CastMinder - The Cast and Splint Monitoring System

Portions of CastMinder are more useful in some cases than in others. A great illustration of this is a buildup of pressure. A patient who has broken her arm but is otherwise healthy and walking around would be immediately able to sense when the pressure in her cast is getting too high, as she would feel pain. However, a patient who is unconscious in a hospital would not be able to communicate such problems. Thus, having a sensor inside of such a patient's cast can prove immensely useful for physicians.

A moisture sensor is useful for most patients. Oftentimes it can be quite difficult to detect the presence of moisture inside of a cast.

The active healing unit is obviously useful for any patient, as an increase in bone formation and a decrease in pain would be beneficial to anyone.

I've only recently posted the CastMinder code in full on GitHub. This is split between two repositories, the repository for the Arduino code and the repository for the iOS application. Most of the code for the Arduino portion is quite simple - after all, all these nodes are doing is collecting data, sending it, and going to sleep. The iOS app is where things start to get a bit more complicated.

There's thousands of lines of code that I'm working hard to document in the CastMinder iOS app. There's also a lot of cleaning up that I need to do before I'm really proud of the code I produced. Thus, what I have posted on GitHub is by no means a finished product, but it should work. If you have any suggestions please do not hesitate to contact me at info[at]coniferapps.com.

I recently produced a video that shows some of the components of the CastMinder system in action. If you're curious about what the app actually looks like in real life, you can check out the video below.

Soon I'll have build instructions and a formal component list available for the active healing unit. Here's a preliminary and general list:

• The first component is a LigthBlue Bean. This allows the device to be controlled from the CastMinder app via Bluetooth
• A 9v battery is used to power the whole system
• A step-down voltage converter is used to provide the Bean with a nice steady 3v (no CR2032 needed!)
• A step-up voltage converter is used to provide power to the electrodes.
• Rather than controlling the input to the step-up converter, I'm using a power MOSFET to control the output. This way I don't need to worry about charging and draining the capacitors in the step-up converter which can mess with the frequency of the output

The Transcutaneous Electrical Nerve Stimulation (TENS) unit is an integral part of the CastMinder Active Healing system. Regular TENS units work by providing relatively high voltage (approx. 50v) pulses to the skin through electrodes. These pulses are generally in the 1-10hz range, or slow enough that a human can differentiate between the different pulses.

I'm currently working on getting the correct pulse width dialed in. This will involve lots of trial and error, as It'll just be a matter of seeing what lessens pain the most. The pulse width and timing is also somewhat subjective, as different people respond in different ways.

These pulses stimulate blood flow around the affected area, lessening patient pain. My Active Healing unit combines a TENS unit with a Bone Growth Stimulator, which I'll talk about in another log.

One of both the benefits and drawbacks of the CastMinder system is its tendency to produce data. A lot of data. One sensor log per second from up to fifteen sensors means thousands of sensor logs an hour. One of the questions I was faced with while designing CastMinder was how can I use this data to help detect complications?

I've developed a few solutions. Lately, I have been experimenting with a great GitHub library called Swift AI that makes it easy to add Machine Learning to iOS apps. I can use the GPU-accelerated algorithm to parse thousands and thousands of data points, then develop correlations between them to predict future cast conditions.

Currently, CastMinder has been tested for over 210 hours. This is a lot of testing data, but these tests were only performed on healthy testers--testers with no broken or fractured limbs. After the entire system is developed further, we'll begin testing CastMinder on individuals with real broken bob

The CastMinder Bluetooth system currently relies on the Lightblue Bean as the primary Bluetooth chip and microcontroller. The Bean is great at saving battery, but it can only go so far on a coin cell. A CR2032 lasts about a month and a half with optimized Bluetooth settings. I could decrease the packet transmission time to get it to last a bit longer, but that would be at the expense of real-time data analytics. An easy solution is to include a bigger battery. I've gotten my device to last upwards of 3 months on bigger coin cells.

My RF24 system is much more power efficient as it uses a lower-power radio to communicate, and doesn't advertise as frequently. This system can last almost half a year on a single CR2032 battery.

CastMinder is currently patent-pending, a distinction afforded to it when I registered and filed a provisional patent for the project. I'm currently in talks with lawyers to get full patent protection.

Recently I've been developing two separate CastMinder systems. The "smart" system is the one I describe in the most detail. It consists of Bluetooth sensor nodes and an iOS application. However, I've also been working on a "dumb" system, which uses RF24 chips to talk to a receiver. The dumb system does have certain advantages, such as the fact that the battery lasts much longer, and it costs almost 4 times less than the Bluetooth system, not even including the cost of a phone.