The final device was 3D printed using the Stratasys Dimension SST 1200es and made out of ABS plastic. The interior of the device was the painted black to diminish any background noise when taking the picture of the skin.
Verification
1.1) After constructing and training the initial python based statistical vector machine (SVM) it was used to conduct verification testing for the specification 1.1 as this algorithm would be the primary component within the final prototype design to allow for inflammation detection. In accordance with the original testing protocol for the verification of this specification the algorithm was evaluated by testing it on a test of skin images that the algorithm was not exposed to during training. The final test set used to test the algorithm consisted of 22 skin images that were derived from the original training dataset and set aside for testing. This dataset included both healthy and inflamed skin images to allow for testing on both image classes that the algorithm would have to identify. After evaluating the algorithm on the test set this component was able to correctly identify 85% of the images from the inflammation class. This verified that the algorithm component of the device met the criteria for specification 1.1.
1.2) For the second specification of our first requirement there was no verification testing done. For this specification we did not need to conduct testing to ensure that we meet a measurable criteria instead this specification involved design constraints for the design of the inflammation detection algorithm. The final algorithm incorporated into the device used images of the patient’s skin to access inflammation. The algorithm had as part of its design no storage of a patient’s personally identifiable data (PID) that ensured that no data leakage could occur exposing the patient’s information to security risk and allowed the algorithm to pass the criteria for this specification.
2.1) After ordering the parts, it turns out that the ROHM PSRESENSE BLE EVK 701 is not a bluetooth chip but rather a chip that makes another chip compatible with the Spresense board. This other chip is called the MK71251, which has two versions(Table A.1). The order came in with the MK71251-02 chip, which only supports peripheral mode BLE communication protocols. This meant that we couldn’t communicate with the bluetooth oximeter as in order to establish a bluetooth connection, a manager module has to initiate the connection to a peripheral bluetooth module.
Table A. The two microchips that are provided for the ROHM SPRESENSE bluetooth module.
| Product Name | Supported Roles |
| ML71251-01 | Peripheral or Manager |
| MK71251-02 | Peripheral Only |
2.2) Even though we couldn’t receive Oximeter data using the prototype, we were able to obtain recorded data using a phone app and sending it over G-Mail. This data was used to simulate the bluetooth connection and the threshold was tested using this method.
3.1 & 3.2) Verify that the size of the device will be incrementally changeable to accommodate the desired ranges (18.0 – 28.0 cm) by measuring the device’s length’s to the nearest 0.10 cm. The device after measuring was 30.0 cm. This measurement passed our verification protocol for ensuring that the device can accommodate patients of different sizes. In addition the strap is adjustable and can be incrementally changed to any length between 10.0 cm to 30.0 cm.
4.1) Calculate the power consumption of the device and its constituent parts. The device consumes approximately 0.0532 W (0.0032 W Sony board + 0.05 W LED). The battery can run the device for 1127 minutes ~ 19 hours.
5.1) Estimate the power output for the LED’s from the power delivered to it. The LED’s have a power output of approximately 158.4 mW. This was calculated by the set voltage of the device to power the LED’s of 3.3 volts and the measured current going through the LED’s of 0.048 Amps. We measured the current going through the LED using a multimeter and measured to the nearest 0.01 µAmp. Using these variables we were able to calculate the power output of the LED seen below.
5.2)After using the device for 30 minutes with the lights off, the subjects did not have any redness on their legs. Additionally, the power delivered to the LED light was calculated to be around 150 mW, which when we assume that it was distributed over an area of 18 Cm2, we get 8.8 mW/ cm2, which is lower than our passing criteria of 180 mW/cm2. Using this power we can also calculate an upper limit for the time it takes for the light to deliver enough energy to start changing skin properties. From the calculations done below, the time is around 15 minutes. By keeping the light off while not taking a picture, the energy delivered to the skin will never surpass the passing criteria.
time=EP=(8100 mJ/cm)(28.8 mw/cm2) = 920.45 seconds = 15.34 min
6.1) Verification testing for this specification involved artificially triggering the device alarm and measuring the time delay programmatically using the time function in python. The device successfully triggered an audible alarm when inflammation was detected. This alarm sounded within 1 second of the algorithm detecting inflammation following proper connection to the computer using MATLAB. The time to trigger an alarm after inflammation is read was 0.2 sec. The time required for this verification was the time between when an inflammation signal and when an alarm is heard. Since the alarm was triggered within an acceptable range this verification test was passed.
6.2 & 6.3) The DVT risk assessment portion of the algorithm was tested with different parameters for SpO2 and erythema threshold hold values that ranged between all possible values for each threshold based on the ranges and resolutions outlined in the specifications. However the prototype of the device did not include a UI component for inputting threshold values. So this verification testing was not possible due to the lack of UI but all possible thresholds values were tested and the algorithm performed as intended. This verification testing failed due to a lack of UI components for the device.
7.1) Measure the leakage current on the chassis of the prototype. This specification and verification testing were removed from the final set of design inputs for this device as there was not enough theoretical maximum current passing through the device to justify testing for the risks of leakage current.
8.1) The device will be tested to ensure that any object > 50 mm and water splashing against the enclosure will not penetrate the device for a minimum of 10 minutes. The device was not tested to meet the IP14 guidelines. This is primarily due to the fact that we believed the device would not pass such a test due to the risk of damaging the device from the water. The speaker holes and the device make it difficult to test against waterproofing. In addition to the camera being exposed underneath the device, the camera can be interfered with and cause distortions with the quality of the images as well as damage the device/patient.
8.2) Strap must survive 10 wash cycles at least. This was not tested as it was seen that the device can no longer pass the IP14 guidelines. This was not tested.