CN115307945A - Rotary blood pump device integrating multiple sensors and running state monitoring method - Google Patents

Abstract

The invention discloses a rotary blood pump device integrating multiple sensors and an operation state monitoring method. Comprises a rotary blood pump provided with a flow sensor, an acceleration sensor and a current sensor. Signals of the three sensors are combined in parallel after being preprocessed, feature extraction is carried out through a one-dimensional convolution network, and finally a Softmax classifier is adopted to monitor the running state of the blood pump. The invention can monitor the operation of the rotary blood pump after the rotary blood pump is implanted into a human body, and monitor four states of normal operation of the blood pump, blood backflow of the blood pump, ventricular suction collapse and thrombus blockage in the pump so as to ensure the safe operation of the rotary blood pump.

Description

Rotary blood pump device integrating multiple sensors and running state monitoring method

Technical Field

The invention relates to a rotary blood pump device and an operation state monitoring method in the field of biomedical engineering, in particular to a rotary blood pump device integrating multiple sensors and an operation state monitoring method.

Background

The rotary blood pump is a high-end medical equipment, is a core component of VAD (VAD) of the conventional artificial heart and ventricular assist device, and is used for supporting the blood circulation of a human body under the condition that the heart cannot work normally. Through the development of many years, the rotary blood pump enters the clinical use stage, is the latest generation, the most mainstream and the most widely applied ventricular assist device in the world at present.

Stable operation of rotary blood pumps is crucial for clinical use. If the running state of the blood pump changes, the rotary blood pump runs in an adverse working condition, and a great risk is caused to the life of a patient. Due to changes in blood pump speed, heart beating capacity, and human motion state, etc., it may result in some cases in which the rotary blood pump operates in an undesirable state: the rotating speed is too high, so that the ventricles are sucked empty to collapse; blood in the blood pump flows backwards due to too low rotating speed; the pump is clogged by foreign matter such as thrombus, which causes blood flow. Therefore, the method has important function and significance for timely monitoring and classifying the normal running state and the poor running state of the rotary blood pump.

At present, related patents on monitoring the running state of the rotary blood pump are few, only a few methods for simply judging single states of blood pump backflow or ventricular aspiration and the like are performed by extracting single signals of blood pump flow and the like through artificial features, and the signals of flow and the like are measured by measuring instruments such as an external flowmeter and the like, so that the monitoring method is not suitable for the implantation requirements of the rotary blood pump.

Disclosure of Invention

In order to solve the problems existing in the background technology and solve the defects, the invention provides a rotary blood pump device integrating various sensors and an operation state monitoring method.

The invention adopts the following technical scheme:

1. a device for monitoring the running state of a rotary blood pump integrating a plurality of sensors comprises:

the device adopts a rotary blood pump provided with a flow sensor, an acceleration sensor and a current sensor; the flow sensor is arranged at an outlet pipe of the rotary blood pump, the acceleration sensor is arranged at the bottom of the rotary blood pump, and the current sensor is arranged in a controller of the rotary blood pump; the running state of the rotary blood pump is judged according to signals obtained by the flow sensor, the acceleration sensor and the current sensor.

The flow sensor is a non-contact ultrasonic flow sensor; the acceleration sensor is a piezoelectric acceleration sensor; the current sensor is integrated in the controller of the rotary blood pump.

And the flow sensor, the acceleration sensor and the current sensor synchronously acquire and store signals on the same time axis.

2. A method for monitoring the running state of a rotary blood pump integrating multiple sensors comprises the following steps:

the monitoring process of the monitoring method specifically comprises the following steps:

1.1 Flow signals, axial acceleration signals, radial acceleration signals and current signals of the rotary blood pump measured in clinical experiments under four operating states are collected, and the collected flow signals, axial acceleration signals, radial acceleration signals and current signals under the four operating states are used as training sets;

1.2 Respectively carrying out filtering pretreatment on all flow signals, axial acceleration signals, radial acceleration signals and current signals in the training set;

1.3 Respectively optimizing the flow signal, the axial acceleration signal, the radial acceleration signal and the current signal which are subjected to filtering pretreatment in different running states to obtain an N x 4 signal combined data matrix in different running states;

1.4 All the N x 4 signal combination data matrixes are sequentially input into a one-dimensional convolution neural network for feature extraction to respectively obtain one-dimensional vectors of different running states;

1.5 Inputting each one-dimensional vector into a full-connection layer to obtain output probabilities of a state one, a state two, a state three and a state four of the one-dimensional vector; the state with the maximum output probability corresponds to the operating state corresponding to the input signal combination data matrix of N × 4, and the determination of the operating states corresponding to the state one, the state two, the state three, and the state four is completed.

