Optical signal processing algorithm based on biological agent fluorescence detection

Abstract

The invention relates to the technical field of optical signal processing, in particular to an optical signal processing algorithm based on biological agent fluorescence detection, which comprises the following steps: S1, collecting fluorescence signals with different wavelengths emitted by each probe, and generating a multi-channel fluorescence spectrum data set; s2, carrying out spectral de-mixing on the fluorescence signals in the fluorescence spectrum data set, separating a plurality of overlapped fluorescence signals, and decoupling the fluorescence signals with different wavelengths; s3, for highly overlapped fluorescence signals, carrying out secondary overlapping separation treatment on the fluorescence signals by using a nonlinear optimization method; s4, extracting the intensity, wavelength and time change characteristics of each fluorescent probe signal, and constructing a multi-dimensional fluorescent signal characteristic vector; and S5, by amplifying the characteristic peak value of the weak fluorescence signal, it is ensured that the signal is not covered by strong signals of other probes in multi-probe detection. According to the invention, the resolution of fluorescence detection is improved, so that the high-precision detection effect can be maintained under complex experimental conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of optical signal processing, and in particular to an optical signal processing algorithm based on fluorescence detection of biological agents. Background Art

[0002] As a highly sensitive and selective analytical method, fluorescence detection technology is widely used in biomedicine, environmental monitoring, chemical analysis and other fields. Through the specific reaction of fluorescent probes with target substances, fluorescence detection can achieve quantitative and qualitative analysis of trace substances. However, with the complexity of application scenarios, multi-probe fluorescence detection technology faces many challenges, including the overlap of fluorescence signals, the difficulty of detecting weak fluorescence signals, and the influence of signal interference in complex environments.

[0003] In some multi-labeled fluorescence detection applications, multiple fluorescent probes are used simultaneously, and each probe emits fluorescence of a different wavelength. However, in actual detection, these fluorescence signals often overlap or interfere with each other, resulting in difficulty in signal separation, especially when the signal intensity is weak and the background noise is high. How to efficiently and accurately separate fluorescence signals of different wavelengths is a technical challenge. Current separation algorithms, such as linear separation or simple filtering methods, have limited accuracy and are difficult to use in complex environments, making it difficult to accurately detect low-concentration samples or trace target substances. Summary of the invention

[0004] The present invention provides an optical signal processing algorithm based on fluorescence detection of biological agents.

[0005] The optical signal processing algorithm based on fluorescence detection of biological agents includes the following steps:

[0006] S1, multi-channel acquisition of fluorescence signals: multiple fluorescent probes are excited simultaneously by a multi-channel fluorescence detection device, and the fluorescence signals of different wavelengths emitted by each probe are collected to generate a multi-channel fluorescence spectrum data set;

[0007] S2, spectral unmixing: spectral unmixing of fluorescence signals in the fluorescence spectral dataset is performed, and multiple overlapping fluorescence signals are separated using the non-negative matrix factorization (NMF) algorithm. Based on the spectral characteristics of each fluorescent probe, the fluorescence signals of different wavelengths are decoupled, and the separated signals are ensured to have high independence and accuracy in wavelength;

[0008] S3, secondary de-overlapping: combining dynamic wavelength selection and signal peak matching method, for highly overlapping fluorescence signals, a nonlinear optimization method is used to perform secondary overlapping separation on the fluorescence signals, ensuring that the signal of each fluorescent probe can be accurately separated and is not interfered by other probes;

[0009] S4, multidimensional feature analysis: Based on the separated fluorescence signals, a multidimensional feature analysis algorithm is used to extract the intensity, wavelength and time variation characteristics of each fluorescent probe signal, and construct a multidimensional fluorescence signal feature vector;

[0010] S5, signal enhancement and compensation: Based on the constructed multi-dimensional fluorescence signal feature vector, weak fluorescence signals are identified. For the fluorescent probes corresponding to the weak fluorescence signals, an adaptive signal enhancement algorithm is used to amplify the characteristic peak of the weak fluorescence signal to ensure that the signal is not masked by the strong signals of other probes in multi-probe detection. At the same time, nonlinear signal compensation technology is used to restore part of the fluorescence data affected by other signals.

[0011] Optionally, the S1 specifically includes:

[0012] S12, using multiple independent excitation sources of a multi-channel fluorescence detection device to simultaneously excite multiple fluorescent probes in the biological sample, each excitation source independently controls the excitation light of different wavelengths to ensure that different fluorescent probes are excited at their optimal excitation wavelengths;

[0013] S13, fluorescence signal acquisition: the multi-channel fluorescence detection device is equipped with multiple independent spectral detectors, each spectral detector corresponds to a different wavelength range, and is used to capture the fluorescence signals emitted by different fluorescent probes. The wavelength of the spectral detector covers the complete spectral range from short wavelength to long wavelength to avoid signal omission;

[0014] S14, spectral separation and data synchronization: spectrally separate the fluorescence signals of different wavelengths through the grating element, and guide the separated fluorescence signals to the corresponding spectral detectors respectively, so that the fluorescence signals of each probe do not interfere with each other;

[0015] S15, the fluorescence signals collected from multiple channels are aggregated to generate a multi-channel fluorescence spectral data set, ensuring that the fluorescence signals of different probes are recorded synchronously in time, providing a unified spatiotemporal data benchmark for subsequent spectral unmixing and signal processing.

