American journal of hypertension · 2026
Testing Optical Character Recognition of Home Blood Pressure Measurements, a Randomized Trial.
СтатьяРандомизированное исследованиеПоддержка NIH: внешняя программаНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Strom Austin J, Khader Esra'a, Polgreen Philip M, Francis Shelby L, Segre Alberto M, Herman Ted, Antes Lisa M, Karanikolas Amanda R, Polgreen Linnea A
Аннотация
BACKGROUND: Diagnosing hypertension typically requires repeated blood pressure (BP) readings taken over multiple days, but obtaining accurate BP measurements from patients at home is known to be challenging, at least in part due to failure to accurately report those measurements. Here, we evaluate 2 low-cost electronic reporting strategies, text messaging, and a novel image-based alternative, for acceptability and accuracy.
METHODS: We developed and tested a 2-stage optical character recognition (OCR) model to "read" a BP monitor display from patient-provided cellphone photographs. We then conducted a crossover trial to test the accuracy and acceptability of image-based BP reporting with having patients report BP measurements by text messaging. We compared the response rates and values obtained from photo uploads to those from text messages.
RESULTS: The majority of the 50 respondents submitted most of the requested data. We received an average of 12/14 texted BP responses and 11.3/14 photo responses from participants. Only 4 participants sent neither text nor photo responses. In addition, our OCR model achieved 98.3% accuracy when "reading" BP values from images of sufficient quality (17.9% were rejected for quality issues), and the average BP reported by subjects did not differ between our text and image-based methods.
CONCLUSION: We were able to successfully "read" BP values from photos of BP monitors sent by participants. Both text messaging and photo uploads of monitor displays appear to be accurate and acceptable approaches for collecting home BP readings.
METHODS: We developed and tested a 2-stage optical character recognition (OCR) model to "read" a BP monitor display from patient-provided cellphone photographs. We then conducted a crossover trial to test the accuracy and acceptability of image-based BP reporting with having patients report BP measurements by text messaging. We compared the response rates and values obtained from photo uploads to those from text messages.
RESULTS: The majority of the 50 respondents submitted most of the requested data. We received an average of 12/14 texted BP responses and 11.3/14 photo responses from participants. Only 4 participants sent neither text nor photo responses. In addition, our OCR model achieved 98.3% accuracy when "reading" BP values from images of sufficient quality (17.9% were rejected for quality issues), and the average BP reported by subjects did not differ between our text and image-based methods.
CONCLUSION: We were able to successfully "read" BP values from photos of BP monitors sent by participants. Both text messaging and photo uploads of monitor displays appear to be accurate and acceptable approaches for collecting home BP readings.