Intraocular fluid marker rapid detection method based on artificial intelligence

Abstract

The invention relates to the technical field of biological material detection, in particular to an intraocular fluid marker rapid detection method based on artificial intelligence. The method comprises the following steps: firstly, collecting an intraocular fluid sample to be detected; obtaining a physical interference factor according to the color characteristic, the turbid characteristic and the flow characteristic of the intraocular fluid sample; acquiring a nucleic acid quality factor according to the concentration, purity and dominant characteristics of nucleic acid in the intraocular fluid sample; acquiring a matrix inhibition factor according to dripping morphological characteristics and pipetting liquid level characteristics of the intraocular fluid sample; and inputting the physical interference factor, the nucleic acid quality factor and the matrix inhibition factor of the intraocular fluid sample into a pre-trained circulation parameter recommendation model, and carrying out rapid detection on the marker of the intraocular fluid sample according to recommended circulation parameters output by the model. According to the method, the sample characteristic factors which reflect the influence of the intraocular fluid sample on the detection are quantified from different angles, and then the loop parameters are recommended through the neural network model for rapid detection, so that the detection efficiency and effect of the intraocular fluid marker are improved.

Description

Technical Field

[0001] This invention relates to the field of biomaterial detection technology, specifically to a rapid detection method for intraocular fluid biomarkers based on artificial intelligence. Background Technology

[0002] Intraocular fluid is an important biological medium for ophthalmic disease research, containing various biomarkers closely related to diseases, such as inflammatory factors, immune factors, growth factors, metabolites, and proteins. The expression of these biomarkers directly reflects the intraocular state, providing precise molecular-level evidence for early diagnosis and personalized treatment strategies in ophthalmic diseases. Many ophthalmic diseases have irreversible pathological processes, and delayed diagnosis can lead to permanent visual impairment; therefore, rapid detection of intraocular fluid biomarkers is crucial.

[0003] However, intraocular fluid samples are characterized by low volume, low abundance of biomarkers, and high matrix complexity, leading to drawbacks such as low sensitivity and efficiency in traditional detection methods. Currently, nucleic acid amplification technology is mainly used to amplify trace amounts of nucleic acids in intraocular fluid millions of times, enabling the joint detection of multiple biomarkers in limited intraocular fluid samples, thus improving detection sensitivity and efficiency. Cyclic parameters are crucial parameters in nucleic acid amplification technology, determining amplification efficiency and specificity. Inappropriate cyclic parameter settings can lead to low amplification efficiency and non-specific amplification, thereby reducing the accuracy of intraocular fluid biomarker detection. Summary of the Invention

[0004] To address the technical problem of poor detection results for intraocular fluid biomarkers, the present invention aims to provide a rapid detection method for intraocular fluid biomarkers based on artificial intelligence. The specific technical solution adopted is as follows: A rapid detection method for intraocular fluid biomarkers based on artificial intelligence, the method comprising: Collect intraocular fluid samples to be tested; Based on the color, turbidity, and flow characteristics of intraocular fluid samples, physical interference factors were obtained; based on the concentration, purity, and dominant characteristics of nucleic acids in intraocular fluid samples, nucleic acid quality factors were obtained; and based on the drop morphology and pipetting surface characteristics of intraocular fluid samples, matrix inhibition factors were obtained. The physical interference factor, the nucleic acid quality factor, and the matrix inhibition factor of the intraocular fluid sample are input into a pre-trained cyclic parameter recommendation model, and the biomarkers of the intraocular fluid sample are rapidly detected based on the recommended cyclic parameters output by the model.

[0005] Furthermore, the intraocular fluid sample includes at least an aqueous humor sample and a vitreous humor sample.

