At SENSIMAT we are creating the first and only mHealth application for pressure relief, designed specifically for wheelchair users. We have had the device sending and receiving data for quite a while, but this is the first time we have put it to use for an extended period of time in real-world settings. We collected a huge amount of data throughout the week – 3,628,800 data points to be exact. Although this week has highlighted the fact that we have lots of testing left to do, this is a pretty big milestone for us as it marks the first time we have populated a database for analysis and damn, it’s exciting!
Pressure mapping and analysis is really nothing brand new – binary relief detection has been going on as far back as the early-80s. So why do this? The uptake of smartphones plus the ability for Bluetooth Low-Energy to constantly stream data allows us to collect data outside of the clinic during a wheelchair user’s actual daily routine. We performed a full search of the literature (we will be publishing a systematic review later this year) and found that this is uncharted territory. This data has already helped us to develop specific algorithms to warn wheelchair users when they are at risk of developing a pressure ulcer.
Lastly, it is important to note that this is in no way a properly controlled scientific study, but it gives us some great insight into seating pressure in real-world situations.
Figure 1: SENSIMAT hardware alone, ROHO Cushion Alone, SENSIMAT hardware inserted underneath the air pockets of the ROHO cushion – enclosed in the zipped-up fabric.
I took the SENSIMAT hardware and inserted it underneath a ROHO Dry Floatation Cushion. This is actually a fairly high-end wheelchair cushion that has been specifically designed to reduce the risk of pressure build-up. The Operations Manual shows that this is indicated for users who are:
Wherever I went during the week and wherever I sat down, I put this cushion down first. I also kept a physical diary log of all my actions during the week, so that we can provide some context to the data. This blog is not intended to be a cushion review but I felt that I should state that the cushion felt pretty good to sit on. When I sit straight it feels quite comfortable, soft, and stable. However, when I moved into more flexible positions, for example heavily leaning on one side with one foot up on an ottoman, it can feel a little odd. I connected the SENSIMAT hardware to my iPhone 4s using Bluetooth, which was a seamless process. I experimented a little bit with the range and found that the Bluetooth connection didn’t cut out even at opposite ends of the house, about 60 feet apart. The iPhone was connected through my wifi and pushed data to our servers at prescribed intervals. I didn’t actually interact with the mobile app for this experiment at all – I just left the SENSIMAT app running in the background, let it collect data, and tried to act as normal as possible. It’s also important to note that there didn’t seem to be any significant effect on battery life for the iPhone, which was a huge concern for us.
Figure 2: Random 90 minute timeframe from the database
Let’s look at some data! I pulled a random 90 minute excerpt from the database to do some analysis. Matching this with my diary shows that I was using my Xbox 360 sitting on a completely wood rocking chair, meaning that the ROHO cushion was sitting on a flat, hard surface on the bottom. It also means that I rocked a little bit, which can be seen by small fluctuations in the data. Since the rocking chair has a slight incline, my feet did not sit flat on the ground, meaning that my legs had a little bit of “hang” that put extra pressure on the front. I also occasionally put my foot up onto the coffee table in front of me. The very first thing that I noticed with this experiment is that seating is amazingly complex and measurement can be difficult. I found that I rarely sit straight – I lean side to side, cross my legs, lean forward to take a sip of coffee, etc. Each different seating position drastically changes the amount of seating pressure to various locations.
Each letter on the graph corresponds to the following positions:
– A to B – leaning back, rocking
– B to C – leaning back, left foot on coffee table
– C to D – leaning back, right foot on coffee table
– D to E – no entry
– E to F – sitting straight
There are numerous pros and cons to this graph. On one hand, it’s quite detailed and paints a nice picture of the way that I was sitting on the SENSIMAT. If I had not kept a diary, I still can estimate the seating position simply by looking at the differences between sensors. This also gives us a great database that we can retroactively perform analyses on. On the other hand, there is such thing as data overkill, and the data we have is quite “noisy”. If we think about how we are going to use this in wheelchair users, it will be perfect for detecting seating position, but quite difficult to tell where a deliberate pressure relief has occurred.
Figure 3: Figure 2 data adjusted for noise, combining sensors using moving averages and switching it from absolute sensor values to relative sensor values
After taking the previous data and eliminating some noise, we have a graph that is much easier to interpret: If the line is up, seated; if the line is down, not seated. This effectively turns the graph into an ON/OFF switch. This makes our data much cleaner and gives us a much easier time trying to find the number of pressure reliefs. However, we lose the ability to figure out seating position – all we know is whether the user is on or off the seat.
Detecting Pressure Relief
Looking back at Figure 3, how do we detect a relief of pressure? Well, looking at each sensor individually gives us some interesting numbers – a sensor was completely relieved 8390 times over the 90 minute interval, or about 28% of the time. It will be extremely interesting to compare this with a spinal cord injury patient. However brief those reliefs are, it still indicates either a shift in weight or a reposition. There was also very low variability in sensors hitting zero – range of 27.29% to 28.22%. Generally, a sensor is not relieved on it’s own, but in combination with a few other sensors.
While this is great information, what we really need to detect is a deliberate pressure relieving exercise. We can do this by scanning for at least 60 seconds of seated pressure, followed by at least 10 seconds of relieved pressure, which mimics the act of doing a pressure relieving exercise in a wheelchair. So how many of these pressure reliefs did I perform?
For context, many occupational therapists recommend that their spinal cord injury patients perform a pressure relieving exercise every 15-30 minutes, for anywhere from 10-30 seconds. My five exercises equals about one every 18 minutes. It is important to note that I was not deliberately relieving pressure, simply acting naturally. Due to the lack of seat-sensation in most spinal cord injury patients, this needs to be a deliberate action.
This was a great test – we collected some useful data that will help guide us forward. However, we still have lots of work to do. This test was in a single cushion, and we are unsure of the effect each specific wheelchair cushion will have on the sensors. Not only that, but we are also unsure of the effect the “bottom” has. I tested it on sitting on a leather chair, fabric chair, wood chair, and wood rocking chair and found slight differences – how will this change depending on the wheelchair base (ie. The “hammock” fabric of a folding wheelchair vs a flat base). Lift-offs are generally easy to detect, especially if done properly and held for extended periods of time – but what about patient’s who don’t have the ability to perform lift-offs perfectly? What about quadraplegics who use the wheelchair’s tilt function? What about users who can’t move themselves but require a caregiver? Things will get really interesting when we start testing in with wheelchair users who have spinal cord injuries.
SENSIMAT Systems Inc.
910 Rowntree Dairy Rd. Unit 13
Canada L4L 5W5