Contactless Assessment of Physiological Parameters with a 61 GHz CW Medical Radar System
summary
Background
Implementation of sensor technology and analysis supported by artificial intelligence (AI) are important drivers of medical research in the healthcare sector. Studies demonstrate the reliability, robustness, and efficiency of existing systems and illustrate synergistic benefits of clinical expertise, sensor technology and AI.
Technical Description and Use
This study uses a high-frequency 61 GHz continuous-wave (CW) radar system for contactless detection of cardiorespiratory parameters. The newly developed radar system offers unique advantages including higher level of integration, improved sensitivity, and smaller size compared to systems operating at lower frequencies. Respiration and heart beat rates are key parameters in the monitoring of physiologic and pathologic settings. Thus, the proposed approach may improve quality of medical care in hospitals. In addition, and in contrast to conventional monitoring based on electrodes and registration-lines, patient comfort is increased and applicability for the clinical staff is improved. The proposed radar system incorporates three stacked boards including frontend, baseband, and microcontroller components. The frontend board has a 30.5 GHz voltage-controlled oscillator (VCO) chip integrated with a divide-by-16 chain that is stabilized in a phase-locked loop (PLL) and a 61 GHz transceiver (TRX) chip equipped with an in-phase/quadrature (I/Q) receiver. Both VCO and TRX chips have been designed and fabricated in an advanced silicon-germanium (SiGe) BiCMOS technology. Intermediate frequency (IF) outputs are dc-coupled to enable CW measurements of the low frequency spectrum of physiologic parameters. A series-fed 2×8 patch antenna array on the frontend board is equipped on a 254-μm-thick Megtron6 substrate with a gain of 12.7 dB and an RF bandwidth of 5.1 GHz (range 57.5 to 62.6 GHz). The array has a beamwidth of 15° E-plane and 50° H-plane, and a simulated radiation efficiency of 72%. A wideband planar RF balun is developed at 61 GHz to interface the single-ended antenna array input with the differential transmitter and receiver outputs of the TRX chip. The integration of 61 GHz TRX with the balun is achieved by 25-μm-diameter gold bondwires.
Post-Processing
The raw IF data output is digitized by a baseband board and transferred to a computer by an STM32 microcontroller evaluation board. Raw radar I/Q signals are band-pass filtered and transformed in pulse-wave and heart sound signals. Based on earlier datasets, a deep learning model was pre-trained and applied for the analysis of the heart beat and the beat-to-beat-interval (IBI) out of the filtered signals. Initial results comparing the accuracy of the radar system to a conventional electrocardiogram (ECG) recorder show a high F1 score up to 0.99 for heart beat detection. In addition, effects of patient movement, e.g., random- and speaking associated movements, and body-position are part of the presented study.