[0016] Optionally, the S2 specifically includes:

[0017] S21, extracting the characteristic wavelength, fluorescence intensity and time variation characteristics of each fluorescence signal from the fluorescence spectrum data set, and constructing a preliminary fluorescence signal characteristic matrix, where each column in the matrix represents the fluorescence signal intensity distribution of each probe at different wavelengths;

[0018] S22, using the known spectral characteristics of the fluorescent probe as input, initializing a non-negative matrix decomposition model, wherein the input matrix is ​​the mixed fluorescence signal of each probe, and the decomposition target matrix represents the spectral weight matrix and the characteristic signal matrix of the fluorescence signal respectively;

[0019] S23, iterative optimization decomposition: a non-negative matrix decomposition algorithm is used to perform iterative optimization in combination with the characteristic information of the fluorescent probe and the background noise constraint. In each iteration, the spectral weight matrix W and the characteristic signal matrix H are updated to ensure that the signal of each probe can be gradually separated from the mixed signal, maintaining non-negativity and reducing overlap.

[0020] Optionally, during the iterative optimization decomposition process, a signal sparsity constraint is introduced to ensure that the signals corresponding to each wavelength in the decomposed signal matrix are independent, minimize signal overlap between probes, and ensure that the separated fluorescent signals do not interfere with each other in the spectrum.

[0021] Optionally, the sparsity constraint is expressed as:

[0022] in,‖·‖ F represents the Frobenius norm, which is used to measure the matrix difference, α is the sparsity regularization coefficient, ‖H‖ 1 It is the sparsity constraint term of the matrix H, which ensures the sparsity of the matrix H and thus reduces signal overlap.

[0023] Optionally, the S3 specifically includes:

[0024] S31, dynamic wavelength selection: Based on the fluorescence spectral data set, for highly overlapping fluorescence signals, the wavelength characteristics of the fluorescent probes are analyzed in real time to dynamically adjust the wavelength range of signal detection. By using the known fluorescence spectrum peak of each probe, the signal collection focus is concentrated on the most characteristic wavelength range of the probe, reducing interference between signals and optimizing wavelength selection to maximize separation efficiency.

[0025] S32, signal peak matching: After dynamic wavelength selection, the separated fluorescence signals are peak matched. By detecting the peak position and intensity of each fluorescence signal, combined with the standard spectral data of each probe, the actual detection signal is matched with the known signal peak to ensure accurate signal separation. Gaussian fitting is used to optimize the peak position to maximize the separation effect of overlapping signals. The goal of the matching process is to minimize the peak position error, which is expressed as: Among them, E i is the peak matching error of probe i, is the peak wavelength actually detected by probe i, is the peak wavelength of probe i in the standard spectrum;

[0026] S33, nonlinear optimization separation: Based on the results of dynamic wavelength selection and peak matching, a nonlinear optimization method is used to perform secondary separation of overlapping signals. By constructing the objective function f(X), where X is the fluorescence intensity matrix after separation of the signal, the objective function f(X) simultaneously considers the signal peak position, intensity distribution and wavelength variation trend. The constrained optimization algorithm is used to iteratively optimize the overlapping signals to ensure that the signal of each probe can be accurately separated.

[0027] Optionally, the objective function uses a constrained optimization algorithm to minimize the error caused by signal overlap, expressed as:

[0028] Among them, I ij is the actual detected wavelength of probe i at λ j The fluorescence intensity at is the wavelength λ of probe i estimated by optimization. j The fluorescence intensity at , p is the number of probes, n is the number of wavelength collection points, and the objective function f(X) is minimized by nonlinear optimization. The optimization steps are:

[0029] initialization is the separated signal;

[0030] Calculate the objective function f(X) and the gradient;

[0031] Iteratively update the signal estimate using the conjugate gradient algorithm

[0032] Repeat the calculation and iteration until f(X) converges to the minimum value.

[0033] Optionally, the S4 specifically includes:

[0034] S41, data set construction after signal separation: Based on the fluorescence signals processed by S2 spectral unmixing and S3 secondary overlap separation steps, a separated multidimensional fluorescence signal data set X is constructed. sep , including information about each fluorescent probe at different wavelengths, time points, and intensity distribution. sep ∈R p×m×n , where p represents the number of fluorescent probes, m represents the number of wavelengths, and n represents the number of different time points;

[0035] S42, multi-dimensional feature extraction: Analyze the signal changes of each fluorescent probe in different dimensions, and use multi-dimensional feature analysis algorithms (principal component analysis and wavelet transformation method) to extract the following features:

[0036] Intensity characteristics: The intensity distribution characteristic vector of each probe is constructed by calculating the fluorescence intensity I(λ, t) of the fluorescence signal at each wavelength λ and time point t;

[0037] Wavelength characteristics: Extract the characteristic wavelength of each probe That is, the maximum signal response area of ​​the probe at the wavelength after demixing, and the wavelength change pattern is recorded;

[0038] Time variation characteristics: Analyze the decay or growth trend of the fluorescence signal intensity over time I(t) and construct a time variation characteristic curve to reflect the dynamic changes of the fluorescence signal;

[0039] S43, constructing a multidimensional feature vector: constructing a multidimensional feature vector V for each fluorescent probe based on the extracted intensity, wavelength and time features i :

[0040] in, Indicates that probe i is at wavelength Location, time point t j The strength of is the characteristic wavelength of probe i, T i Represents the time-varying feature vector of probe i, describing the curve of its intensity variation over time.

