| Popis: |
Measurement of renal function in awake rats or mice can be accomplished by an intelligent plaster device that fits on the back of animals. The device performs a percutaneous measurement of the kinetics of a labeled fluorescent dye exclusively eliminated by the kidney. During the measurement, relative motion between plaster and skin leads to a variation of the illumination conditions, which emerge as artifacts in the data. In this paper, a novel strategy to detect and eliminate artifacts is suggested. The method combines cluster analysis and nonlinear regression with a priori knowledge about signal morphology to correct data. The performance of the proposed method is demonstrated on real and simulated data. Simulations were performed on data with two artifact amplitude ranges: (1) shifts in the recorded data with amplitude exceeding 3% of the signal amplitude for a combined duration of 10% of the total measurement time and (2) shifts greater than 3% for approximately 30% of the total measurement time. Prior to artifact removal, the MAE was calculated to be 10.3% and 21.9%, respectively. Following artifact removal using the proposed method, results showed that, when determining the half-life, the mean absolute error (MAE) was 0.88% for range type 1 and 10.4% for the more substantial range of the type 2 artifacts. When examining real data, the mean difference (bias) while determining the half-life was 7.5%. Results show that novel technique outperforms a number of state-of-the-art techniques when removing artifacts from the signal recorded while an animal is allowed to move freely. In this case, the signal acquires shifts and random changes with large amplitudes, which make it impossible to use standard methods. |
Full Text from ScienceDirect