Diabetic patients use therapy from insulin pump, a type of implantable medical device, for the infusion of insulin to control blood glucose level. While these devices offer many clinical beneﬁts, there has been a recent increase in the number of cases, wherein, the wireless communication channel of such devices has been compromised. This not only causes the device to malfunction, but also potentially threatens the patient’s life. Neural Network based, real-time deep learning classiﬁer (speciﬁcally the Multilayer Perceptron Model) was designed for wireless medical device security. Deep learning is among the most effective and broadly utilized systems for classiﬁcation, identiﬁcation, segmentation etc. Although they are effective, they are both computationally and memory intensive, making them hard to be deployed on low-power embedded frameworks. We present an on-chip neural system execution for securing diabetic treatment. The model achieved 98.1% accuracy in classifying fake vs genuine glucose measurements. The proposed model was comparatively evaluated with Support Vector Machine which achieved only 90.17% accuracy with negligible precision and recall.
Professionals who make and use medical devices focus their ingenuity on serving patients. Persons on the dark side of the industry focus their ingenuity on abusing those devices for criminal ends.
Deep learning provides new algorithms to counteract this type of crime. Device edge processing is a fairly novel paradigm in which much of the processing takes place at the edge of the network (not in the network core).
Field Programmable Gate Array (FPGA) is proving out to be a promising hardware computing platform for deep learning algorithms, which can accomplish extensive performance while essentially enhancing energy proﬁciency contrasted to product equipment.
Edge computing for real-time local data analysis We implemented a framework in LabVIEW that allows us to train the Neural Network in TensorFlow and inference it on NI myRIO NI myRIO (FPGA) provides an ideal prototyping platform with capabilities to test the performance of the algorithm and resource utilization. Once this initial prototyping is completed and performance is validated, the FPGA co-processing can be migrated to a SoC.
As the name implies, a Multi-Layer Perceptron (MLP) is a perceptron with multiple layers. Those layers form a network that can be allowed to share information through what is called forward and back propagation. In fact, multiple instances of a single perceptron exist in an MLP.
We manage data preparation as follows…
• We adapted algorithms from the TensorFlow library to craft a new system for training and inferencing…
• We took advantage of the several unusual features in the NI LabVIEW development environment including these:
• Exploring the frontiers of chip-deployed inferencing in real-world conditions of the Internet of Things.