[0006] Furthermore, the method for obtaining the physical interference factor includes: A preset first dose of aqueous humor subsample is taken from the aqueous humor sample, and the RGB image of the aqueous humor subsample in a static state is acquired. The light transmittance is obtained, and the explicit parameters of the aqueous humor subsample are obtained based on the light transmittance and the R channel value of the pixel in the RGB image. A second dose of vitreous fluid was taken from the vitreous fluid sample and a rheological test was performed. The viscosity parameters of the vitreous fluid sample were obtained based on the flow pressure and flow velocity of the vitreous fluid sample at each moment. The explicit parameters and the viscosity parameters are fused, and the normalized value of the fusion result is used as the physical interference factor.

[0007] Furthermore, the method for obtaining the explicit parameters includes: In an RGB image, the sum of the R channel values ​​of all pixels is used as the numerator, and the sum of the B and G channel values ​​of all pixels is used as the denominator. The ratio of the fractions is used as the bloodiness index. The negative correlation mapping result of the light transmittance is used as the turbidity index. The bloodiness index and the turbidity index are fused, and the fusion result is used as an explicit parameter.

[0008] Furthermore, the method for obtaining the nucleic acid quality factor includes: For each type of intraocular fluid sample, a subsample of the third preset dose was taken for quantitative detection to obtain the nucleic acid concentration parameter of the corresponding subsample; a subsample of the fourth preset dose was taken, and the nucleic acid purity parameter was obtained based on the protein content and organic matter content of the corresponding subsample; a subsample of the fifth preset dose was taken, and the dominant nucleic acid parameter was obtained based on the RNA concentration and DNA concentration of the corresponding subsample. The nucleic acid concentration parameter, nucleic acid purity parameter, and nucleic acid dominant parameter of each intraocular fluid sample are fused, and the normalized value of the fusion result is used as the nucleic acid quality factor.

[0009] Furthermore, the method for obtaining the nucleic acid purity parameter includes: For each intraocular fluid sample, a subsample of a preset fourth dose was subjected to spectrophotometric detection. The A260 / A280 ratio was used as a protein purity parameter, and the A260 / A230 ratio was used as an organic contaminant purity parameter. The protein purity parameter and the organic contaminant purity parameter were fused together, and the fusion result was used as the nucleic acid purity parameter.

[0010] Furthermore, the method for obtaining the matrix repressor factor includes: A sixth dose of vitreous fluid was taken from a vitreous fluid sample and a drop test was performed to obtain a diffusion image of the corresponding vitreous fluid sub-sample after it was dropped. The sample area in the diffusion image was determined, and the gel parameters of the corresponding vitreous fluid sub-sample were obtained based on the boundary ruggedness of the sample area. A pre-set seventh dose of vitreous fluid subsample was taken from the vitreous fluid sample and a pipetting experiment was performed to obtain the liquid surface image of the corresponding vitreous fluid subsample during the pipetting process; the reflective patches in the liquid surface image were determined based on grayscale features, and the contamination parameters of the corresponding vitreous fluid subsample were obtained based on the boundary ruggedness features and grayscale values ​​of the reflective patches. The gel parameters and the contamination parameters are fused, and the normalized value of the fusion result is used as the matrix inhibition factor.

[0011] Furthermore, the method for obtaining the gel parameters includes: Edge detection is performed on the diffused image, and the area enclosed by the edges is taken as the sample area; the variance of the slope between all adjacent edge pixels in the sample area is taken as the gel parameter.

[0012] Furthermore, the method for obtaining the pollution parameters includes: The foreground region in the liquid surface image is obtained based on a threshold segmentation algorithm, and the foreground region is taken as a reflective patch. The variance of the slope between adjacent boundary points in the reflective patch is taken as a boundary ruggedness parameter. The product of the boundary ruggedness parameter and the gray-scale mean of the pixels in the reflective patch is taken as a contamination parameter.

[0013] Furthermore, the method for obtaining the cyclic parameter recommendation model includes: The physical interference factor, nucleic acid quality factor, and matrix inhibition factor of different historical intraocular fluid samples are obtained, and the optimal cycling parameters are labeled for each historical intraocular fluid sample to construct a training sample set; a feedforward multilayer perceptron model is trained based on the training sample set, and the trained model is used as a cycling parameter recommendation model.

