Peripheral perfusion is one of those measurements that clinicians check constantly but technology has been slow to automate. A nurse presses a fingernail, watches the color return, and estimates capillary refill time. A pulse oximeter clips onto a finger and reports a perfusion index number. These are useful but limited — they capture a single point in time, at a single location, and they require physical contact with the patient.
Camera-based photoplethysmography is changing the math on what contactless monitoring can do beyond heart rate and respiratory rate. Researchers are now using standard cameras to assess microcirculation, map skin perfusion across tissue regions, and detect vascular abnormalities that previously required specialized equipment. The work is early, but the clinical implications are real: peripheral vascular disease affects over 200 million people globally, and most cases go undiagnosed until complications appear.
"rPPG was more sensitive in detecting perfusion changes than automated capillary refill time, showing significant differences between bacterial and COVID-19-associated sepsis groups during fluid resuscitation." — Marcinkevics et al., Medicina (2024)
How cameras detect perfusion in skin tissue
Every heartbeat sends a pulse of blood through the body's vascular network, all the way down to the capillary beds in the skin. This produces tiny fluctuations in skin color — invisible to the human eye but measurable by a camera sensor analyzing pixel intensity changes over time. That is the basic principle behind remote photoplethysmography (rPPG), and it works for perfusion assessment the same way it works for heart rate: by reading the optical signature of blood moving through tissue.
The perfusion index (PI) takes this a step further. Standard pulse oximeters calculate PI as the ratio of pulsatile (AC) blood flow to static (DC) blood in peripheral tissue. A high PI means good perfusion; a low PI means blood is not reaching the extremities well. The question researchers have been asking: can a camera replicate this measurement without touching the patient?
Xu et al. at the IEEE Engineering in Medicine and Biology Conference (2024) investigated exactly this. Their study used camera-PPG signals to calibrate perfusion index measurements through ice water stimulation experiments that induced controlled perfusion changes. The results were split. Personalized regression models achieved an R² of 83%, meaning camera-based PI measurement worked well when calibrated to an individual patient. But generalized models — the kind you would need for a plug-and-play clinical tool — showed negative R² values, meaning they performed worse than simply guessing the average. The culprit was skin property variation between subjects: melanin content, subcutaneous fat thickness, and skin texture all affect how light interacts with tissue.
This is an honest picture of where the technology stands. Camera-based perfusion measurement works, but it is not yet a one-size-fits-all solution. The path forward likely involves patient-specific calibration during an initial setup period.
| Assessment method | Contact required | Spatial coverage | Continuous monitoring | Clinical maturity |
|---|---|---|---|---|
| Manual capillary refill time | Yes (finger press) | Single point | No (spot check) | Established but subjective |
| Pulse oximeter perfusion index | Yes (finger clip) | Single point | Yes | Widely used in ICU/OR |
| Laser Doppler flowmetry | Yes (probe) | Single point | Yes | Research and specialty clinics |
| Laser speckle contrast imaging | No (but fixed setup) | 2D tissue area | Yes | Research stage |
| Camera-based rPPG perfusion | No | Configurable ROI | Yes | Early clinical validation |
| Imaging photoplethysmography (iPPG) | No | Full 2D perfusion map | Yes | Research stage |
Microcirculation monitoring in critical care
The clinical case for perfusion monitoring is strongest in critical care, where the gap between what systemic hemodynamics show and what is actually happening at the tissue level can be the difference between a patient recovering or developing organ failure.
The ANDROMEDA-SHOCK trial, published by Hernández et al. in JAMA (2019), was a landmark study that tested whether targeting peripheral perfusion — specifically capillary refill time — during sepsis resuscitation produced better outcomes than targeting serum lactate levels. The results showed a trend toward lower 28-day mortality in the CRT-guided group. The follow-up ANDROMEDA-SHOCK-2 trial, completed in 2025 and published in JAMA, went further: a personalized hemodynamic resuscitation protocol targeting capillary refill time in early septic shock demonstrated improved outcomes compared to usual care.
