Novel smartphone-based technology tracks 'imperceptible facial blood flow change.'
This article was first published on Tuesday, August 6, 2019 in MedPage Today.
By Ashley Lyles, Staff Writer, MedPage Today
Selfies taken with a smartphone with transdermal optical imaging (TOI) may soon be a way to conveniently measure blood pressure, a study found.
TOI "processes imperceptible facial blood flow changes from videos captured with a smartphone camera and uses advanced machine learning to determine blood pressure from the captured signal," explained Kang Lee, PhD, of the University of Toronto, and colleagues.
Their smartphone software model predicted BP measures with an average accuracy of 95.8% for pulse pressure, 95.7% for diastolic BP, and 94.8% for systolic BP in normotensive adults, they wrote in Circulation: Cardiovascular Imaging.
The average prediction biases were 0.52 mm Hg for pulse pressure, 0.39 mm Hg for systolic BP, and −0.20 mm Hg for diastolic BP, the investigators found.
Monitoring BP is key in the managment and prevention of hypertension, noted Lee, but "The current cuff-based methods are not convenient and uncomfortable for regularly and repeated measurement."
"The present study shows the new method may make convenient and comfortable monitoring of blood pressure a reality," Lee told MedPage Today.
Co-author Zhong-Ping Feng MD, PhD, also of the University of Toronto, added that "This convenient contactless tool provides the option for individuals regularly monitoring and recording their blood pressure over time, allowing for better detection of early onset of blood pressure changes and time-lapse- based blood pressure monitoring by the increased population of smartphone users, and thus permits early prevention and management of high blood pressure of the general [public] at none or little cost of health care funds."
These findings point to the promise of obtaining BP measurement with a video camera, noted Ramakrishna Mukkamala, PhD, of Michigan State University in East Lansing, in an accompanying editorial.
Although privacy concerns would have to be addressed, "such rich information in facial video may render this video camera approach to be more promising for BP measurement than a contact measurement of the finger [photo-plethysmography] waveform," Mukkamala stated.
Future studies should include heterogeneous participants, and focus on gathering possible BP-related data from facial video, such as facial expressions and head ballistocardiography, he added.
The researchers evaluated 1,328 normotensive Chinese and Canadian adults using a smartphone with TOI to record BP. Recordings were 2-minutes long. The participants were recruited from the Toronto institution and the Physical Examination Center of the Affiliated Hospital of Hangzhou Normal University in Hangzhou, China.
To create computational models that predicted reference pulse pressure, diastolic and systolic BPs from facial blood flow information, the investigators used an advanced machine learning algorithm. They used 15% of the cohort to test these models, 70% to train them, and 15% of the sample to validate the models' performance.
Study limitations included the focus on individuals with a diastolic BP range of ≥60 to <90 mm Hg and a normotensive systolic BP range of ≥100 to <140 mm Hg. The study did not have a cohort with hypotensive and hypertensive BPs. Also, the cohort was racially homogeneous as the majority was of East Asian descent, but "our participants displayed a reasonable degree of skin tone variation and this factor did not impact model prediction accuracy," the authors pointed out.
"Future work will determine whether these models meet the clinically accepted accuracy threshold of 5±8 mm Hg when tested on a full range of blood pressures according to international accuracy standards," they stated.
(The study was supported by the Canadian Institutes of Health Research, the Natural Sciences and Engineering Research Council of Canada, and the National Science Foundation of China.
Lee and a co-author disclosed relevant relationships with Transdermal Optical Imaging. Feng disclosed no relevant relationships with industry.
Mukkamala disclosed support from the NIH, patents for cuffless BP measurements, and a relevant relationship with Digitouch Health.)
“The present study shows the new method may make convenient and comfortable monitoring of blood pressure a reality.”
Kang Lee, PhD, of the University of Toronto
KEY TAKEAWAYS
Transdermal optical imaging processes imperceptible facial blood flow changes from videos captured with a smartphone camera and uses machine learning to determine blood pressure.
The smartphone software model predicted BP measures with an average accuracy of 95.8% for pulse pressure, 95.7% for diastolic BP, and 94.8% for systolic BP in normotensive adults.