1.6 The method comprises the steps of) collecting a flow signal, an axial acceleration signal, a radial acceleration signal and a current signal of the rotary blood pump to be measured in the same operation state, preprocessing the flow signal, the axial acceleration signal, the radial acceleration signal and the current signal of the rotary blood pump to be measured in the same operation state, inputting the preprocessed flow signal, the preprocessed axial acceleration signal, the preprocessed radial acceleration signal and the preprocessed current signal of the rotary blood pump to be measured into a one-dimensional convolutional neural network to perform characteristic extraction to obtain a one-dimensional vector of the rotary blood pump to be measured in the operation state, inputting the one-dimensional vector into a full-connection layer to obtain four output probabilities, wherein the operation state corresponding to the maximum output probability is the operation state of the rotary blood pump to be measured, and monitoring the operation state of the rotary blood pump to be measured.

The step 1.3) is specifically as follows: respectively intercepting discrete data units in the same section of time interval from the flow signal, the axial acceleration signal, the radial acceleration signal and the current signal which are subjected to filtering preprocessing in the same operation state, taking the number N of discrete data points of the discrete data unit with the most discrete data points in four sections of discrete data units as a reference N, amplifying the discrete data points of the other three sections of discrete data units by an interpolation method until the number of the discrete data points is equal to the reference N, and finally connecting the amplified four sections of discrete data units together in parallel to form a signal combination data matrix of N4.

The step 1.4) is specifically as follows:

1.4.1 Inputting the signal combination data matrix of N by 4 into the first convolution kernel and the second convolution kernel in sequence to obtain a data matrix of (N-9) by 100 and a data matrix of (N-99) by 100 respectively;

1.4.2 Simultaneously performing maximum pooling on the (N-9) × 100 data matrix and the (N-99) × 100 data matrix to obtain a (N-9)/6 × 100 data matrix and a (N-99)/6 × 100 data matrix, respectively;

1.4.3 The (N-9)/6 x 100 data matrix and the (N-99)/6 x 100 data matrix are respectively input into a first convolution kernel and a second convolution kernel for convolution, and then the maximum pooling is carried out on the convolved (N-9)/6 x 100 data matrix and the convolved (N-99)/6 x 100 data matrix respectively, so as to obtain the (N-9)/6 x 100 data matrix after the maximum pooling and the (N-99)/6 x 100 data matrix after the maximum pooling is obtained;

1.4.4 D repeats steps 1.4.3) for the maximum pooled (N-9)/6 x 100 data matrix and the maximum pooled (N-99)/6 x 100 data matrix, finally obtaining a first 1 x 100 data matrix and a second 1 x 100 data matrix, respectively;

1.4.5 The first data matrix of 1 × 100 and the second data matrix of 1 × 100 are superposed and combined to obtain a one-dimensional vector having a structure of the data matrix of 1 × 100.

The four operation states comprise a normal working state, a blood backflow state of a blood pump, a ventricular suction collapse state and a thrombus blockage state in the pump.

The interpolation method is piecewise linear interpolation, bilinear interpolation or spline interpolation; the activation function of the one-dimensional convolution adopts a linear rectification function RELU, and the activation function of the full connection layer adopts a normalized exponential function Softmax.

The sizes of convolution kernels of the first convolution kernel and the second convolution kernel are respectively 10 × 4 and 100 × 4, the number of the convolution kernels is 100, and the step length is 1; the maximum pooled nuclei size was 6 with a step size of 6.

The number of repetitions D of step 1.4.3) is from 1 to 3.

The invention has the beneficial effects that:

the invention integrates various sensors in the rotary blood pump, improves the implantability of the sensors and can measure various relevant parameters of the rotary blood pump on line.

The method for monitoring the running state of the rotary blood pump by fusing the signals of the multiple sensors has better monitoring accuracy and reliability.

