As we age, the number of pills we have to take increases. It is thus a growing challenge for the elderly to remember what pills they have to take. Pill ML is conceived to be connected to the elderly patient’s healthcare provider. The healthcare provider (which could be the primary doctor or community nurse) could then input the type of pill the patient needs to take. This input determines the correct pill for the patient to take, which the algorithm recognises from its training sets. When the patient is supposed to take the pill, the patient places the pill under Pill ML’s electronic camera to check if the patient is consuming the right pill.
The current prototype is limited to identifying whether the pill is correct. Future versions of the software are envisioned to include features like
– Notifying the user to consume different pills at different times
– Identifying different types of pills within a mixture of pills
– Recognizing a “pills have been consumed” state
The algorithm was derived from a flexible classification program written on P5js by Andreas Refsgaard. Modifications included the addition of a state machine, introduction of an interface for a healthcare provider to input the type of pill, and also the training of the algorithm to recognise different types of pills.