Testing Optical Character Recognition of Home Blood Pressure Measurements, a Randomized Trial.

Area of Science:

• Biomedical Engineering
• Health Informatics
• Cardiovascular Medicine

Background:

• Accurate home blood pressure (BP) monitoring is essential for diagnosing hypertension but is often hindered by challenges in patient self-reporting.
• Existing methods for remote BP data collection face limitations in accuracy and patient adherence.

Purpose of the Study:

• To evaluate the acceptability and accuracy of two low-cost electronic strategies for home BP measurement reporting: text messaging and a novel image-based optical character recognition (OCR) method.
• To compare the response rates and data accuracy between text-based and image-based BP reporting.

Main Methods:

• Development and testing of a two-stage OCR model to extract BP values from patient-submitted cellphone photographs of BP monitors.
• A crossover trial involving 50 participants to compare the accuracy and acceptability of image-based reporting versus text messaging for BP readings.
• Analysis of response rates and BP values obtained from both reporting methods.

Main Results:

• High participant engagement, with most respondents submitting the majority of requested BP data via both text (average 12/14) and photo uploads (average 11.3/14).
• The OCR model achieved 98.3% accuracy in reading BP values from suitable quality images, though 17.9% of images were rejected due to quality issues.
• No significant difference was found in the average BP values reported between the text messaging and image-based methods.

Conclusions:

• Successfully demonstrated the feasibility of extracting BP values from cellphone photographs of BP monitors using an OCR model.
• Both text messaging and image-based photo uploads are validated as accurate and acceptable methods for collecting home BP readings, offering practical solutions for remote patient monitoring.