The invention can provide guarantee for the clinical safe use of the rotary blood pump, and use the one-dimensional convolution neural network to carry on the fusion and characteristic extraction of the multiple signals, has proposed the new one-dimensional convolution neural network structure; the accuracy and the reliability of monitoring the running state of the rotary blood pump are ensured.

The invention can monitor the operation of the rotary blood pump after the rotary blood pump is implanted into a human body, and detect four states of normal operation of the blood pump, blood backflow of the blood pump, ventricular suction collapse and thrombus blockage in the pump so as to ensure the safe operation of the rotary blood pump.

Drawings

FIG. 1 is a schematic view of an integrated multi-sensor rotary blood pump implantation;

FIG. 2 is a schematic top view of an integrated position of a flow and current sensor;

FIG. 3 is a side view of a schematic of the integrated position of the flow, acceleration and current sensors;

FIG. 4 is a bottom view of a schematic of the integrated position of the flow and acceleration sensors;

FIG. 5 is a flow chart diagram of a method of monitoring the operating state of a rotary blood pump;

fig. 6 is a convolution network structure diagram for classifying the running state of the rotary blood pump.

Shown in the figure: 1. a rotary blood pump; 2. a left ventricle; 3. the aorta; 1-1, a flow sensor; 1-2, an acceleration sensor; 1-3, current sensor.

Detailed Description

The invention is described in further detail below with reference to the figures and the embodiments.

As shown in fig. 2, 3 and 4, the device adopts a rotary blood pump 1 provided with a flow sensor 1-1, an acceleration sensor 1-2 and a current sensor 1-3; the flow sensor 1-1 is arranged at an outlet pipe of the rotary blood pump 1 and is used for measuring the outlet blood flow of the rotary blood pump, the acceleration sensor 1-2 is arranged at the bottom of the rotary blood pump 1 and is used for measuring the axial and radial vibration acceleration of the rotary blood pump, and the current sensor 1-3 is arranged in a controller of the rotary blood pump 1 and is used for measuring the current waveform of a direct current motor of the rotary blood pump; so that the rotary blood pump 1 can judge the running state of the rotary blood pump 1 according to the signals obtained by the flow sensor 1-1, the acceleration sensor 1-2 and the current sensor 1-3. The flow signal measured by the flow sensor 1-1 is subjected to high-frequency interference removal through a low-pass filter, and the cut-off frequency is set to be 5-10 times of the heart rate; the blood pump vibration acceleration signal measured by the acceleration sensor 1-2 passes through a band-pass filter, and the upper cut-off frequency and the lower cut-off frequency of the blood pump vibration acceleration signal are respectively set to be 10Hz and 500Hz; the current signal of the motor of the blood pump measured by the current sensor 1-3 passes through a low-pass filter, and the cut-off frequency is set to be 1-5 times of the low-pass filter of the flow signal.

The flow sensor 1-1 is a non-contact ultrasonic flow sensor; the acceleration sensor 1-2 is a piezoelectric acceleration sensor; the current sensors 1-3 are integrated in the controller of the rotary blood pump.

The flow sensor 1-1, the acceleration sensor 1-2 and the current sensor 1-3 synchronously acquire and store signals on the same time axis.

As shown in fig. 5, the monitoring process of the monitoring method specifically includes:

1.1 Collecting flow signals, axial acceleration signals, radial acceleration signals and current signals of the rotary blood pump 1 measured in clinical experiments under four operating states with known occurrence probability respectively, and taking the collected flow signals, axial acceleration signals, radial acceleration signals and current signals under the four operating states as a training set;

1.2 Respectively carrying out filtering pretreatment on all flow signals, axial acceleration signals, radial acceleration signals and current signals in the training set;

1.3 Respectively optimizing the flow signal, the axial acceleration signal, the radial acceleration signal and the current signal which are subjected to filtering pretreatment and under different operation states to obtain a signal combination data matrix of N x 4 under different operation states;

1.4 All the N x 4 signal combination data matrixes are sequentially input into a one-dimensional convolution neural network for feature extraction to respectively obtain one-dimensional vectors of different running states;

1.5 Inputting each one-dimensional vector into a full-connection layer to obtain output probabilities of a state one, a state two, a state three and a state four of the one-dimensional vector; the state with the maximum output probability corresponds to the operating state corresponding to the input signal combination data matrix of N × 4, that is, the flow signal, the axial acceleration signal, the radial acceleration signal and the current signal of the rotary blood pump 1 in the ventricular suction collapse state are input into the one-dimensional convolution neural network for processing, and finally the state with the maximum output probability after being output by the full connection layer corresponds to the ventricular suction collapse state. And finishing the determination of the running states corresponding to the state I, the state II, the state III and the state IV respectively. In this embodiment, the first state, the second state, the third state and the fourth state after the one-dimensional vector is input into the full connection layer correspond to a normal working state, a blood pump blood backflow state, a ventricular suction collapse state and a pump thrombus blockage state respectively.