[0014] The present invention has the following beneficial effects: This invention collects intraocular fluid samples to prepare for rapid detection of intraocular fluid biomarkers. Then, based on the color, turbidity, and flow characteristics of the intraocular fluid samples, it identifies physical interference factors reflecting hemoglobin contamination or excessive viscosity encapsulating nucleic acids, thus hindering detection. Based on the concentration, purity, and dominant characteristics of nucleic acids in the intraocular fluid samples, it identifies nucleic acid quality factors reflecting the quality of the nucleic acid template. Based on the drop morphology and pipetting surface characteristics of the intraocular fluid samples, it identifies matrix inhibition factors reflecting the inhibition of enzyme activity by proteins, lipids, or interfacial contaminants, thus hindering detection. Finally, the physical interference factors, nucleic acid quality factors, and matrix inhibition factors of the intraocular fluid samples are input into a pre-trained cyclic parameter recommendation model. Rapid biomarker detection in intraocular fluid samples is then performed based on the recommended cyclic parameters output by the model. This invention quantifies sample characteristic factors that affect PCR detection from different perspectives, and then uses a pre-trained model to recommend cyclic parameters for rapid detection, improving the detection efficiency and effectiveness of intraocular fluid biomarkers. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating a rapid detection method for intraocular fluid biomarkers based on artificial intelligence, provided as an embodiment of the present invention; Figure 2 This is a flowchart illustrating a method for obtaining a matrix inhibitory factor according to an embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a rapid detection method for intraocular fluid biomarkers based on artificial intelligence proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] The following description, in conjunction with the accompanying drawings, details a specific scheme for a rapid detection method for intraocular fluid biomarkers based on artificial intelligence provided by this invention.

[0020] Please see Figure 1 The diagram illustrates a flowchart of a rapid detection method for intraocular fluid biomarkers based on artificial intelligence, according to an embodiment of the present invention, specifically including: Step S1: Collect the intraocular fluid sample to be tested.

[0021] In one embodiment of the present invention, an intraocular fluid sample to be tested is first collected to prepare for the rapid detection of intraocular fluid markers; wherein, the intraocular fluid sample includes at least an aqueous humor sample and a vitreous humor sample.

[0022] It should be noted that collecting intraocular fluid samples must be performed by professional operators, as this is an existing technology. Here, we will only briefly describe the general process and precautions.

[0023] (1) First, collect intraocular fluid: Prepare 27-30G sterile microsyringes (30G syringes for collecting aqueous humor to reduce corneal damage; 27G syringes for collecting vitreous fluid to ensure flow rate), disposable sterile puncture needles, and nuclease-free centrifuge tubes (labeled with individual information in advance and pre-cooled to -20℃). After normal intraocular pressure and pre-treatment such as anesthesia and disinfection, insert a 30G syringe needle into the anterior chamber at a 30° angle to the cornea, 1 mm inside the limbus (superior temporal region, avoiding the lens), to a depth of approximately 1-2 mm (entering only the anterior chamber, avoiding contact with the iris or lens), and collect 50-200 μL of aqueous humor; then, insert a 27G needle perpendicularly to the sclera at the pars plana (3.5-4 mm posterior to the limbus), to a depth of 10-12 mm, taking care to avoid contact with the retina, and collect 100-500 μL of vitreous fluid; (2) Treat the aqueous humor and vitreous fluid immediately: Slowly inject the intraocular fluid from each syringe into a nuclease-free centrifuge tube (to avoid aerosol contamination caused by air bubbles); if the sample volume is greater than 200 μL, it needs to be aliquoted into 2-3 tubes (50-100 μL per tube) for subsequent sampling and analysis, and to avoid repeated freeze-thaw cycles; centrifuge at 4℃ and 5000×g for 10 minutes to remove cell debris and tissue particles, and transfer the supernatant to a new tube; the supernatant is the intraocular fluid sample used subsequently. Store the intraocular fluid sample at 4℃ to prevent RNA degradation and preserve sample activity; it should be noted that the aqueous humor and vitreous fluid are processed separately to obtain two types of intraocular fluid samples.

