초록 ▼

A method for the analysis of a component of a medical sample with the aid of an autoanalyzer, in which the sample is reacted with a reagent system and a physically measurable quantity X resulting from the reaction of the sample with the reagent system is measured in order to determine at least one m

A method for the analysis of a component of a medical sample with the aid of an autoanalyzer, in which the sample is reacted with a reagent system and a physically measurable quantity X resulting from the reaction of the sample with the reagent system is measured in order to determine at least one measured value R for a specified sample and in which this at least one measured value R is further processed with the aid of a processing unit in a processing stage in order to determine an analytical result A. In the processing stage of the invention, use is made of the results of a neural net training, in which, for a number of standard samples for which the analytical result A is known, at least one measured value R, or a measurement result derived from several measured values Ri, is applied to the input of a neural net.

대표청구항 ▼

A method for the kinetic analysis of a component of a medical sample by means of an autoanalyzer to determine the concentration C of the component or a concentration-related parameter for the medical sample, comprising the steps of: a. reacting the sample with a reagent system, the step of reacting

A method for the kinetic analysis of a component of a medical sample by means of an autoanalyzer to determine the concentration C of the component or a concentration-related parameter for the medical sample, comprising the steps of: a. reacting the sample with a reagent system, the step of reacting the sample with the reagent system resulting in a time-dependent alteration in a physically measurable quantity X; b. measuring the physically measurable quantity X at various measurement times (ti=t1, t2, . . . , tn) and determining therefrom a plurality of time-dependent input values for a specified sample, said plurality of time-dependent input values being selected from the group consisting of: (1) several measured values Ri(ti) of the same physically measurable quantity X for a specified sample, (2) several measurement results derived from several measured values Ri(ti) for a specified sample, and (3) combinations thereof, and c. processing said plurality of time-dependent input values as defined in step b for the specified sample by means of a processing unit having a trained neural net and determining therefrom an analytical result A, wherein, for a specified sample, the analytical result A is the concentration C of a component or a concentration-related parameter for the sample; wherein the processing step further includes the step of using the trained neural net resulting from a neural net training, the neural net training including, for each of a large number of standard samples for which the analytical result A is known, applying to an input layer of a neural net a plurality of time-dependent input values as defined in step b; wherein the neural net comprises a substantially parallel data processing system including a plurality of neuronal processing elements arranged in a distributed topology which includes at least: (i) a plurality of layers, each layer having at least one neuronal processing element, the at least one neuronal processing element having an input and an output, one of the plurality of layers being an input layer comprising at least one neuronal processing element, said input layer having applied thereto a plurality of input signals corresponding to said plurality of time-dependent input values as defined in step b, another one of the plurality of layers being an output layer, comprising at least one neuronal processing element, said output layer providing an output signal corresponding to the analytical result A; and (ii) a plurality of neuronal interconnects, each neuronal interconnect connecting an output of a neuronal processing element in one layer to an input of a neuronal processing element in another layer, and having a weight which is modifiable during said neural net training.