The problem with capillary refill time, despite its clinical value, is that it relies on a clinician physically pressing tissue and visually estimating the time for color to return. It is subjective, variable between observers, and impossible to perform continuously.
Marcinkevics et al. at Pauls Stradins Clinical University Hospital in Riga addressed this gap directly. Their 2024 study, published in Medicina, used rPPG alongside automated capillary refill time (aCRT) to assess microcirculation in 20 ICU patients with sepsis and septic shock. The study compared bacterial sepsis against COVID-19-associated sepsis during fluid resuscitation. The results: rPPG was more sensitive than aCRT in detecting perfusion changes between groups, and aCRT measured by camera was more sensitive than manual CRT assessment performed by clinicians. The bacterial sepsis group showed higher initial lactate levels and more pronounced microcirculatory dysfunction, and the camera-based tools picked up these differences that manual assessment missed.
This is where contactless monitoring adds something that contact-based devices cannot easily replicate: continuous, objective, spatially distributed perfusion data without requiring a clinician at the bedside.
Peripheral vascular disease screening
Peripheral arterial disease (PAD) affects an estimated 236 million people worldwide, according to the Global Burden of Disease study. In the United States, approximately 8 to 12 million adults have PAD, with prevalence climbing sharply after age 60. The disease narrows arteries supplying the extremities, most commonly the legs, reducing blood flow and increasing the risk of amputation and cardiovascular events.
PAD screening matters because the disease is frequently asymptomatic in its early stages. The ankle-brachial index (ABI) remains the standard screening tool, but it requires a trained operator with a Doppler probe and takes several minutes per patient. In primary care settings, where PAD screening would have the most impact, ABI testing is often skipped.
PPG waveform analysis offers a simpler alternative. A 2022 study in Physiological Measurement evaluated the systolic rise time — the interval between the onset and peak of the PPG pulse wave — as a screening parameter for PAD. In diabetic patients, the systolic rise time measured by PPG at the toe showed meaningful differences between those with and without PAD. The approach requires only a simple optical sensor, making it far more accessible than Doppler-based ABI.
Camera-based plantar perfusion imaging (CPPI) takes this concept further by removing the sensor entirely. A 2024 proof-of-concept study proposed using a standard camera to image the plantar surface of the foot, extracting perfusion maps from the video to screen for PAD. The logic is sound: PAD primarily affects blood flow to the feet, and perfusion imaging of the foot sole captures exactly where disease impact shows up first. While still at the proof-of-concept stage, CPPI represents a path toward PAD screening that could be performed with a smartphone during a routine office visit.
Diabetic foot and wound assessment
Diabetic foot complications are among the most costly and debilitating consequences of poor peripheral perfusion. Approximately 15% of people with diabetes will develop a foot ulcer during their lifetime, and inadequate perfusion is a primary reason these wounds fail to heal.
Imaging photoplethysmography (iPPG) has been validated against laser speckle contrast analysis (LASCA) for measuring plantar skin perfusion. A 2021 validation study published in the Journal of Clinical Medicine found that iPPG measurements at the plantar surface correlated with LASCA readings, supporting its use as a perfusion assessment tool for diabetic foot management. The ability to map perfusion across the entire foot surface, rather than checking a single point, gives clinicians a spatial view of where blood flow is adequate and where it is compromised.
Lai et al. demonstrated that iPPG perfusion maps can detect skin perfusion perturbations including temperature changes and irritation responses. For wound management, this spatial perfusion data could guide decisions about debridement margins, off-loading strategies, and the likelihood of healing — all currently assessed through clinical judgment rather than objective measurement.
The signal processing behind perfusion imaging
Extracting perfusion data from video is technically more demanding than extracting heart rate. Heart rate measurement relies on detecting the dominant frequency in the pulsatile signal — a relatively clean periodic pattern. Perfusion assessment requires quantifying the amplitude of that pulsatile component relative to the baseline signal, which introduces sensitivity to motion artifacts, ambient lighting changes, and tissue heterogeneity.
