Uncertainty-Aware Environment Simulation of Medical Devices Digital Twins

TL;DR

This paper introduces EnvDT, a model-based approach for simulating uncertain environmental factors in medical device digital twins, enhancing testing accuracy and scenario diversity in IoT healthcare applications.

Contribution

It presents a novel EnvDT method for modeling environmental uncertainties in medical device digital twins, improving testing coverage and scenario diversity.

Findings

Achieves 61% environment model coverage

Generates diverse uncertain scenarios with a diversity value of 0.62

Validated on three real-world IoT medical devices

Abstract

Smart medical devices are an integral component of the healthcare Internet of Things (IoT), providing patients with various healthcare services through an IoT-based application. Ensuring the dependability of such applications through system and integration-level testing mandates the physical integration of numerous medical devices, which is costly and impractical. In this context, digital twins of medical devices play an essential role in facilitating testing automation. Testing with digital twins without accounting for uncertain environmental factors of medical devices leaves many functionalities of IoT-based healthcare applications untested. In addition, digital twins operating without environmental factors remain out of sync and uncalibrated with their corresponding devices functioning in the real environment. To deal with these challenges, in this paper, we propose a model-based approach (EnvDT) for modeling and simulating the environment of medical devices' digital twins under uncertainties. We empirically evaluate the EnvDT using three medicine dispensers, Karie, Medido, and Pilly connected to a real-world IoT-based healthcare application. Our evaluation targets analyzing the coverage of environment models and the diversity of uncertain scenarios generated for digital twins. Results show that EnvDT achieves approximately 61% coverage of environment models and generates diverse uncertain scenarios (with a near-maximum diversity value of 0.62) during multiple environmental simulations.