[0024] Step S2: Based on the color, turbidity, and flow characteristics of the intraocular fluid sample, obtain the physical interference factor; based on the concentration, purity, and dominant characteristics of nucleic acids in the intraocular fluid sample, obtain the nucleic acid quality factor; based on the drop morphology and pipetting surface characteristics of the intraocular fluid sample, obtain the matrix inhibition factor.

[0025] Intraocular fluid samples are characterized by low volume, low abundance of biomarkers, and high matrix complexity. Nucleic acid amplification technologies such as polymerase chain reaction (PCR) can simultaneously amplify multiple target sequences (such as multiple cytokines and pathogen genes), enabling the joint detection of multiple biomarkers in limited samples and significantly improving detection efficiency.

[0026] Considering that when red blood cells rupture, the intraocular fluid sample will contain a large amount of hemoglobin and Fe. 2+ Hb plasma components, these substances will inhibit Taq enzyme activity, and Taq enzyme is the most important and core DNA polymerase in the PCR detection process. The inhibition of Taq enzyme activity will directly affect the subsequent PCR detection process. Furthermore, considering that vitreous fluid is rich in collagen and hyaluronic acid, excessively viscous vitreous fluid may embed nucleic acids, leading to insufficient release of nucleic acids; in addition, collagen has an inhibitory effect on PCR enzymes, which may lead to uneven distribution of reaction components in the reaction system. The stronger the gelation degree, the more obvious the impact on the subsequent PCR detection process. Since the color and turbidity characteristics of intraocular fluid samples can help assess the presence of ruptured red blood cells, and the flow characteristics can help assess whether the sample is too viscous, thus helping to assess whether it interferes with PCR detection; based on this, embodiments of the present invention will obtain physical interference factors based on the color, turbidity, and flow characteristics of intraocular fluid samples; physical interference factors initially reflect the degree of interference of intraocular fluid samples with PCR detection at the physical level, preparing for subsequent evaluation of PCR cycling parameters to improve PCR detection results.

[0027] Preferably, in one embodiment of the present invention, considering that aqueous humor is usually a colorless and transparent liquid, when ruptured blood cells are present, the sample will be reddish-brown and relatively turbid, with a relatively high R channel value and poor light transmittance; and considering that vitreous fluid usually has a high viscosity and gel-like characteristic, vitreous fluid can be sampled and analyzed. The higher the flow pressure and the lower the flow rate during the rheological test, the more viscous the vitreous fluid is, which may encapsulate nucleic acids and inhibit PCR enzyme activity. Based on this, the method for obtaining physical interference factors includes: A preset first dose of aqueous humor subsample is taken from the aqueous humor sample, and the RGB image of the aqueous humor subsample in a static state is acquired. The light transmittance is obtained, and the explicit parameters of the aqueous humor subsample are obtained based on the light transmittance and the R channel value of the pixel in the RGB image. A second dose of vitreous fluid was taken from the vitreous fluid sample and a rheological test was performed. The viscosity parameters of the vitreous fluid sample were obtained based on the flow pressure and flow velocity of the vitreous fluid sample at each moment. The explicit parameters and viscosity parameters are fused, and the normalized value of the fusion result is used as the physical interference factor.

[0028] As an example, the preset first dose is set to 5 μl. A 5 μl aqueous humor subsample is taken from the aqueous humor sample for optical analysis. The aqueous humor sample is loaded onto a dedicated optical analysis instrument to avoid the formation of air bubbles, and a light source (white light or light of a specific wavelength) is set for transmittance detection. RGB images are acquired and the transmittance intensity is recorded. Then, based on the transmittance intensity and the R channel value of the pixels in the RGB image, the explicit parameters of the aqueous humor subsample are obtained. The explicit parameters reflect the red turbidity characteristics of the aqueous humor subsample, indirectly reflecting the degree of contamination of the intraocular fluid sample by ruptured blood cells, and also reflecting the degree of influence on PCR detection.

