Graduation Year


Document Type




Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Electrical Engineering

Major Professor

Stephen E. Saddow, Ph.D.

Committee Member

Gokhan Mumcu, Ph.D.

Committee Member

Robert D. Frisina, Ph.D.

Committee Member

Sylvia W. Thomas, Ph.D.

Committee Member

Lennox Hoyte, Ph.D.


Non-invasive sensing, implantable sensing, blood sugar level, blood permittivity, antenna patch


The purpose of this research was to design a blood glucose sensing system based on the induced shift in the resonant frequency of an antenna patch operating in the ISM band (5.725 – 5.875 GHz). The underlying concept is the fact that when a person has variations in their blood glucose levels, the permittivity of their blood varies accordingly. This research analyzed the feasibility of using an antenna patch as a blood glucose sensing device in three configurations: 1) as an implantable active sensor, 2) as an implantable passive antenna sensor, and 3) as a non-invasive sensor. In the first arrangement, the antenna is to be implanted inside the body as an active antenna, requiring that its power supply and internal circuitry to be implanted. In the second arrangement, the antenna is also implanted, but would not require a power supply or internal circuity since it would be passive. For the third arrangement, the non-invasive sensing approach, the antenna is placed facing the upper arm while mounted outside the body. In order to evaluate the best approach all the three approaches were simulated using the electromagnetic field tool simulator ANSYS EM15.0 HFSSTM, along with a human tissue model. The tissue model included physiological and electrical characteristics of the human abdomen for simulating the active and passive approaches, and the upper arm for the non-invasive approach. The electromagnetic boundaries were set with perfectly matched layers to eliminate any reflections which would cause a non-physical resonance in the results. Simulation of the active sensing configuration resulted in a resonant frequency shift from 5.76 to 5.78GHz (i.e., a 20 MHz shift) for a simulated blood permittivity variation of 62.0 to 63.6. This corresponds, theoretically, to an approximate glucose shift of 500 mg/dL. The passive configuration simulations did not yield conclusive variations in resonant frequency and this approach was abandoned early on in this research. Thirdly, the non-invasive approach resulted in a simulated shift of resonant frequency from 5.797 to 5.807 (i.e., a 10MHz shift) for simulated blood permittivity variation of 51.397 to 52.642 (an approximate variation of 2000 mg/dL in glucose). In the literature planar, continuous blood-rich layers are used to simulate RF sensing of glucose, which is not applicable when measuring glucose in actual human veins, which are tubular in geometry and of finite extent. Therefore the model employed assumed a 1.8 mm diameter blood vessel, buried under a fatty layer that was capped with skin. The above results, both simulated and verified experimentally, used this more realistic model which is further proof that a practical non-invasive blood glucose measurement system should be possible.

The non-invasive approach was tested experimentally by using oil in gel phantoms to mimic the electrical properties of skin, fat, blood and muscle. A fat phantom was placed over a muscle phantom, with a strip of blood phantom within and a skin phantom was placed on top. The blood phantom had a 2000mg/dL variation of D-glucose in the phantom mixture which decreased the relative permittivity from 52.635 to 51.482 and resulted in a shift of resonant frequency from 5.855 to 5.842 (i.e., a 13MHz shift). This is consistent with the non-invasive simulated results thus validating our model of the non-invasive sensing approach. While this variation in blood glucose is non-physical (typical human glucose range can range in the extremes from 30 to 400 mg/dL, where healthy glucose levels vary from 70mg/dL to 180mg/dL) it was necessary to provide a high confidence fit between the simulated and experimental data. This is because the level of precision with which the physical phantoms could be fabricated with was insufficient to match the highly precise simulated data.

Analysis on the effect of lateral displacement of the antenna from the blood vessel, its elevation above the skin and variations caused by different skin thickness, and blood vessel depth were evaluated. A calibration technique to correct physical misalignment by the user is proposed in which two additional antennas, located diagonally with respect to the sensing antenna, serve as reference point for placement over the upper arm in line of sight with the blood vessel.

Once the non-invasive sensor approach was shown to be viable for continuous glucose monitoring, a sensor platform was designed whereby an RF generator was used to drive the antenna with a frequency sweep between 5.725 to 5.875GHz. A fraction of its output power was coupled to both the antenna and the system analysis circuitry through a directional coupler. The transmitted and received power were then processed with demodulating logarithmic amplifiers which convert the RF signal to a corresponding voltage for downstream processing. Both inputs were then fed into a microcontroller and the measured shift in resonant frequency, fO, converted to glucose concentration which was displayed on glucose meter display.