[0041] Optionally, the S5 specifically includes:

[0042] S51, weak fluorescence signal recognition: based on the constructed multi-dimensional fluorescence signal feature vector V i , by calculating the signal intensity of each probe With standard intensity threshold I threshold Compare and identify weak fluorescence signals below the standard intensity threshold if the following conditions are met: The fluorescent signal corresponding to the probe is marked as a weak signal;

[0043] S52, Adaptive signal enhancement: For the identified weak fluorescence signal, its characteristic peak is amplified through the adaptive signal enhancement algorithm Enhancement factor G i Dynamically adjust according to the ratio of the strength of the weak signal to the threshold;

[0044] S53, nonlinear signal compensation: for the enhanced weak fluorescence signal, the nonlinear signal compensation technology is used to restore the part of the fluorescence signal distorted by the strong signal interference of other probes, and a compensation function C(λ,t) is established to represent the correction amount of the signal distortion after signal enhancement.

[0045] Optionally, the compensation function C(λ, t) is expressed as:

[0046] C(λ,t)=δ·(I enhanced (λ,t)-I reference( λ,t)), where δ is the compensation coefficient, Ienhanced (λ,t) is the enhanced fluorescence signal, I reference (λ, t) is the reference target signal value after compensation. The distorted part of the signal is corrected by the compensation function, making the overall waveform of the weak fluorescence signal closer to the real signal.

[0047] Beneficial effects of the present invention:

[0048] The present invention, by introducing dynamic wavelength selection and peak matching method, combines with improved non-negative matrix decomposition algorithm, effectively solves the problem of multi-probe fluorescence signal overlap, and through spectral unmixing and secondary overlap separation processing, can not only separate highly overlapping fluorescence signals in complex environments, but also ensure that the signal of each probe has high independence and accuracy. Especially in the case of high overlap, residual signal interference is further eliminated through nonlinear optimization technology, so that the separated signal is more stable and accurate. This multi-level signal separation and de-overlapping method improves the resolution of fluorescence detection, so that the present invention can maintain a high-precision detection effect under complex experimental conditions.

[0049] The present invention constructs a multi-dimensional fluorescence signal feature vector after separation and overlap, extracts detailed features of each probe in the dimensions of intensity, wavelength and time variation, and can more accurately describe and analyze the separated fluorescence signal. This feature extraction method provides a basis for subsequent signal enhancement and compensation. By identifying weak fluorescence signals and using an adaptive signal enhancement algorithm, the enhancement amplitude of the weak signal can be dynamically adjusted so that it is not masked by the strong signals of other probes. At the same time, the nonlinear signal compensation technology can further restore the interfered part of the signal to ensure the integrity and authenticity of the signal. This adaptive signal enhancement and compensation mechanism improves the detection sensitivity of weak signals and is particularly suitable for detection under multi-probe and weak signal conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 A schematic diagram of an optical signal processing algorithm flow chart of an embodiment of the present invention;

[0052] Figure 2 Schematic diagram of fluorescence signal separation according to an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0054] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0055] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0056] like Figure 1-Figure 2 As shown, the optical signal processing algorithm based on fluorescence detection of biological agents includes the following steps:

[0057] S1, multi-channel acquisition of fluorescence signals: multiple fluorescent probes are excited simultaneously by multi-channel fluorescence detection equipment, and the fluorescence signals of different wavelengths emitted by each probe are collected to generate a multi-channel fluorescence spectrum data set, ensuring that the fluorescence signals of all probes are recorded in real time on different wavelength channels, covering the entire range from short wavelength to long wavelength;

[0058] S2, spectral unmixing: spectral unmixing of fluorescence signals in the fluorescence spectral dataset is performed, and multiple overlapping fluorescence signals are separated using the non-negative matrix factorization (NMF) algorithm. Based on the spectral characteristics of each fluorescent probe, the fluorescence signals of different wavelengths are decoupled, and the separated signals are ensured to have high independence and accuracy in wavelength;

[0059] S3, secondary de-overlapping: combining dynamic wavelength selection and signal peak matching method, for highly overlapping fluorescence signals, a nonlinear optimization method is used to perform secondary overlapping separation on the fluorescence signals, ensuring that the signal of each fluorescent probe can be accurately separated and is not interfered by other probes;

[0060] S4, multidimensional feature analysis: Based on the separated fluorescence signals, a multidimensional feature analysis algorithm is used to extract the intensity, wavelength and time variation characteristics of each fluorescent probe signal, and construct a multidimensional fluorescence signal feature vector;

[0061] S5, signal enhancement and compensation: Based on the constructed multi-dimensional fluorescence signal feature vector, weak fluorescence signals are identified. For the fluorescent probes corresponding to the weak fluorescence signals, an adaptive signal enhancement algorithm is used to amplify the characteristic peak of the weak fluorescence signal to ensure that the signal is not masked by the strong signals of other probes in multi-probe detection. At the same time, nonlinear signal compensation technology is used to restore part of the fluorescence data affected by other signals.