[0029] It should be noted that the acquisition of RGB images and the determination of light transmittance are well-known technical methods and will not be elaborated further; light transmittance, also known as light transmittance, refers to the ratio of the light intensity transmitted through the aqueous humor sample to the incident light intensity, which is a dimensionless ratio.

[0030] In a preferred embodiment of the present invention, considering that a larger R channel value relative to the other two channels in an RGB image indicates a greater likelihood of the presence of hemoglobin, and that a lower light transmittance indicates a higher degree of turbidity, the method for obtaining explicit parameters includes: In an RGB image, the sum of the R channel values ​​of all pixels is used as the numerator, the sum of the B and G channel values ​​of all pixels is used as the denominator, and the ratio of the fractions is used as the blood index; the negative correlation mapping result of the light transmittance is used as the turbidity index; the blood index and the turbidity index are fused, and the fusion result is used as an explicit parameter. Specifically, the logic is adjusted by taking the reciprocal of the light transmittance and using a negative correlation mapping, so that the lower the light transmittance, the higher the turbidity index. Finally, the bloodiness index and the turbidity index are multiplied and fused to obtain the explicit parameter. In other examples, the implementer can also use other negative correlation mapping methods, such as mapping to a negative exponential function, or weighted fusion of the bloodiness index and the turbidity index, which will not be elaborated further.

[0031] Then, the preset second dose was set to 10 μl, and a 10 μl sample of vitreous fluid was taken from the vitreous fluid sample for rheological testing. Specifically, the rheological test was performed using a microfluidic chip. The vitreous fluid sample was injected into the inlet of the microfluidic chip, and the micropump was started to push the vitreous fluid sample through the detection channel at a gradient flow rate (0.5→5.0 μL / min), while the flow pressure and flow velocity in the channel were recorded in real time. The ratio of flow pressure to flow velocity at the same time was used as the viscosity sub-parameter of the vitreous fluid sample at the corresponding time. The higher the flow pressure and the lower the flow velocity, the larger the ratio, and the more viscous the vitreous fluid sample. The mean of the viscosity sub-parameter at all times during the rheological test was used as the viscosity parameter.

[0032] Finally, the explicit parameters and viscosity parameters are multiplied and fused, and the product is mapped to tanh and normalized to obtain the physical interference factor of intraocular fluid samples. This means that the more red, turbid and viscous the intraocular fluid sample is, the greater its impact on PCR detection, thus preparing for subsequent accurate recommendation of cycling parameters based on artificial intelligence algorithms.

[0033] It should be noted that all parameters in the embodiments of the present invention need to undergo standardization processing such as removal of dimensions before calculation.

[0034] In other embodiments of the present invention, the implementer may also take aqueous humor samples and vitreous fluid samples separately, perform analysis and calculation of display parameters and viscosity parameters on both samples, and then average the calculation results of the two samples to obtain the physical interference factor.

[0035] Considering that nucleic acid concentration is a direct indicator of the total amount of nucleic acid template in intraocular fluid, and that intraocular fluid itself is an avascular transparent matrix with very few cells, low nucleic acid concentration is the norm. Too low a concentration will lead to insufficient amplification of the target sequence, and the greater the impact on PCR detection, the more necessary it is to increase the number of cycles. Furthermore, considering that intraocular fluid samples are prone to contamination with proteins (such as hemoglobin) or organic contaminants (such as anesthetics and povidone-iodine) due to puncture damage, these substances can not only reduce PCR efficiency by inhibiting Taq enzyme activity and interfering with primer binding, but also affect nucleic acid purity and reduce the accuracy of subsequent PCR testing. Furthermore, considering that intraocular fluid biomarkers include DNA (such as pathogen genomes and tumor driver genes) and RNA (such as cytokine mRNA and viral RNA), and that their amplification mechanisms differ (RNA requires reverse transcription to cDNA first), if the sample has a high RNA content (i.e., RNA-dominant), failing to add a reverse transcription step will lead to missed detection of RNA biomarkers; conversely, if the sample has a high DNA content (i.e., DNA-dominant), excessive reverse transcription will waste reagents and prolong the detection time; failing to clearly define nucleic acid dominance will reduce the efficiency and effectiveness of subsequent PCR detection. Therefore, in this embodiment of the invention, nucleic acid quality factors are obtained based on the concentration, purity, and dominant characteristics of nucleic acids in intraocular fluid samples. The nucleic acid quality factors further reflect the degree of interference of intraocular fluid samples with PCR detection at the level of nucleic acid template quality, preparing for subsequent evaluation of cycling parameters to improve PCR detection results.