According to the error between the maximum output probability corresponding to each of the four operation states and the occurrence probability of each of the four operation states, the parameters of the one-dimensional convolutional neural network are iteratively optimized through a back propagation algorithm of the one-dimensional convolutional neural network, the one-dimensional convolutional neural network is optimized, and the monitoring accuracy is improved.

1.6 The method comprises the steps of) collecting a flow signal, an axial acceleration signal, a radial acceleration signal and a current signal of the rotary blood pump to be measured in the same operation state, preprocessing the flow signal, the axial acceleration signal, the radial acceleration signal and the current signal of the rotary blood pump to be measured in the same operation state, inputting the preprocessed flow signal, the preprocessed axial acceleration signal, the preprocessed radial acceleration signal and the preprocessed current signal of the rotary blood pump to be measured into a one-dimensional convolutional neural network to perform characteristic extraction to obtain a one-dimensional vector of the rotary blood pump to be measured in the operation state, inputting the one-dimensional vector into a full-connection layer to obtain four output probabilities, wherein the operation state corresponding to the maximum output probability is the operation state of the rotary blood pump to be measured, and monitoring the operation state of the rotary blood pump to be measured.

The step 1.3) is specifically as follows: respectively intercepting discrete data units in the same section of time interval from the flow signal, the axial acceleration signal, the radial acceleration signal and the current signal which are subjected to filtering preprocessing in the same operation state, taking the number N of discrete data points of the discrete data unit with the most discrete data points in four sections of discrete data units as a reference N, amplifying the discrete data points of the other three sections of discrete data units by an interpolation method until the number of the discrete data points is equal to the reference N, and finally connecting the amplified four sections of discrete data units together in parallel to form a signal combination data matrix of N4.

The duration of the same time interval is 1-3 s, the time length of 1-3 s in the intercepted discrete data is selected according to the size of the cardiac cycle, and the time length enables the measuring signal to completely show the cycle characteristics of the measuring signal.

As shown in fig. 6, step 1.4) specifically includes:

1.4.1 Inputting the signal combination data matrix of N by 4 into the first convolution kernel and the second convolution kernel in sequence to obtain a data matrix of (N-9) by 100 and a data matrix of (N-99) by 100 respectively;

1.4.2 Simultaneously performing maximum pooling on the (N-9) × 100 data matrix and the (N-99) × 100 data matrix to obtain a (N-9)/6 × 100 data matrix and a (N-99)/6 × 100 data matrix, respectively;

1.4.3 The (N-9)/6 x 100 data matrix and the (N-99)/6 x 100 data matrix are respectively input into a first convolution kernel and a second convolution kernel for convolution, and then the maximum pooling is carried out on the convolved (N-9)/6 x 100 data matrix and the convolved (N-99)/6 x 100 data matrix respectively, so as to obtain the (N-9)/6 x 100 data matrix after the maximum pooling and the (N-99)/6 x 100 data matrix after the maximum pooling is obtained;

1.4.4 Repeating the data matrix of maximum pooled (N-9)/6 x 100 and the data matrix of maximum pooled (N-99)/6 x 100 for D times step 1.4.3) to finally obtain a first data matrix of 1 x 100 and a second data matrix of 1 x 100, respectively;

1.4.5 The first data matrix of 1 x 100 and the second data matrix of 1 x 100 are combined in a superposition mode, and neural units of some networks are temporarily and randomly discarded during training to reduce overfitting, so that one-dimensional vectors with the structure of the data matrix of 1 x 100 are obtained.

Preferably, the four operating states include a normal operating state, a blood pump blood backflow state, a ventricular suction collapse state and a pump thrombus blockage state.

Preferably, the interpolation method is piecewise linear interpolation, bilinear interpolation or spline interpolation; the activation function of the one-dimensional convolution adopts a linear rectification function RELU, and the activation function of the full connection layer adopts a normalized exponential function Softmax.

