3.1. CW-NIRS Devices
For the initial phase of the HEMOCOVID-19 clinical trial (see
Section 2), 10 devices were required to launch the project in 8 clinical centers, and 2 devices were kept in-house at ICFO for continued quality control testing. Additional partners have joined the consortium since then. The following characteristics were required when selecting the devices for use:
Readily available for delivery within 30 days;
Total cost within the limited project budget;
Should provide both the trends and an estimate of the absolute value of the blood oxygen saturation;
Should be suitable for use at an intensive care unit with regard to the restrictions introduced by the COVID-19 pandemic, including features such as the following:
- –
Wireless/remote controlled;
- –
Disinfectable (with alcohol) between patients;
- –
Easy-to-operate with remote-training only;
- –
No disposable parts;
- –
Can be re-utilized without leaving the containment zone;
- –
Minimal footprint in contact with the tissue;
- –
Should come with customer support directly from the company during the pandemic lock-downs.
Should be suitable to use at the measurement site—the brachioradialis muscle of the forearm.
These constraints made instrument selection a challenge. For example, none of the devices in the market with a medical-device authorization (CE, FDA, or other equivalent approvals) met these requirements. Eventually, we decided to utilize PortaMon (Artinis Medical Systems, NL) [
30] devices for the clinical trial. This device found particular success during recent years and has been validated and extensively used, primarily in the sport/athletic field, to monitor local muscle oxygenation [
4,
22,
23,
24,
25,
26,
27].
Briefly, PortaMon is a portable, wireless, battery-operated CW-NIRS system consisting of three couples of light emitting diodes (LED) as sources at nominal wavelengths of 760 and 850 nm at different distances from the receiver (30, 35, and 40 mm). It is capable of continuous monitoring with a temporal resolution of 0.1 s, reporting the local tissue oxygen saturation index (TSI), oxy- and deoxy-hemoglobin concentrations (respectively,
and
), and the total hemoglobin concentration (THC). TSI is an index proportional to the more commonly utilized
. The absolute values are retrieved by using so-called spatially resolved spectroscopy and the modified Beer–Lambert law [
2,
4,
8,
31].
After the initial tests were conducted, we acquired 13 (with 1 device as an additional back-up) devices. Two additional devices were loaned by the manufacturer and were added to the tests after the initial order since (as described below) significant inter-device variability was found, with some units systematically out of the empirically defined acceptance range. In this manuscript, the devices are identified by their “id” number (id36, id38, id40, etc.). A photo of two devices is reported in
Figure 2a, showing both top and bottom views.
3.2. Tissue-Simulating Phantoms, Type 1
In our laboratory, a set of commercial tissue-simulating, homogeneous, solid phantoms (Biomimic Optical Phantoms, INO, Québec, Canada) was available. This type of phantoms is commonly utilized for testing diffuse optical devices as they provide a relatively well established prior knowledge of the optical properties, good homogeneity, and stability over long periods (years) [
32,
33].
For this part of the study, we utilized two phantoms (Type 1 phantoms): (1) Biomimic PB300 (INO PB300) with nominal values of reduced scattering coefficient
cm
and absorption coefficient
cm
at
nm, (2) Biomimic PB312 (INO PB312)
cm
and
cm
at
nm. For all the measurements, the phantoms were prepared with a custom mask, which assists in the repeatable and reliable placement of the device on the phantom surface (see
Figure 2b).
We note here that the wavelength dependence of the optical properties of these phantoms were not tuned to provide a specific or equivalent value. Since our goal was not to validate the absolute values of these parameters with these phantoms, we took an approach whereby we used the mean value from all devices as the expected nominal value. Two phantoms were utilized to provide two different measured light levels, partially testing the devices’ dynamic range.
Figure 2.
Experimental setup. (a) Two PortaMon (Artinis Medical Systems) devices used in this study; top and bottom view. (b) Biomimic PB312 phantom, with the custom mask for PortaMon placement. (c) One of the BioPixS phantoms, together with the custom mask (left panel, top view) and the PortaMon placed (right panel). (d) PortaMon placement for forearm muscle measurements. (e) Sketch of the VOT procedure together with a visual representation of the measured relevant parameters.
Figure 2.
Experimental setup. (a) Two PortaMon (Artinis Medical Systems) devices used in this study; top and bottom view. (b) Biomimic PB312 phantom, with the custom mask for PortaMon placement. (c) One of the BioPixS phantoms, together with the custom mask (left panel, top view) and the PortaMon placed (right panel). (d) PortaMon placement for forearm muscle measurements. (e) Sketch of the VOT procedure together with a visual representation of the measured relevant parameters.
3.3. Tissue-Simulating Phantoms, Type 2
After the initial tests, we observed that there was a systematic variability between different devices. This difference, as discussed below, is quite minimal, but since it has been detected and as its dependence on environmental conditions, the age of the device, hours of use, and other factors is unknown, we sought a solution that would involve utilizing identical phantoms alongside each system. As previously discussed, this is not a trivial problem since the manufacturing, characterization, and maintenance of such phantoms is a complex matter that is still being tackled by both academia and the industry [
28,
33,
34].
