Connected Consumer Healthcare: An IoT-Enabled FBG Vibration Sensing and Machine Learning Framework for Predictive Maintenance of MRI/CT Systems

The increasing number of connected consumer healthcare devices is generating the demand for smart, trusted, real-time diagnostic support. We introduce an IoT-based framework that will be used to couple Fibre Bragg Grating (FBG) vibration sensors with modern machine learning (ML) algorithms to predictively maintain MRI/CT systems to stabilise the imaging capabilities and positively improve consumer confidence. FBG sensors, which are not subject to electromagnetic interference and are also beneficial compared to standard piezoelectric sensors, are used to record vibration signals, which are then denoised and transformed into diagnostic features via filtering, Fast Fourier Transform (FFT) and Discrete Wavelet Transform (DWT). The dataset contains around 864x105vibration samples collected during the 24 hours of work of MRI/ CT systems in a healthy and a faulty condition. The Fluvio producer pipeline shows that there is scalable streaming to consumer interfaces, i.e. smartphones and IoT dashboards. We consider the models of the Random Forest (RF), Support Vector Machine (SVM), Deep Neural Network (DNN), and Recurrent Neural Network (RNN). RF performs the best, with an R2= 0.9865, an MAE = 0.0421, and an RMSE = 0.0613, where it is also resilient to noise (R2= 0.9743). It is also the shortest training (15.2 s), low inference latency (1.3ms per sample), and as high as 95 % classification accuracy with less than 5% error - the real-time requirement. The framework makes it possible to detect faults early, access them remotely, and perform predictive analytics on the attached devices, which is directly related to clinical outcomes: motion artifacts will decrease, better image quality, and more accurate diagnosis in MRI/CT. This solution is an improvement of the goals of Healthcare 4.0 as it enhances a dependable, scalable, and patient-friendly diagnostic ecosystem.