[0062] S1 specifically includes:

[0063] S12, using multiple independent excitation sources of a multi-channel fluorescence detection device to simultaneously excite multiple fluorescent probes in the biological sample, each excitation source independently controls the excitation light of different wavelengths to ensure that different fluorescent probes are excited at their optimal excitation wavelengths;

[0064] S13, fluorescence signal acquisition: The multi-channel fluorescence detection equipment is equipped with multiple independent spectral detectors, each of which corresponds to a different wavelength range and is used to capture the fluorescence signals emitted by different fluorescent probes. The wavelength of the spectral detector covers the complete spectral range from short wavelength to long wavelength to avoid signal omission;

[0065] S14, spectral separation and data synchronization: spectrally separate the fluorescence signals of different wavelengths through the grating element, and guide the separated fluorescence signals to the corresponding spectral detectors respectively, so that the fluorescence signals of each probe do not interfere with each other, ensuring the independence and accuracy of data collection, as follows:

[0066] S141, the fluorescence signal first enters the grating element through the detection device. The grating is an optical element with equally spaced lines. When the fluorescence signals of multiple wavelengths are irradiated to the grating, since different wavelengths of light have different diffraction angles, the grating will diffract these wavelengths at different angles;

[0067] S142, grating diffraction separation: The grating diffracts the incoming fluorescent signal. According to the different wavelengths of light, short-wavelength light will be diffracted to a smaller angle, while long-wavelength light will be diffracted to a larger angle. This diffraction effect separates light of different wavelengths in space, forming an angle range corresponding to different wavelengths, thus completing the separation of the spectrum.

[0068] S143, optical path adjustment: The separated fluorescence signals are guided into different optical paths according to the wavelength and diffraction angle. Each optical path uses lenses or optical fibers to transmit the signals to a specific detector channel for a specific wavelength range. The purpose of optical path adjustment is to ensure that signals of different wavelengths enter a dedicated detector to avoid confusion between wavelengths.

[0069] S144, signal-guided detector: each separated fluorescence signal is guided to a specific spectral detector according to its wavelength. The spectral detector may be a photomultiplier tube, a charge-coupled device or a CMOS sensor. Each detector independently detects the separated wavelength signal. Each detector only receives the light signal within the wavelength range it is responsible for, thereby ensuring accurate detection and recording of the fluorescence signal.

[0070] S15, the fluorescence signals collected from multiple channels are aggregated to generate a multi-channel fluorescence spectral data set, ensuring that the fluorescence signals of different probes are recorded synchronously in time, providing a unified spatiotemporal data benchmark for subsequent spectral unmixing and signal processing.

[0071] S2 specifically includes:

[0072] S21, the principal component analysis (PCA) method is used to extract the characteristic wavelength, fluorescence intensity and time variation characteristics of each fluorescence signal from the fluorescence spectrum data set, and a preliminary fluorescence signal characteristic matrix is ​​constructed. Each column in the matrix represents the fluorescence signal intensity distribution of each probe at different wavelengths;

[0073] Based on the multi-channel fluorescence spectrum data set, the characteristic wavelength, fluorescence intensity and time variation of the fluorescent probe are extracted to construct a preliminary fluorescence signal feature matrix. The fluorescence signal data set is assumed to be a matrix X∈R m×n , where m represents the number of wavelengths for spectral measurement and n represents the number of fluorescence signal acquisitions at different time points.

[0074] The extracted features are:

[0075] Wavelength characteristics: Assume that the fluorescence peak wavelength of the probe is at λ i , then the signal intensity at this wavelength is extracted from the fluorescence signal;

[0076] Fluorescence intensity: For different probes, the signal intensity I(t) is calculated by the fluorescence spectrum intensity;

[0077] Time-varying features: Record the changes in fluorescence intensity over time I(t), which reflects the attenuation or change trend of the fluorescence signal. Time-varying features can be extracted using a more suitable wavelet transform method.

[0078] Each column in the preliminary feature matrix X represents the fluorescence intensity at different wavelengths at a certain moment:

[0079] Among them, I(λ i , t j ) indicates the wavelength λ i and time point t j Collected fluorescence intensity.

[0080] S22, using the known spectral characteristics of the fluorescent probe as input, initializing a non-negative matrix decomposition model, wherein the input matrix is ​​the mixed fluorescence signal of each probe, and the decomposition target matrix represents the spectral weight matrix and the characteristic signal matrix of the fluorescence signal respectively;

[0081] Non-negative matrix factorization (NMF) is the core algorithm for decoupling mixed signals. The goal of NMF is to decompose the input mixed fluorescence signal matrix X into the product of two non-negative matrices:

[0082] X≈WH, where W∈R m×k Represents the spectral weight matrix, representing the characteristic spectra of different probes at each wavelength, H∈R k×n represents the characteristic signal matrix, which represents the fluorescence intensity of different probes over time. k represents the number of probes. NMF requires that both W and H are non-negative, that is, all matrix elements must be non-negative.