A further consideration was the availability of such phantoms at short notice and their cost-effectiveness. The most suitable candidate was identified to be the devices produced by BioPixS (Cork, Ireland,
www.biopixstandards.com, acessed on 6 September 2021). Ten (nominally) identical phantoms (BioPixS-Matrix-CCB5d, Type 2) from the same class of solid phantoms as Type 1 (see above) phantoms (
cm ) that were manufactured from the same batch of materials with a nominal reduced scattering coefficient
cm
and absorption coefficient
cm
at 740 nm were produced and characterized by the manufacturer. The declared inter-phantom difference is 0.8% for the absorption coefficient and 1.4% for the scattering coefficient at 740 nm.
As for the Type 1 phantoms, we prepared a custom mask for reliable positioning on Type 2 phantoms as shown in
Figure 2c.
3.4. In Vivo Measurements
All the in vivo studies were conducted according to the guidelines of the Declaration of Helsinki and approved by the local Ethics Committee. Subjects were asked to provide informed consent.
In these protocols, the goal was to evaluate the repeatability and variability during the resting condition for each device and also the variability of the vascular occlusion test (VOT) with a single device. To evaluate the repeatability and variability during the resting condition, the forearm brachioradialis muscle of the same subject (age: 26; gender: male) was measured. To reduce the variability to physiological changes in the muscle, the subject was at rest sitting on a chair, with the arm resting in a stable position on the arm of the chair.
The vascular occlusion test protocol consisted of continuously monitoring the
during a baseline period of three minutes, a period of complete arterial occlusion of three minutes, and a period of recovery of five minutes. Arterial occlusion—i.e., ischemia—was induced by inflating a typical arm blood pressure cuff placed on the biceps at a pressure of 50 mmHg above the systolic pressure of the subject. This protocol was repeated on a separate subject for 20 different days during the same month, on the same healthy subject (age: 41; gender: male) in the supine position. A sketch of the measurement procedure is reported in
Figure 2e together with the visual representation of the relevant parameters that were evaluated:
baseline, deoxygenation slope—
, reoxygenation slope—
, and hyperemic response—
.
is the area under the hyperemic peak.
is calculated by linearly fitting the first minute of the curve
vs.
during the occlusion period.
is calculated by linearly fitting the same curve from the instant the occlusion is released up to the instant the
returns back to its baseline value.
3.5. Description of Tests
A summary of all the tests performed is reported in
Table 1. As mentioned above, the aim of these tests was to assess the performance of the devices in terms of the variability and reproducibility of the measurements when performed under different conditions. The tests were devised to match the overall need to acquire data from multiple locations with different devices, by different operators and over a minimum time-span of a year. Further considerations included that fact that the tissue of interest—a muscle—is soft; therefore, the hemodynamics are affected by the probe pressure but cannot be controlled with a quantitative feedback mechanism in this particular case. In the following sections, we describe each test in detail.
Table 1 uses capital letters as an identifier for each test.
We began by evaluating the the warm-up time of the device (Test A) and stability of and over eight hours of continuous acquisition starting immediately after turning the device on. Type 1 phantoms were utilized in this test and three different devices were utilized. We opted not to use the whole set of devices since the results from the first three were quite similar and due to the urgency of starting the clinical trial. The goal of this test was to provide instructions for the user about the use of the device. It was important to identify this point and if there was any variability between different devices, since, as a battery-operated device, it could not be switched on continuously. The warm-up time was evaluated by assuming that the measured and would stabilize around a mean value once the device was ready for use.
All the subsequent tests (B to F) also used Type 1 phantoms where each measurement consisted of five subsequent acquisitions of 20 s each, after removing and rapidly replacing the device in position. This measurement was repeated for different devices, by different operators, and in different periods (different days, different months). In addition, to test the variability at different times of the same day, we repeated one measurement with one device at the beginning and at the end of the same day.
Furthermore, as indicated in
Table 1, several of these tests (
) were also performed in vivo on the
brachioradialis muscle, in the forearm, as detailed above. Each device was placed carefully, as would be done during the clinical trial, and covered with a black bandage to avoid background light, as shown in
Figure 2d. For each device, four subsequent single acquisitions of 20 s each were repeated by rapidly removing and replacing the device in the same position. This procedure was repeated two times (two sets of four subsequent acquisitions per device) by randomizing the order of the devices. This precaution was taken in order to avoid that changes of muscle hemodynamics over time would not impact the same device twice in the same manner.
We also characterized (Test G) a set of 10 phantoms of Type 2 (see above) to evaluate their potential to be utilized with every device for on-site quality control on a day-to-day basis. These phantoms were characterized with one PortaMon (id50) on three different days. Each day, five subsequent acquisitions of 20 s each were performed after removing and replacing the device rapidly in the same position.
Lastly, we tested (Test H) the reproducibility of VOT; that, is the measurement protocol of the HEMOCOVID-19 clinical trial as described above. This protocol was repeated with a single device (id63), once per day, for 20 different days of the same month, on a single healthy subject laying supine for at least 30 min prior to measurement. Please note that this device (id63) was acquired independently by a HEMOCOVID-19 partner and is not reported in the other tests.
For all the tests, the variability and reproducibility of a measurement and the differences between devices were evaluated by calculating the coefficient of variation , defined as , where x is the measured quantity (i.e., and ), is the standard deviation, and is the average.