Several open-loop neuromodulation devices are now available on the market for indications including Parkinson’s disease, essential tremor, epilepsy, major depression, and chronic pain. Although they provide significant relief of disease symptoms, they also require substantial input and time from clinicians to tailor the settings to each patient. Closed-loop systems eliminate this step by completely automating data interpretation, with the output being the stimulation signal. Progress is already being made towards this goal, with systems featuring reduced human involvement currently being pursued – but with the first fully closed-loop devices only now starting to be launched.
Converting sensor data to therapeutic signals
One of the reasons why fully closed-loop systems are a tough challenge is that the real-time feedback needs to be supremely robust and reliable, in order to ensure device functionality and patient safety under all scenarios – and this requires a lot of work on the data-processing algorithms.
A major aspect of data-processing is how to turn the sensor data into the biomarker signal . However, that’s usually dealt with during the biomarker exploration phase of a project, so by the time that a closed-loop product is being pursued, it’s more a case of implementing that algorithm within the device’s software architecture. Nevertheless, there’s still scope for adjustment – for example, in how often a signal is sampled, and how often changes to therapy parameters have to made based on the signals being measured.
Another challenge is cleaning up the initial signal from the electrode or sensor array – which in real-world scenarios can show more variability than in a research setting. Analogue and digital filters, frequency domain analysis, and dimensionality-reduction techniques, among other methods, can be employed to remove the large stimulation signal, deal with electronic noise and artefacts, as well as correcting for signal drift. To help in this task, machine learning models can also be applied to classify events and so improve reliability.
The energy cost of data-processing
So far, so good, but each additional processing step comes with a cost – its energy usage and effect on battery life, and the complexity of the processing hardware.
One perspective is that carrying out all the processing within the implant is preferable, to avoid the extra hardware needed for signal transmission and receipt, and eliminate problems with signal latency or external connectivity. We’ve shown ourselves how machine learning can be embedded in an implant without exceeding the power budget, and in some cases, carrying out that processing could even save power, if it means that the therapy current is more effectively managed.
On the other hand, the benefits of processing the data on a wearable device or even in the cloud could outweigh the drawbacks. For example, off-device processing makes it possible to run more complex algorithms, and update them more easily (in contrast, changing firmware within an implant faces more hurdles). A further advantage is that the device manufacturer can gather large amounts of real-world performance data, including on ‘edge cases’, which can help inform future improvements to the algorithms. There’s also a strong case to be made for using a hybrid approach, which maintains rapid responses with onboard processing, while also transmitting data externally to allow in-depth analysis.
Read the e-book: Closing the loop in neuromodulation
What’s clear is that the final choices over algorithms – which ones are used, when they’re used, and where they’re used – vary according to the needs of the application. And at TTP we’re keen to bring our expertise to solving your challenges around the design of both the software and hardware of neuromodulation devices.
To find out more, download our e-Book on “Closing the loop in neuromodulation therapy”.
About TTP's Neurotechnology team
From proof-of-concept studies to manufacturing scale-up, TTP's dedicated neurotechnology consulting services can help you rapidly engineer advanced neuromodulation solutions, guiding you every step of the way. With our multidisciplinary team of engineers, scientists and human factors designers, you can hit the ground running. Combining deep expertise with a proven track record in end-to-end product development, we will help you create technologies and devices that push the limits of what's possible in neurotechnology. Find out how our neurotechnology product development team can help you start strong and finish ahead.
TTP's Neurotechnology team is part of a broader MedTech team. Learn more about TTP's approach to medical device design and development and our medical device consulting services.
.jpg)
.jpg)
