Abstract Increased use of portable and intelligent diagnostic devices has spurred the use of ultrasound imaging, cloud computing, and machine learning (ML). This paper outlines the design and implementation of an intelligent diagnosis framework for networked portable ultrasound systems based on cloud infrastructure. The system is structured around a modular pipeline that mimics cloud transmission effects, extracts waveform features, and uses machine learning models for anomaly detection. Functional disturbances such as signal delay, packet loss, and overheating were simulated, and signal-based characteristics were derived to detect anomalies. A combination of autoencoder, isolation forest, and one-class support vector machine (SVM) models was shown to achieve a detection rate of up to 94% for four anomaly classes. Simulations for adaptive routing also illustrated a power efficiency gain of 18%. The results verify the practicability of real-time monitoring and ML-aided diagnostics being incorporated in cloud-linked ultrasound machines.
Ahmed H. Ahmed Albegli (Sun,) studied this question.