[0083] Initialization: Using the known spectral characteristics λ of the fluorescent probe i , W can be initialized to an approximately known spectral weight matrix: Each column of the matrix W represents the characteristic wavelength spectral distribution of a probe.

[0084] S23, iterative optimization decomposition: a non-negative matrix decomposition algorithm is used to perform iterative optimization in combination with the characteristic information of the fluorescent probe and the background noise constraint. In each iteration, the spectral weight matrix W and the characteristic signal matrix H are updated to ensure that the signal of each probe can be gradually separated from the mixed signal, maintaining non-negativity and reducing overlap.

[0085] Unlike the multidimensional feature extraction in S4, the feature extraction in the spectral unmixing step is for preliminary signal separation. In the spectral unmixing step, the purpose of feature extraction is to provide a basis for unmixing and help the non-negative matrix factorization algorithm to better separate overlapping signals. Therefore, the feature extraction here is more based on the known spectral characteristics of each fluorescent probe, such as its typical wavelength, fluorescence intensity pattern, etc., which are the input parameters or initial conditions provided for the unmixing algorithm. In S4, after the fluorescent signal has been successfully separated by the spectral unmixing step, the feature extraction here is to conduct an in-depth analysis of the separated signal. By using a multidimensional feature analysis algorithm, the intensity, wavelength and time variation characteristics of the separated fluorescent signal are extracted and analyzed more finely. It not only extracts the signal of each probe, but also constructs a multi-dimensional feature vector for subsequent signal enhancement, compensation and further data analysis. This step is to perform feature analysis on the separated signal instead of serving unmixing.

[0086] Iterative optimization decomposition is iteratively optimized through NMF, gradually updating the matrices W and H. The iterative update rule is:

[0087]

[0088] After each iteration, the spectral weight matrix W and the characteristic signal matrix H are updated respectively, and the fluorescence signals of different probes are gradually separated from the mixed signals. T is the transpose of the characteristic signal matrix, W T It is the transpose of the spectral weight matrix, Wij represents the spectral weight of the i-th wavelength for the j-th probe; Hij represents the fluorescence intensity of the j-th probe at the i-th time point.

[0089] During the iterative optimization decomposition process, signal sparsity constraints are introduced to ensure that the signals corresponding to each wavelength in the decomposed signal matrix are independent, minimize the signal overlap between probes, and ensure that the separated fluorescence signals do not interfere with each other in the spectrum.

[0090] The sparsity constraint ensures that the signal matrix H of each probe has significant signal values ​​only within the characteristic wavelength range of the probe, ensuring signal decoupling, which is expressed as:

[0091] in,‖·‖ F represents the Frobenius norm, which is used to measure the matrix difference, α is the sparsity regularization coefficient, ‖H‖ 1 It is the sparsity constraint term of the matrix H, which ensures the sparsity of the matrix H and thus reduces signal overlap.

[0092] After decomposition, the separated characteristic signal matrix H and spectral weight matrix W are obtained. Next, through the verification step, the separated signal is compared with the spectral characteristics of the known fluorescent probe, the separation accuracy is adjusted, and the characteristic wavelength λ is compared. i range, ensure that the spectral weight W after separation conforms to the known spectral characteristics of the probe, and verify whether the separated signal meets the expectations based on the known intensity distribution of the fluorescent probe. Through verification and adjustment, ensure that the fluorescent signal of each probe is separated correctly, and finally complete the demixing.

[0093] S3 specifically includes:

[0094] S31, dynamic wavelength selection: based on the fluorescence spectrum data set, for highly overlapping fluorescence signals, the wavelength characteristics of the fluorescent probes are analyzed in real time, the wavelength range of signal detection is dynamically adjusted, and the known fluorescence spectrum peak of each probe is used to focus the signal collection on the most characteristic wavelength range of the probe, reduce interference between signals, and optimize wavelength selection to maximize separation efficiency. The most characteristic wavelength range of the probe refers to the strongest fluorescence emission peak and the wavelength range near the fluorescence signal of the probe after excitation. Each fluorescent probe has its own specific fluorescence emission wavelength, which is manifested as one or more peaks on the spectrum. These peak wavelengths are the areas that best represent the signal characteristics of the probe; specifically: the peak wavelength (denoted as λmax) is the wavelength of the strongest light signal generated by the probe in the fluorescence emission spectrum, and the characteristic wavelength range refers to the main wavelength range of the fluorescence signal emitted by the probe, forming a narrow range around the left and right parts of the peak wavelength λmax. For example, a probe with a peak wavelength of 520nm may have a characteristic wavelength range between 510nm and 530nm; dynamic wavelength selection is to focus the collection on the characteristic wavelength range of each probe to reduce interference between fluorescence signals. Suppose there are p fluorescent probes, and the characteristic wavelength range of each probe is in and are the minimum and maximum characteristic wavelengths of probe i, respectively, expressed as: in, is the optimal acquisition wavelength of probe i, that is, the position where the signal intensity I(λ) is the largest within its characteristic wavelength range. I(λ) is the fluorescence intensity at wavelength λ. Through the dynamic wavelength selection process, the focus is on collecting nearby fluorescent signals and reduce interfering signals at other wavelengths.

