Machine Learning Enabled IoT Embedded Platform for RealTime Anomaly Detection in Drug Delivery Devices

Abstract

The reliability of modern drug delivery devices such as infusion pumps and insulin pumps is critically dependent on accurate delivery of medication.However, despite advances in recent years, such devices are prone to failure with reported events including flow abnormalities, blockages, inaccurate dosage delivery and mechanical failure.These events have potential to cause serious harm to patients with risk of overdose or underdose.Current monitoring of such devices includes analysis of threshold-based alarms raised by the device, which often fail to provide early intervention and lack of intelligent detection of abnormalities.Therefore, there is a growing need for implementation of intelligent real-time monitoring in these devices to guarantee their reliability.This paper outlines an IoT-enabled, embedded machine learning implementation that monitors the operation of drug infusion equipment, in real-time, and alerts for potential anomalies such as flow anomalies or occlusions.This novel embedded implementation, operating on the edge of the network, can process in real-time the real-time, raw sensor information streams (i.e.flow rate, pressure and actuator signals) generated by infusion pump(s) for the high speed, on-line classification and thereby results in significantly reduced latency due to minimized communication requirements with cloud-computing architectures.The proposed system was tested on simulated data from sensors of infusion pumps and insulin pumps.The results of the experiments indicate a high accuracy of the system in detecting various types of anomalies, inclu ding occlusion and dosage errors.In addition, the system is capable of operating in real-time, detecting anomalies within a time frame of a few milliseconds.This is due to a low computational complexity of the model.The system is able to process large amounts of data with high precision and thus it can be used in various applications in the field of healthcare.The proposed system enhances patient safety by timely detection of failure of the drug delivery device, and helps in improving the monitoring precision and time to response of the clinical staff for treatment using smart healthcare systems employing IoT and edge intelligence for scalable implementation.