Integrated heterogeneous multimodal 5G-enabled, radio-frequency based, non-invasive, wearable head imaging system for telemedicine applications
This thesis presents a novel, cost-effective, and non-invasive approach to next-generation head imaging systems for telemedicine applications. Current imaging technologies—such as CT scans, MRI, EEG, and X-rays—face significant challenges including high cost, invasive procedures, bulky equipment, radiation exposure, and limited real-time monitoring capabilities. In response, this work introduces innovative wearable solutions that integrate advanced antenna designs, flexible materials, and artificial intelligence to enhance diagnostic accuracy and patient safety. The first contribution details the development of a graphene-polyamide-based flexible wearable 5G communication antenna. Leveraging graphene’s high electrical conductivity and mechanical strength, combined with a fractal design on a flexible substrate, this antenna achieves an ultra-wide bandwidth and robust real-time data transmission necessary for continuous head monitoring. The second contribution presents an AI and Machine Learning-enhanced Ultrawide-band Cu-FR4-Cu fractal microwave antenna designed for telemedicine applications. This design offers an ultra-wide impedance bandwidth and high gain performance across multiple frequency bands—from Sub-6 GHz to mmWave 5G bands—while integrated machine learning algorithms, Adaptive beamforming, Interference mitigation, Self-learning impedance matching, are adopted for optimizing antenna performance. The third contribution presents a wearable RF-based system using integrated circular-patch and M-shape UWB antennas for brain abscess detection. Custom-designed sensors were evaluated through CST simulations and validated on realistic 3D-printed head phantoms. The antennas exhibited wideband matching, high radiation efficiency, and safe SAR levels under 1W excitation. Return-loss and phase shifts were proportional to inclusion size, enabling accurate sizing of brain abscesses. A dual-sensor approach enhanced detection sensitivity and robustness to misalignment. The system supports 5G-enabled telemedicine and machine learning for real-time, remote diagnostics. The final contribution presents a first of its kind wearable ultra-wideband (UWB) RF sensing system for real-time, non-invasive brain abscess detection and staging. An M-shaped antenna with vector network analyzer captured S11 return-loss measurements across multiple gigahertz frequencies from a comprehensive dataset of 6,650 samples, strategically split into training (4,657), validation (995), and test (998) sets with stratified sampling to maintain class distribution. A sophisticated machine learning pipeline implemented LSTM neural networks (77.56% multiclass accuracy) and XGBoost (75.45% multiclass accuracy) for comprehensive four-class brain abscess staging, with optimized deployment targeting Raspberry Pi 5 edge computing platforms. The system demonstrates exceptional clinical performance with LSTM achieving superior multiclass classification accuracy of 77.56% and maintaining robust real-world deployment performance of 81.0% accuracy during edge implementation. The comprehensive validation framework encompasses six critical domains: real-time processing achieving sub-millisecond inference (0.795ms average, representing 1,257× speed improvement over 100ms clinical threshold), clinical-grade accuracy surpassing 75% requirements, dataset robustness with over 14,000% augmentation, 100% system reliability validation, full Raspberry Pi 5 hardware compatibility, and complete 4-class pathology classification capability with integrated GPIO feedback systems. This breakthrough RF-based system enables continuous brain abscess monitoring at the point of care through a 21-dimensional feature space encompassing raw RF measurements, extracted signal characteristics, and clinical parameters. The system achieves productionready clinical translation status with established diagnostic precision, proven system reliability, and preliminary regulatory preparation suitable for pilot testing initiatives in resource-limited telemedicine environments. Collectively, the research offers a transformative approach to wearable head imaging by addressing the limitations of conventional systems through comprehensive validation and edge deployment optimization.