There are numerous challenges along the development path which must be resolved for each new clinical situation and hardware configuration. For instance, the scalp and skull temperatures will change as well. However, we're developing a multi-frequency radiometer that can distinguish temperatures at different depths and thus overcome such hurdles. In any case, we have demonstrated that a precision and “accuracy” of 0.4 degC relative to the calibration point is possible in vivo when the radiometry system is carefully setup and calibrated for the specific clinical measurement configuration.
I and Paul Stuffer discussed your question further, and here is a more complete answer in case you’re interested:
Remember that MW radiometry is best for relative temperature readings and is dependent on calibration to one or more known temperature references.
The projected accuracy and resolution of microwave radiometry in vivo is highly dependent on specifics of the clinical application as well as radiometer hardware and software configurations.
For instance, accuracy of one large volume weighted average measurement of deep tissue temperature can be very precise and accurate, but that “accuracy” is essentially impossible to prove experimentally since any other measurement approach would measure a different volume. Certainly no single or multipoint implanted sensor can come close to characterizing the same volume-average tissue measurement, because the radiometric measurement is weighted by the distribution of receive antenna sensitivity defined by the radiation (=receive) pattern of the antenna integrated over the measurement frequency band and further weighted by the non-uniform receiver gain of the MW amplification chain (across radiometer frequency band) AND specific impedance coupling of antenna to tissue volume. Even true 3D MR thermal imaging would read a different volume average temperature from radiometry, since MRTI will weight the relative contributions of tissue temperatures within the ROI differently from the wideband radiometer receive antenna. The MRTI and MW radiometer volume average measurements might read the same if the volume being sensed was entirely uniformly equal temperature, but this is a rather non-interesting clinical case (i.e. dead patient). In that case a simple thermocouple inside the uniform temperature volume also correlates accurately with either MR or MW Rad temperatures. That is how we do laboratory phantom calibrations and tests of MW Rad temperature “accuracy”.
Recall that the radiometer electronics gives one received power measurement for one broadband radiometric reading. To produce an “accurate” radiometric equivalent tissue temperature (i.e. predicted tissue temperature), we perform extensive EM simulations of antenna radiation patterns into tissue at various frequencies across the radiation receive band of the radiometer electronics, and blend them together into an approximate average weighting function for that broadband antenna sensor. So the mathematics of the temperature conversion algorithm calculates an equivalent average temperature of a volume of tissue weighted by the distribution of received energy over the band from the antenna. If you have accurate a priori information about actual tissue temperature (preferably from a similar volume as you are trying to characterize) then you can “calibrate” your received power reading to translate to that specific volume average temperature. Then the radiometer can read this temperature and even changes in that temperature quite accurately (better than 0.4 degC) - as long as the following factors do not change:
i) impedance coupling of antenna to that tissue volume (so it is essential to maintain a stable connection of antenna, or somehow correct for any changes - though any correction would likely put a hit on accuracy. Best to keep antenna connection stable),
ii) outside EMI pickup (so shield EMI from rad sensor and average over longer time periods (10 s or preferably 30 s or more) to average out noise spikes)
iii) amplifier gain (this is systematic and can be calibrated out – we have already done this in the 0.4 degC accuracy demonstration over multiple hour experiment)
iv) temperature of the radiometer electronics and antenna itself (this can be calibrated out by thermistor temperature sensors on antenna and electronics box, and we have done this also as it is critical for long term stability)
Certainly there are foreseeable additional hurdles but we believe that when used carefully in appropriately controlled clinical applications, we can deliver “accuracy” of our calculated weighted average tissue volume temperature measurement on the order of 0.4 degC – even for long times of many hours. We will be moving soon to multiple frequency band, multiple antenna, and correlation mode radiometer systems that provide additional readings of power (temperature) from different and overlapping sense volumes. These multiple readings, from overlapping tissue regions, can be used to distinguish: a) temperature-depth profiles, b) calibrate and correct spurious temperature readings, c) improve overall precision, accuracy and spatial resolution of radiometric temperature predictions.
Sorry for following up on this so late. Thanks for your extensive answer.
Seeing your explanation, I certainly think that an accuracy of 0.4degC would be sufficient. However, the dependence on the clinical setting, which is inherently unstable, might be a threath. Placing antennas at reproducibly is very difficult. Still, also for our simulation guided approach we found ways to obtain a good reproducibility (see e.g. our work on photogrametry). In addition, would an alternative be combine invasive temperature sensors with radiometry to get the complete picture?