Tracer Experiment as a Criterion for the Reliability of Predictive Modeling of Statistical Data in Oncology

Abstract

Prognosis is the process of calculating a prognosis or development a predictive model that predicts the future values of a time series, based on its values in the past, allowing to detect (predict) patterns in large sets of statistical data, using optimal interpolation methods. Modeling of an accurate prognosis in oncology, the reliability of which is confirmed by a tracer experiment, allows establishing an accurate diagnosis in order to start the necessary treatment, predict the long-term results of this treatment and evaluate the effectiveness of this treatment. In the article, along with the description of two previously developed principles of the interpolation predicting model in oncology: the determination of the minimum of difference (error) between the numerical values of the studied symptom, data from the patient's history, and calculated, using various methods of polynomial interpolation; the choice as a given function of the values of the statistical frequencies of the observed patients in the given time intervals of the studied symptom, which made it possible to develop an algorithm, and then a system computer control of the medical process "Medical Commander" for optimal interpolation prediction of the studied symptoms in oncology, a criterion for the reliability of the developed system is presented, which confirmed the universality of its work with absolutely independent statistical data giving the same result. The presented criterion for the reliability of the developed system made it possible not only to identify a surge (“hump”) of a group of patients with metastases of a certain localization of the neoplasm after surgery in a clinical conditions, but also to determine that this surge lies in the same time interval for a group of patients, who are completely independent of each other, united only by the localization of the neoplasm.