Preferably, the convolution kernel sizes of the first convolution kernel and the second convolution kernel are 10 × 4 and 100 × 4, respectively, the number of convolution kernels is 100, and the step size is 1; the maximum pooled nuclei size was 6 with a step size of 6.

Preferably, the repetition number D of step 1.4.3) is 1 to 3 times, which is determined according to the size of the reference N of the discrete data points.

The flow sensor 1-1, the acceleration sensor 1-2 and the current sensor 1-3 are modularized components, whether the flow sensor, the acceleration sensor 1-2 and the current sensor are integrally installed on the rotary blood pump 1 or not can be selected according to needs in actual use, and in addition, a pressure sensor module can be additionally installed at an inlet of the rotary blood pump 1, so that the collectable signals of the rotary blood pump 1 are further increased. The flow sensor 1-1, the acceleration sensor 1-2 and the current sensor 1-3 have different sampling frequencies, the flow sensor 1-1 and the current sensor 1-3 can be set to have a sampling frequency of about several hundred hertz at most, and the acceleration sensor 1-2 has a maximum sampling frequency of several thousand hertz, so that the number of discrete points sampled in the same time period is different. In subsequent processing, the same time signal segments of the three sensors are amplified to the same dimension by an interpolation method.

The invention can train the proposed one-dimensional convolution neural network by adopting flow, acceleration and current data of the rotary blood pump 1 measured in clinic or animal experiments under different known running states, the trained network can be deployed by a computer or an embedded system, four working states of normal work, blood backflow of the blood pump, ventricular suction collapse and thrombus blockage in the pump of the rotary blood pump 1 are monitored and classified in practical clinical use, and an alarm is given when an abnormal working state of the blood pump is found, so that the running safety of the blood pump is ensured.

Claims (10)

1. The utility model provides an integrated multiple sensor's rotary blood pump running state monitoring devices which characterized in that: the device adopts a rotary blood pump (1) provided with a flow sensor (1-1), an acceleration sensor (1-2) and a current sensor (1-3); the flow sensor (1-1) is arranged at an outlet pipe of the rotary blood pump (1), the acceleration sensor (1-2) is arranged at the bottom of the rotary blood pump (1), and the current sensor (1-3) is arranged in a controller of the rotary blood pump (1); the running state of the rotary blood pump (1) is judged according to signals obtained by the flow sensor (1-1), the acceleration sensor (1-2) and the current sensor (1-3).

2. The device for monitoring the operating state of a rotary blood pump integrating a plurality of sensors according to claim 1, wherein: the flow sensor (1-1) is a non-contact ultrasonic flow sensor; the acceleration sensor (1-2) is a piezoelectric acceleration sensor; the current sensor (1-3) is integrated in the controller of the rotary blood pump.

3. The device for monitoring the operating state of a rotary blood pump integrating a plurality of sensors according to claim 1, wherein: the flow sensor (1-1), the acceleration sensor (1-2) and the current sensor (1-3) synchronously acquire and store signals on the same time axis.

4. The method for monitoring the running state of the rotary blood pump integrated with multiple sensors, which is applied to the device of any one of claims 1 to 3, is characterized in that the monitoring process of the monitoring method is as follows:

1.1 Collecting flow signals, axial acceleration signals, radial acceleration signals and current signals of the rotary blood pump (1) measured in clinical experiments under four operating states respectively, and taking the collected flow signals, axial acceleration signals, radial acceleration signals and current signals under the four operating states as a training set;

1.2 Respectively carrying out filtering pretreatment on all flow signals, axial acceleration signals, radial acceleration signals and current signals in the training set;

1.3 Respectively optimizing the flow signal, the axial acceleration signal, the radial acceleration signal and the current signal which are subjected to filtering pretreatment in different running states to obtain an N x 4 signal combined data matrix in different running states;

1.4 All the N x 4 signal combination data matrixes are sequentially input into a one-dimensional convolution neural network for feature extraction to respectively obtain one-dimensional vectors of different running states;

1.5 Inputting each one-dimensional vector into a full-connection layer to obtain output probabilities of a state one, a state two, a state three and a state four of the one-dimensional vector; the state with the maximum output probability corresponds to the operation state corresponding to the input signal combination data matrix of N x 4, and the determination of the operation states corresponding to the state one, the state two, the state three and the state four is completed;

