised contrastive learning on 492,347 hours of unlabeled sensor data from 2,109 participants, using overlapping 15-second blocks of PPG and accelerometer waveforms. Second, features are averaged across wear days, and a regularized logistic regression over the aggregated representation plus participant height returns the probability of hypertension.

Validation on 196 participants
The model was tested in a prospective multicenter study on 196 participants at five centers in the United States, with no known cardiovascular disease and no antihypertensive medication. The reference was 24-hour ambulatory blood pressure monitoring, with seven days of real-world wear. For primary screening, sensitivity was 65.8 percent, specificity 90 percent and positive predictive value 80.6 percent. Initial office measurement in the same cohort reached 55.3 percent sensitivity and 77.8 percent positive predictive value.

What the confirmatory test showed
On a subset of 48 participants with elevated baseline office pressure, a confirmatory test reached 78.4 percent sensitivity, 90.9 percent specificity and 96.7 percent positive predictive value. The model does not estimate systolic and diastolic pressure for a given moment.

What it means for engineers
The model looks for a sustained hypertensive state by aggregating features across several days. Google states that this is early research work, not a diagnosis and not a replacement for cuff measurement, and that the results point to no existing or planned commercial device feature. The study did influence the Health Guardian blood pressure tracking feature in the Google Health app. For engineers, passive screening on mass-market wearable sensors is becoming a real class of tasks.

Key figures

IMPACT Passive screening on sensors that people already wear is turning from a research topic into a measurable engineering problem, and the hard part moves from signal processing to thresholds, validation and error handling.

Sources: research.google - nature.com - medicalxpress.com - the-decoder.com

IMPACT checklist


Continue the conversation: https://t.me/AIAutomationArchitecture · Bluesky

AI Automation

來源:DEV Community · dev.to