[0095] S32, signal peak matching: After dynamic wavelength selection, the separated fluorescence signals are peak matched. By detecting the peak position and intensity of each fluorescence signal, combined with the standard spectral data of each probe, the actual detection signal is matched with the known signal peak to ensure accurate signal separation. Gaussian fitting is used to optimize the peak position to maximize the separation effect of overlapping signals. The goal of the matching process is to minimize the peak position error, which is expressed as: Among them, E i is the peak matching error of probe i, is the peak wavelength actually detected by probe i, is the peak wavelength of probe i in the standard spectrum;

[0096] During the peak matching process, the Gaussian fitting method is used to estimate the position and intensity of the peak signal to maximize the signal separation effect. The Gaussian fitting is expressed as:

[0097] Where A is the peak intensity, is the center position of the Gaussian distribution, that is, the peak wavelength, and σ is the peak width, which indicates the extension range of the spectral peak. By fitting the signal, the actual peak position is determined. And with the standard peak Make a match;

[0098] S33, nonlinear optimization separation: Based on the results of dynamic wavelength selection and peak matching, a nonlinear optimization method is used to perform secondary separation of overlapping signals. By constructing the objective function f(X), where X is the fluorescence intensity matrix after separation of the signal, the objective function f(X) simultaneously considers the signal peak position, intensity distribution and wavelength variation trend. The constrained optimization algorithm is used to iteratively optimize the overlapping signals to ensure that the signal of each probe can be accurately separated.

[0099] The objective function uses a constrained optimization algorithm to minimize the error caused by signal overlap and is expressed as:

[0100] Among them, I ij is the actual detected wavelength of probe i at λ j The fluorescence intensity at is the wavelength λ of probe i estimated by optimization. j The fluorescence intensity at , p is the number of probes, n is the number of wavelength collection points, and the objective function f(X) is minimized by nonlinear optimization. The optimization steps are:

[0101] initialization is the separated signal;

[0102] Calculate the objective function f(X) and the gradient;

[0103] Iteratively update the signal estimate using the conjugate gradient algorithm

[0104] Repeat the calculation and iteration until f(X) converges to the minimum value.

[0105] S4 specifically includes:

[0106] S41, data set construction after signal separation: Based on the fluorescence signals processed by S2 spectral unmixing and S3 secondary overlap separation steps, a separated multidimensional fluorescence signal data set X is constructed. sep , including information about each fluorescent probe at different wavelengths, time points, and intensity distribution. sep ∈R p×m×n , where p represents the number of fluorescent probes, m represents the number of wavelengths, and n represents the number of different time points;

[0107] S42, multi-dimensional feature extraction: Analyze the signal changes of each fluorescent probe in different dimensions, and use multi-dimensional feature analysis algorithms (principal component analysis and wavelet transformation method) to extract the following features:

[0108] Intensity characteristics: The intensity distribution characteristic vector of each probe is constructed by calculating the fluorescence intensity I(λ, t) of the fluorescence signal at each wavelength λ and time point t;

[0109] Wavelength characteristics: Extract the characteristic wavelength of each probe That is, the maximum signal response area of ​​the probe at the wavelength after demixing, and the wavelength change pattern is recorded;

[0110] Time variation characteristics: Analyze the decay or growth trend of the fluorescence signal intensity over time I(t) and construct a time variation characteristic curve to reflect the dynamic changes of the fluorescence signal;

[0111] S43, constructing a multidimensional feature vector: constructing a multidimensional feature vector V for each fluorescent probe based on the extracted intensity, wavelength and time features i :

[0112] in, Indicates that probe i is at wavelength Location, time point t j The strength of is the characteristic wavelength of probe i, T i represents the time-varying characteristic vector of probe i, describing the curve of its intensity variation over time;

[0113] By constructing a multidimensional feature vector V iFurther analysis is performed to identify the fluorescence signal pattern of each probe, combining the multidimensional characteristics of time, wavelength and intensity for classification and pattern recognition to ensure the accuracy and consistency of the signal of each probe under different detection conditions.

[0114] S5 specifically includes:

[0115] S51, weak fluorescence signal recognition: based on the constructed multi-dimensional fluorescence signal feature vector V i , by calculating the signal intensity of each probe With standard intensity threshold I threshold By comparison, weak fluorescence signals below the standard intensity threshold are identified if the following conditions are met: The fluorescence signal corresponding to the probe is marked as a weak signal; threshold The specific value of I is determined by the experimental settings and the characteristics of the fluorescent probe. In general, I threshold It can be set to 10-20% of the fluorescence signal intensity generated by the probe under standard conditions, that is: I threshold =0.1×I max To 0.2×I max , where I max It is the peak fluorescence intensity of the probe under standard excitation conditions (without interference). By setting the threshold in the range of 10-20%, weak signals can be effectively distinguished from normal signals, ensuring that the enhancement algorithm only processes weak signals below this intensity.