1.6 The method comprises the steps of) collecting a flow signal, an axial acceleration signal, a radial acceleration signal and a current signal of the rotary blood pump to be measured in the same operation state, preprocessing the flow signal, the axial acceleration signal, the radial acceleration signal and the current signal of the rotary blood pump to be measured in the same operation state, inputting the preprocessed flow signal, the preprocessed axial acceleration signal, the preprocessed radial acceleration signal and the preprocessed current signal of the rotary blood pump to be measured into a one-dimensional convolutional neural network to perform characteristic extraction to obtain a one-dimensional vector of the rotary blood pump to be measured in the operation state, inputting the one-dimensional vector into a full-connection layer to obtain four output probabilities, wherein the operation state corresponding to the maximum output probability is the operation state of the rotary blood pump to be measured, and monitoring the operation state of the rotary blood pump to be measured.

5. The method for monitoring the operating state of a rotary blood pump integrating multiple sensors according to claim 4, wherein the method comprises the following steps: the step 1.3) is specifically as follows: respectively intercepting discrete data units in the same section of time interval from flow signals, axial acceleration signals, radial acceleration signals and current signals which are subjected to filtering pretreatment in the same operation state, taking the number N of discrete data points of the discrete data unit with the most discrete data points in four sections of discrete data units as a reference N, amplifying the discrete data points of the other three sections of discrete data units to the number N of the discrete data points equal to the reference N by an interpolation method, and finally connecting the amplified four sections of discrete data units in parallel to form an N4 signal combination data matrix.

6. The method for monitoring the operating state of a rotary blood pump integrating multiple sensors according to claim 4, wherein the method comprises the following steps: the step 1.4) is specifically as follows:

1.4.1 Inputting the signal combination data matrix of N by 4 into the first convolution kernel and the second convolution kernel in sequence to obtain a data matrix of (N-9) by 100 and a data matrix of (N-99) by 100 respectively;

1.4.2 Maximum pooling of (N-9) × 100 data matrix and (N-99) × 100 data matrix simultaneously yields (N-9)/6 × 100 data matrix and (N-99)/6 × 100 data matrix, respectively;

1.4.3 The data matrix of (N-9)/6 x 100 and the data matrix of (N-99)/6 x 100 are respectively input into a first convolution kernel and a second convolution kernel for convolution, then the data matrix of (N-9)/6 x 100 after the convolution and the data matrix of (N-99)/6 x 100 after the convolution are respectively subjected to maximum pooling, and the data matrix of (N-9)/6 x 100 after the maximum pooling and the data matrix of (N-99)/6 x 100 after the maximum pooling are obtained;

1.4.4 Repeating the data matrix of maximum pooled (N-9)/6 x 100 and the data matrix of maximum pooled (N-99)/6 x 100 for D times step 1.4.3) to finally obtain a first data matrix of 1 x 100 and a second data matrix of 1 x 100, respectively;

1.4.5 A) the first data matrix of 1 × 100 and the second data matrix of 1 × 100 are combined in superposition to obtain a one-dimensional vector having a 1 × 100 data matrix structure.

7. The method for monitoring the operating state of a rotary blood pump integrating multiple sensors according to claim 4, wherein the method comprises the following steps: the four operation states comprise a normal working state, a blood backflow state of a blood pump, a ventricular suction collapse state and a thrombus blockage state in the pump.

8. The method for monitoring the operating state of a rotary blood pump integrating multiple sensors according to claim 4, wherein the method comprises the following steps: the interpolation method is piecewise linear interpolation, bilinear interpolation or spline interpolation; the activation function of the one-dimensional convolution adopts a linear rectification function RELU, and the activation function of the full connection layer adopts a normalized exponential function Softmax.

9. The method for monitoring the operating state of a rotary blood pump integrating multiple sensors according to claim 4, wherein the method comprises the following steps: the sizes of convolution kernels of the first convolution kernel and the second convolution kernel are respectively 10 × 4 and 100 × 4, the number of the convolution kernels is 100, and the step length is 1; the maximum pooled nuclei size was 6 with a step size of 6.

10. The method for monitoring the operating state of a rotary blood pump integrating multiple sensors according to claim 4, wherein the method comprises the following steps: the number of repetitions D of step 1.4.3) is from 1 to 3.

Images