The purpose of research is to develop and test a technique for forming informative features using descriptors for neural networks designed to assess medical risks based on the analysis of transient processes in biomaterial in a living organism (in vivo).Methods. Studies suggest the use of test electrical effects on areas of the body with unusual conductivity to obtain the amplitude-phase-frequency characteristic of the impe-dance of the biomaterial on which the specified effect was performed. The coordinates of the Cole graph of this biomaterial were used as key para-meters. To form the Cole graph, the Carson transform was used, based on transient data obtained using a four-terminal, where the main element is the impedance of the studied biomaterial. The input signals for the four-terminal were a sequence of sinusoidal pulses.Results. Based on the E20-10 data collection system manufactured by L-Card CJSC, a software and hardware complex has been developed for digitizing transient processes in four-terminal circuits, the element of which is the impedance of biomaterial in anatomical areas with abnormal electrical conductivity. Software in the Delphi programming language was developed to generate test signals and record biomaterial responses to these exposures. A theoretical model was also proposed explaining the conversion of the samples of the transition characteristic of the four-terminal with the impedance of the biomaterial to the Cole graph of this biomaterial.Conclusion. The study confirms that the use of a linear biomaterial impedance model contributes to the formation of descriptors based on the amplitude-phase-frequency characteristic, taking into account its dissipative properties. Building a Cole graph taking into account these dissipative characteristics allows us to develop classifiers of medical risks of socially significant diseases.
Read full abstract