[0116] S52, Adaptive signal enhancement: For the identified weak fluorescence signal, its characteristic peak is amplified through the adaptive signal enhancement algorithm Enhancement factor G i Dynamically adjust according to the ratio of weak signal strength to threshold: Among them, α is the adaptive enhancement coefficient, which ensures that the amplitude of signal enhancement is inversely proportional to the signal strength, dynamically improving the strength of weak signals so that they are not masked by the strong signals of other probes in multi-probe detection;

[0117] S53, nonlinear signal compensation: for the enhanced weak fluorescence signal, the nonlinear signal compensation technology is used to restore the part of the fluorescence signal distorted by the strong signal interference of other probes, and the compensation function C(λ,t) is established to represent the correction amount of the signal distortion after signal enhancement;

[0118] The enhanced signal is verified through the feedback mechanism to ensure that the compensated signal is consistent with the fluorescence characteristics of the actual probe to avoid over-enhancement or under-compensation. If the error between the enhanced signal and the reference signal exceeds the set range, the enhancement factor G is dynamically adjusted. i And compensation coefficient β for precise correction.

[0119] The compensation function C(λ,t) is expressed as:

[0120] C(λ,t)=δ·(I enhanced (λ,t)-I reference (λ,t)), where δ is the compensation coefficient, I enhanced (λ,t) is the enhanced fluorescence signal, I reference (λ, t) is the reference target signal value after compensation. The distorted part of the signal is corrected by the compensation function, making the overall waveform of the weak fluorescence signal closer to the real signal.

[0121] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0122] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An optical signal processing algorithm based on fluorescence detection of biological agents, characterized in that: The following steps are involved: S1, using a multi-channel fluorescence detection device to simultaneously excite multiple fluorescent probes, collect fluorescence signals of different wavelengths emitted by each probe, and generate a multi-channel fluorescence spectrum data set; S2, spectrally unmixing the fluorescence signals in the fluorescence spectral dataset, separating multiple overlapping fluorescence signals using a non-negative matrix factorization algorithm, and decoupling fluorescence signals of different wavelengths based on the spectral characteristics of each fluorescent probe; S3, combining dynamic wavelength selection and signal peak matching method, for highly overlapping fluorescence signals, a nonlinear optimization method is used to perform secondary overlapping separation processing on the fluorescence signal; S4, based on the separated fluorescence signals, a multidimensional feature analysis algorithm is used to extract the intensity, wavelength and time variation characteristics of each fluorescence probe signal, and a multidimensional fluorescence signal feature vector is constructed; S5, signal enhancement and compensation: Based on the constructed multi-dimensional fluorescence signal feature vector, weak fluorescence signals are identified. For the fluorescent probes corresponding to the weak fluorescence signals, an adaptive signal enhancement algorithm is used to amplify the characteristic peak of the weak fluorescence signal to ensure that the signal is not masked by the strong signals of other probes in multi-probe detection. At the same time, nonlinear signal compensation technology is used to restore part of the fluorescence data affected by other signals.

2. The optical signal processing algorithm based on fluorescence detection of biological agents according to claim 1 is characterized in that: The S1 specifically includes: S12, using multiple independent excitation sources of a multi-channel fluorescence detection device to simultaneously excite multiple fluorescent probes in the biological sample, each excitation source independently controls the excitation light of different wavelengths to ensure that different fluorescent probes are excited at their optimal excitation wavelengths; S13, fluorescence signal acquisition: the multi-channel fluorescence detection device is equipped with multiple independent spectral detectors, each spectral detector corresponds to a different wavelength range, and is used to capture the fluorescence signals emitted by different fluorescent probes. The wavelength of the spectral detector covers the complete spectral range from short wavelength to long wavelength; S14, spectral separation and data synchronization: spectrally separate the fluorescence signals of different wavelengths through the grating element, and guide the separated fluorescence signals to the corresponding spectral detectors respectively, so that the fluorescence signals of each probe do not interfere with each other; S15, the fluorescence signals collected from multiple channels are aggregated to generate a multi-channel fluorescence spectral data set, ensuring that the fluorescence signals of different probes are recorded synchronously in time, providing a unified spatiotemporal data benchmark for subsequent spectral unmixing and signal processing.

3. The optical signal processing algorithm based on fluorescence detection of biological agents according to claim 1 is characterized in that: The S2 specifically includes: S21, extracting the characteristic wavelength, fluorescence intensity and time variation characteristics of each fluorescence signal from the fluorescence spectrum data set, and constructing a preliminary fluorescence signal characteristic matrix, where each column in the matrix represents the fluorescence signal intensity distribution of each probe at different wavelengths; S22, using the known spectral characteristics of the fluorescent probe as input, initializing a non-negative matrix decomposition model, wherein the input matrix is ​​the mixed fluorescence signal of each probe, and the decomposition target matrix represents the spectral weight matrix and the characteristic signal matrix of the fluorescence signal respectively; S23, iterative optimization decomposition: a non-negative matrix decomposition algorithm is used to perform iterative optimization in combination with the characteristic information of the fluorescent probe and the background noise constraint. In each iteration, the spectral weight matrix W and the characteristic signal matrix H are updated to ensure that the signal of each probe can be gradually separated from the mixed signal, maintaining non-negativity and reducing overlap.

4. The optical signal processing algorithm based on fluorescence detection of biological agents according to claim 3 is characterized in that: During the iterative optimization decomposition process, signal sparsity constraints are introduced to ensure that the signals corresponding to each wavelength in the decomposed signal matrix are independent, thereby reducing signal overlap between probes.