Pirzada et al. at the University of St Andrews, in a 2024 review published in IEEE Sensors Journal, catalogued the signal processing pipeline for rPPG systems: face detection and region-of-interest selection, color channel extraction, signal filtering, and physiological parameter estimation. For perfusion applications specifically, the AC/DC ratio of the signal is the key metric, and maintaining stable DC baseline measurement is harder than tracking AC periodicity.
Deep learning approaches are improving this. Debnath and Kim (2025), reviewing 145 articles in BioMedical Engineering OnLine, found that deep learning methods consistently outperform traditional signal processing for rPPG parameter extraction. Neural networks can learn to separate perfusion-related signal components from noise without hand-engineered filter parameters, and they handle the skin tone and lighting variations that trip up conventional algorithms.
A 2024 study published in Biomedical Signal Processing and Control examined 0.1-Hz vasomotion — slow oscillations in microcirculation regulated by the autonomic nervous system — using iPPG under local heating conditions. Detecting these low-frequency perfusion oscillations through a camera opens up assessment of microvascular reactivity, a parameter that currently requires laser Doppler equipment.
Where camera-based perfusion monitoring is heading
The technical trajectory is toward multi-parameter spatial perfusion mapping from a single camera. Instead of measuring perfusion at one fingertip, a camera can simultaneously assess perfusion across the face, hands, or feet, generating heat-map-style visualizations of blood flow distribution. This spatial information is something no contact-based sensor can easily provide.
Clinical integration will likely start in settings where perfusion monitoring already happens but is labor-intensive: ICU resuscitation guidance, surgical flap monitoring, and diabetic foot clinics. In each case, the camera replaces a manual or single-point measurement with continuous, spatial data.
The calibration challenge identified by Xu et al. needs solving before generalized clinical tools emerge. Subject-specific skin properties affect the AC/DC ratio in ways that are not yet fully characterized. Approaches being explored include brief calibration protocols at the start of monitoring, machine learning models trained on diverse populations, and multi-wavelength imaging that can account for melanin and tissue thickness variations.
Circadify has developed camera-based vital sign measurement technology that captures heart rate, HRV, respiratory rate, and SpO2 from a standard device camera. As the evidence base for camera-based perfusion assessment grows, the underlying rPPG signal processing that powers these measurements is the same foundation that perfusion monitoring builds on.
Frequently asked questions
Can a camera measure peripheral perfusion?
Camera-based photoplethysmography detects blood volume changes in skin tissue by analyzing subtle color fluctuations in video. These signals correlate with perfusion levels in superficial vascular beds. While personalized calibration models show strong accuracy (R² up to 83%), generalized models remain unreliable across different individuals due to skin property variations, meaning the technology works best when calibrated per patient.
What is the perfusion index and why does it matter?
The perfusion index is the ratio of pulsatile to non-pulsatile blood flow in peripheral tissue, typically measured by pulse oximeters. It reflects how well blood is reaching extremities. Low perfusion index values can indicate peripheral vascular disease, poor cardiac output, or shock. Camera-based systems are being investigated as a way to measure perfusion index without attaching sensors to the patient.
How is microcirculation monitoring used in clinical care?
Microcirculation monitoring helps clinicians assess tissue-level oxygen delivery, which systemic hemodynamic measurements like blood pressure can miss. In critical care, the ANDROMEDA-SHOCK trials demonstrated that targeting capillary refill time during sepsis resuscitation reduced mortality. Camera-based systems offer a way to continuously track perfusion changes that currently require manual bedside assessment.
Can camera-based perfusion imaging detect peripheral arterial disease?
Early research suggests camera-based plantar perfusion imaging can differentiate between healthy subjects and those with peripheral arterial disease. PPG waveform analysis, including systolic rise time measurements, shows promise for general PAD screening, particularly in diabetic populations where routine vascular screening is recommended but often skipped.