[0036] Preferably, in one embodiment of the present invention, the method for obtaining nucleic acid quality factors includes: For each type of intraocular fluid sample, a subsample of the third preset dose was taken for quantitative detection to obtain the nucleic acid concentration parameter of the corresponding subsample; a subsample of the fourth preset dose was taken, and the nucleic acid purity parameter was obtained based on the protein content and organic matter content of the corresponding subsample; a subsample of the fifth preset dose was taken, and the dominant nucleic acid parameter was obtained based on the RNA concentration and DNA concentration of the corresponding subsample. The nucleic acid concentration parameters, nucleic acid purity parameters, and dominant nucleic acid parameters of each intraocular fluid sample were fused, and the normalized value of the fusion result was used as the nucleic acid quality factor.

[0037] It should be noted that each sampling is an additional sample taken from the remaining intraocular fluid sample after the previous sampling, and contamination should be avoided during the sampling process; implementers may also adjust the sampling dose according to actual needs.

[0038] As an example, separate samples were taken from aqueous humor and vitreous fluid for analysis. This example uses vitreous fluid as an example. The preset third dose was set to 5 μl. Two 5 μl vitreous fluid sub-samples were taken from the vitreous fluid sample for fluorescence quantitative detection. Specifically, the Qubit dsDNA HS and RNA HS kits were used for fluorescence quantitative detection. The sum of the measured RNA concentration and DNA concentration was used as the nucleic acid concentration parameter. The fourth dose was set to 5 μl, and another 5 μl of vitreous fluid sample was extracted. The protein content and organic matter content were detected by Nanodrop ultraviolet spectrophotometry to assess the nucleic acid contamination. In a preferred embodiment of the present invention, it is considered that the absorption characteristics of nucleic acids, proteins and organic pollutants to ultraviolet light at specific wavelengths can help assess the concentration or content of each component; the maximum absorption peak of nucleic acids (DNA / RNA) is at a wavelength of 260 nm, the maximum absorption peak of aromatic amino acids in proteins (such as phenylalanine and tyrosine) is at a wavelength of 280 nm, and organic matter has a strong absorption peak near a wavelength of 230 nm in the ultraviolet region. Therefore, the A260 / A280 ratio can help reflect protein contamination of nucleic acids. A larger A260 / A280 ratio indicates a higher relative nucleic acid content compared to protein content, and thus a relatively higher nucleic acid purity. The A260 / A230 ratio can reflect organic contamination of nucleic acids. A larger A260 / A230 ratio indicates a higher relative nucleic acid content compared to organic matter content, and thus a relatively higher nucleic acid purity. Methods for obtaining nucleic acid purity parameters include: For each intraocular fluid sample, a subsample with a preset fourth dose was subjected to spectrophotometric detection. The A260 / A280 ratio was used as the protein purity parameter, and the A260 / A230 ratio was used as the organic contaminant purity parameter. The protein purity parameter and the organic contaminant purity parameter were fused, and the fusion result was used as the nucleic acid purity parameter.

[0039] Specifically, the protein purity parameter and the organic contaminant purity parameter are multiplied and fused, and the fusion result is used as the nucleic acid purity parameter. It should be noted that the A260 / A280 ratio and the A260 / A230 ratio are already known techniques, and their specific processes will not be elaborated further.