5. The optical signal processing algorithm based on fluorescence detection of biological agents according to claim 4 is characterized in that: The sparsity constraint is expressed as: in,‖·‖ F represents the Frobenius norm, which is used to measure the matrix difference, α is the sparsity regularization coefficient, and ‖H‖1 is the sparsity constraint term of the matrix H, which ensures the sparsity of the matrix H and thus reduces signal overlap.

6. The optical signal processing algorithm based on fluorescence detection of biological agents according to claim 1 is characterized in that: The S3 specifically includes: S31, dynamic wavelength selection: Based on the fluorescence spectrum data set, for highly overlapping fluorescence signals, the wavelength characteristics of the fluorescent probes are analyzed in real time to dynamically adjust the wavelength range of signal detection. By using the known fluorescence spectrum peak of each probe, the signal collection focus is concentrated on the most characteristic wavelength range of the probe; S32, signal peak matching: After dynamic wavelength selection, the separated fluorescence signals are peak matched. By detecting the peak position and intensity of each fluorescence signal, combined with the standard spectral data of each probe, the actual detection signal is matched with the known signal peak to ensure accurate signal separation. Gaussian fitting is used to optimize the peak position to maximize the separation effect of overlapping signals. The goal of the matching process is to minimize the peak position error, which is expressed as: Among them, E i is the peak matching error of probe i, is the peak wavelength actually detected by probe i, is the peak wavelength of probe i in the standard spectrum; S33, nonlinear optimization separation: Based on the results of dynamic wavelength selection and peak matching, a nonlinear optimization method is used to perform secondary separation of overlapping signals. By constructing the objective function f(X), where X is the fluorescence intensity matrix after separation of the signal, the objective function f(X) simultaneously considers the signal peak position, intensity distribution and wavelength change trend, and uses a constrained optimization algorithm to iteratively optimize the overlapping signals.

7. The optical signal processing algorithm based on fluorescence detection of biological agents according to claim 6 is characterized in that: The objective function uses a constrained optimization algorithm to minimize the error caused by signal overlap and is expressed as: Among them, I ij is the actual detected wavelength of probe i at λ j The fluorescence intensity at is the wavelength λ of probe i estimated by optimization. j Where p is the number of probes, n is the number of wavelength collection points, and the objective function f(X) is minimized by nonlinear optimization.

8. The optical signal processing algorithm based on fluorescence detection of biological agents according to claim 1 is characterized in that: The S4 specifically includes: S41, Dataset construction after signal separation: Construct the separated multidimensional fluorescence signal dataset X sep , including information about each fluorescent probe at different wavelengths, time points, and intensity distribution. sep ∈R p×m×n , where p represents the number of fluorescent probes, m represents the number of wavelengths, and n represents the number of different time points; S42, multi-dimensional feature extraction: Analyze the signal changes of each fluorescent probe in different dimensions, and use the multi-dimensional feature analysis algorithm to extract the following features: Intensity characteristics: The intensity distribution characteristic vector of each probe is constructed by calculating the fluorescence intensity I(λ, t) of the fluorescence signal at each wavelength λ and time point t; Wavelength characteristics: Extract the characteristic wavelength of each probe That is, the maximum signal response area of ​​the probe at the wavelength after demixing, and the wavelength change pattern is recorded; Time variation characteristics: Analyze the decay or growth trend of the fluorescence signal intensity over time I(t) and construct a time variation characteristic curve to reflect the dynamic changes of the fluorescence signal; S43, constructing a multidimensional feature vector: constructing a multidimensional feature vector V for each fluorescent probe based on the extracted intensity, wavelength and time features i : in, Indicates that probe i is at wavelength Location, time point t j The strength of is the characteristic wavelength of probe i, T i Represents the time-varying feature vector of probe i, describing the curve of its intensity variation over time.

9. The optical signal processing algorithm based on fluorescence detection of biological agents according to claim 8, characterized in that: The S5 specifically includes: S51, weak fluorescence signal recognition: based on the constructed multi-dimensional fluorescence signal feature vector V i , by calculating the signal intensity of each probe With standard intensity threshold I threshold By comparison, weak fluorescence signals below the standard intensity threshold are identified if the following conditions are met: The fluorescent signal corresponding to the probe is marked as a weak signal; S52, Adaptive signal enhancement: For the identified weak fluorescence signal, its characteristic peak is amplified through the adaptive signal enhancement algorithm Enhancement factor G i Dynamically adjust according to the ratio of the strength of the weak signal to the threshold; S53, nonlinear signal compensation: for the enhanced weak fluorescence signal, the nonlinear signal compensation technology is used to restore the part of the fluorescence signal distorted by the strong signal interference of other probes, and a compensation function C(λ,t) is established to represent the correction amount of the signal distortion after signal enhancement.

10. The optical signal processing algorithm based on fluorescence detection of biological agents according to claim 9, characterized in that: The compensation function C(λ,t) is expressed as: C(λ,t)=δ·(I enhanced (λ,t)-I reference (λ,t)), where δ is the compensation coefficient, I enhanced (λ,t) is the enhanced fluorescence signal, I reference (λ, t) is the reference target signal value after compensation. The distorted part of the signal is corrected by the compensation function, making the overall waveform of the weak fluorescence signal closer to the real signal.

Images