Suppressing the system error in the measurement model of the prediction-based object recognition algorithm: ovarian follicle detection case

A heuristic procedure for suppressing system error in the measurement model of a prediction algorithm is presented. This error is suppressed by modifying the measurements. The procedure consists of two steps. Firstly the decision whether a measurement should be modified or nor is taken, and secondly, the measurement is actually modified. Mathematical mechanisms are developed for an integration of the modified measurement model into the prediction algorithm. The new algorithm was tested on sequences of ovarian ultrasound images with follicles. The follicles are recognised about 3% more accurately when compared to the results obtained using the basic prediction algorithm.