Clinically relevant cardiovascular parameters, such as pulmonary blood volume (PBV) and ejection fraction (EF), can be assessed through indicator dilution techniques. Among these techniques, which are typically invasive due to the need for central catheterization, contrast ultrasonography provides a new emerging minimally invasive option. PBV and EF are then measured by a dilution system identification algorithm after detection of multiple dilution curves by an ultrasound scanner. In this paper, dilution systems are represented by parametric models. Since the measured indicator dilution curves (IDCs) are corrupted by measurement artifacts and outliers, the use of conventional least square error (LSE) estimator for estimating system parameters is not optimal. Different estimators are therefore proposed for estimating the system parameters. Comparison of these estimators with the LSE estimator in assessing EF and PBV is performed on simulated, in vitro and patient data. The results show that the proposed total least absolute deviation estimator (TLAD) outperforms other estimators. The measured IDCs are highly corrupted by noise, which affect the estimation of EF and PBV. Therefore, a two stage denoising method capable of removing outliers is also proposed for removing noise in IDCs.
Bharath, H. N., Prabhu, K. M. M., Korsten, H. H. M., & Mischi, M. (2012). System modeling and identification in indicator dilution method for assessment of ejection fraction and pulmonary blood volume. Biomedical Signal Processing and Control, 7(6), 640-648. https://doi.org/10.1016/j.bspc.2012.03.006