Big data is used for lots of different things. Marketers use it to target specific groups of consumers. Financial institutions use it to track customer transactions and predict future trends. Even medical science puts big data to use. Above and beyond things like electronic record-keeping, big data is being harnessed to improve medical device development.
Medical devices before the age of electronics were mechanical devices with limited benefits. When electronics and computer technology were introduced to medical devices a few decades ago, we saw a corresponding jump in their usefulness. Now we are taking that next leap by integrating medical device development with big data.
How does this help the average patient? According to Rock West Solutions, a California company specializing in signal processing and big data for medical device development and applications, patients benefit by way of new and improved medical devices capable of helping them lead healthier lives.
Improving Seizure Prediction with Data
A good example of how big data is improving medical device development is found in a study published by eBioMedicine earlier this year. The study related to the development of better predictive analytics for helping epileptics predict and deal with seizures. Clinicians already use random predictors for this purpose, but such predictors are not all that reliable given the unique nature of epileptic seizures.
Researchers hoping to come up with a better means of predictive analysis turned to deep learning for a solution. They came up with a predictive system that combines individual user data with much larger random data sets that can be applied to machine learning to produce better results than random prediction is capable of.
Testing their system was a matter of coming up with a predictive system that could be fine-tuned by patients. Over a predetermined amount of time, seizure data was collected along with data explaining how patients tuned their systems. The data was then combined with larger data sets and fed into a machine learning system.
The results of the machine learning process were deployed on a wearable device utilizing a low-power computer chip. Researchers were pleased to observe a mean sensitivity for their device of 68.6%. More importantly, their device outperformed a comparable random predictor by more than 41%.
Researchers say that clinicians can use their predictive system to customize sensitivity and alarm settings to individual patients based on patient fine-tuning. This is something that random predictive systems cannot do. Customization allows for a level of personalization that can make epilepsy a lot more manageable.
Deep Learning and Signal Processing
Rock West Solutions explains that the key to making all this work is signal processing. Deep learning is an excellent tool for predictive analysis, but only to the extent that it has the right data to work with. Therein lies the inherent weakness of big data.
In its purest form, big data does not concern itself with how large data sets are used. Its main goal is to harvest and store as much data as possible. It is up to those who need use of the data to figure out what to do with it. Thus, raw data and machine learning do not necessarily produce the best results when combined.
On the other hand, signal processing is the science of analyzing big data for the purposes of isolating only the data that is valuable for a particular purpose. Signal processing makes it possible to extract the data necessary for deep learning to occur.
Big data is already improving medical device development considerably. Who knows what researchers and developers will be able to produce in the future?