August 8, 2014
Dear Indiegogo Backers,
Since our last update, we’ve been focused on working through the technical issues with the Scanadu Scout™ investigational devices. We know many of you are hungry for an update on what’s been happening at Scanadu HQ. We’d like to share with you some color on our quest to bring the Scanadu Scout™ to life and put it in your hands.
RECAP: Diagnosing the Issues
The Scanadu Scout™ was originally set to ship in March, however in April, we had to pause shipment because the units that rolled out of manufacturing did not meet our standards of accuracy. To diagnose the situation, we pulled out all of the guts of the device – the sensors, the interconnects, the components – to identify the exact sources of the problems. We had some assumptions, but needed to test each detail to be sure.
We invited you to our headquarters to donate data in May and reported our initial results from the Scanathons in June. We were excited to meet with so many of you! As we mentioned in our latest blog post in July, finding the ultimate solution would require a significant amount of work and testing. Thanks to the participation and the data from many of you, we were able to confirm most of our suspicions about what was wrong.
It turned out that some last minute design decisions earlier this year about the placement of the sensors resulted in them being moved by a few millimeters (particularly the temperature and SPO2 sensors). That very small decision had significant impact in corrupting a lot of the raw sensor data. Because the raw sensor data is the essence – the life blood – of the device, it polluted the entire system.
Here’s more detail:
TECHNICAL UPDATE: Straight from the Workbench
In the original version, accuracy of temperature measurements was affected by heat transfer between the electrocardiogram (ECG) electrode and the skin in cases when, just before the scan, the ECG electrode had been exposed to temperature levels significantly deviating from room temperature.
Several possible solutions to this problem were identified and each of these has been built as a “prototype” and is being tested for performance, accuracy, repeatability and manufacturability.
Above, Printed Circuit Board (PCB) variations being used for performance, accuracy, repeatability and manufacturability testing and evaluation.
We continue to make significant progress towards collection of a larger volume of data to support the training of our temperature algorithm and detailed validation of the temperature sub-system. The Scanadu clinical team has identified a location for additional data collection, where we will initially focus on gathering data over a wider temperature range.
Our March build of Scanadu Scout™ also suffered from parasitic light reflections around the photoplethysmography (PPG) sensor (because of the last minute design change we mentioned above). These reflections were caused by a flaw in the optical design and led to corruption of the PPG signals. These PPG signals also affect three other measurements including: blood pressure, SpO2 level and pulse rate.
In order to ensure that we completely understood the problem and had a definitive solution, we decoupled the hardware development from algorithm development. We made a prototype with PPG sensors exposed outside of the device and exposed the sensors directly to the skin, using this for collection of raw data signals on a large group of individuals. Reference levels of the vital signs were also measured. The data was collected in a number of Scanathon sessions as well as in the Hypoxia lab in San Francisco (to test SpO2). This gave us very clean sensor signals paired with the corresponding reference levels, which allowed development of revised blood pressure and SpO2 algorithms.
Below is the original version of the acrylic window on the left and a potential new solution on the right.
Our blood pressure algorithm relies on features found in the electrocardiogram (ECG) and photoplethysmography (PPG) signals. For instance, one such feature - Pulse Wave Transit Time - is a measure of time between contractions of the heart, identified from the R-wave of the ECG and the peaks of the peripheral pulse waves in the PPG.
Those signals are processed to remove noise and artifacts (or filtering). We had to redesign our PPG sensor and retrain the algorithms coming from those features extractions coming from both the ECG and PPG.
We are continuing testing and validating our algorithms and the quality of our readings with help from recent and upcoming Scanathons.
NEXT STEPS: Test, Test and Test Again
We are inviting our backers to come back to our headquarters and help us do more data collection to further verify our solutions (and to help us train our updated algorithms). Our next data collection is planned for Monday and Tuesday of next week. We‘re practically full because many of the backers who participated in previous Scanathons are returning, but there are a few spots still open: please register here to receive the formal invitation.
Our design team was ready with both iOS and Android apps last March but have taken advantage of the delay to further enhance the user experience and implement version 2 of the App. During our upcoming Scanathon we’ll ask our backers to test our new app on iOS as well as on Android devices.
Here’s a sneak peak of the second version of the app:
We still cannot give a definitive ship date - but we have developed potential solutions for the problems we have encountered and have an aggressive schedule for data collection, testing and validation. We are determined and hope to resume shipping of the investigational devices before the end of the year. Of course if you prefer not to wait, we are happy to return your donation if you email a request directly to email@example.com.
This has been an incredibly hard journey. We are putting our blood, sweat and tears (not to mention our vitals) into rigorously testing our solutions so we can deliver to you.
Thank you for your continued patience and support. If you have any additional questions, contact us at firstname.lastname@example.org.