An Efficient Method for Detection of Outliers in Tracer Curves Derived from Dynamic Contrast-Enhanced Imaging

الملخص EN

Presence of outliers in tracer concentration-time curves derived from dynamic contrast-enhanced imaging can adversely affect the analysis of the tracer curves by model-fitting.

A computationally efficient method for detecting outliers in tracer concentration-time curves is presented in this study.

The proposed method is based on a piecewise linear model and implemented using a robust clustering algorithm.

The method is noniterative and all the parameters are automatically estimated.

To compare the proposed method with existing Gaussian model based and robust regression-based methods, simulation studies were performed by simulating tracer concentration-time curves using the generalized Tofts model and kinetic parameters derived from different tissue types.

Results show that the proposed method and the robust regression-based method achieve better detection performance than the Gaussian model based method.

Compared with the robust regression-based method, the proposed method can achieve similar detection performance with much faster computation speed.