Smart Sensor Network Architecture with Machine Learning-Based Predictive Monitoring for High-Complexity Computed...

Area of Science:

• Medical Imaging Systems
• Machine Learning Applications
• Predictive Maintenance

Background:

• Ensuring operational reliability in high-complexity medical imaging systems like Computed Tomography (CT) scanners is challenging.
• Unexpected downtime in advanced CT platforms necessitates improved monitoring solutions.
• Integrating distributed sensing and data-driven analytics offers a path to enhanced system reliability.

Purpose of the Study:

• To propose a smart sensing architecture for the Revolution EVO CT scanner.
• To enable intelligent condition monitoring through data-driven analytics.
• To identify abnormal operating conditions in CT scanners for proactive maintenance.

Main Methods:

• Processed system logs from August 2024 to October 2025 into 10-min intervals.
• Created a structured dataset with 76 features from operational events, scanning parameters, and temporal dynamics.
• Trained Support Vector Machine (SVM) and Artificial Neural Network (ANN) supervised learning models to detect anomalies.

Main Results:

• Both SVM and ANN models achieved high classification accuracy (0.973).
• SVM demonstrated balanced performance with precision, recall, and F1-score of 0.973.
• ANN showed superior ranking and anomaly sensitivity with AUROC of 0.993 and AUPRC of 0.976.

Conclusions:

• The proposed sensor-driven machine learning framework effectively detects system anomalies in CT scanners.
• This approach has the potential to optimize maintenance planning in clinical environments.
• Intelligent condition monitoring enhances the reliability and uptime of advanced medical imaging equipment.