[0040] Furthermore, the preset fifth dose can be set to 2 μl, and two additional 2 μl vitreous fluid samples can be extracted. The RNA and DNA concentrations in the corresponding vitreous fluid samples can be measured to assess the dominant nucleic acid parameters. Specifically, multiplex quantitative PCR (qPCR) experiments can be designed to specifically distinguish and detect DNA and RNA markers in the two vitreous fluid samples. The core principle is to use different enzymes and reaction procedures to achieve targeted amplification and detection of specific nucleic acid molecules. This is a well-known technology, and the specific operation process will not be elaborated further. Only the general conditions and process for specific detection will be briefly described: 1. Reagent preparation: DNA detection system: 5 μL of 2×PCR Mix (containing Taq enzyme), 0.5 μL of DNA-specific primers (such as β-actin DNA primers), 2 μL of vitreous fluid sample, and nuclease-free water to a final volume of 10 μL; RNA detection system: 5 μL of 2×RT-PCR Mix (containing reverse transcriptase), 0.5 μL of RNA-specific primer (such as β-actin RNA primer), 2 μL of vitreous fluid sample, and nuclease-free water to a final volume of 10 μL; 2. Amplification procedure: DNA detection system: Initial denaturation: 95℃ for 3 minutes; 35 cycles (denaturation: 95℃ for 15 seconds, annealing: 60℃ for 30 seconds, extension: 72℃ for 20 seconds); final extension: 72℃ for 5 minutes; RNA detection system: Reverse transcription: 42℃ for 30 minutes; Enzyme inactivation: 95℃ for 5 minutes; 35 cycles (denaturation: 95℃ for 15 seconds, annealing: 60℃ for 30 seconds, extension: 72℃ for 20 seconds); Final extension: 72℃ for 5 minutes; 3. Data Analysis: Ct values ​​were obtained from two vitreous fluid samples using real-time quantitative PCR. The concentrations were then converted using a standard curve. RNA / DNA = RNA concentration / DNA concentration. This value was recorded as the dominant nucleic acid parameter. In another example of the present invention, the ratio of RNA concentration to DNA concentration measured by the Qubit dsDNA HS and RNA HS kits can be directly used as the dominant nucleic acid parameter; the larger the RNA / DNA ratio, the higher the RNA dominance.

[0041] Thus far, vitreous fluid samples have been sampled and analyzed to obtain nucleic acid concentration, purity, and dominant nucleic acid parameters. These three parameters are multiplied and fused to obtain the first nucleic acid quality parameter. The same analysis is performed on aqueous humor samples to obtain the second nucleic acid quality parameter. Finally, the first and second nucleic acid quality parameters are added together, and the sum is normalized by mapping it to tanh to obtain the nucleic acid quality factor.

[0042] Considering that in the detection of intraocular fluid biomarkers, especially when using multiplex PCR technology, the chemical state of the sample matrix may seriously interfere with the efficiency and specificity of the amplification reaction; for example, factors such as proteins, lipids, surfactants or degraded nucleic acids can cause multiple interferences to the PCR system, such as enzyme activity inhibition, primer mismatch, and template instability. Furthermore, considering that vitreous fluid normally exhibits a certain degree of colloidal rheology and diffuses in a spherical manner when dripping; however, under pathological conditions such as inflammation, infection, and tumors, cell rupture and exudation can lead to a significant increase in protein concentration, causing the vitreous fluid to become highly viscous. Due to strong intermolecular forces, the highly viscous components will form continuous fine filaments when dripping, resulting in irregular edges of the dripping area; if it contains excessive liposomes, lysozyme, or surfactant contaminants, irregular reflective spots will be produced during the transfer of intraocular fluid samples. Based on this, embodiments of the present invention will obtain matrix inhibition factors according to the drop morphology and pipetting surface characteristics of intraocular fluid samples; the matrix inhibition factors further reflect the degree of interference of intraocular fluid samples on PCR detection at the level of contaminants, preparing for subsequent evaluation of PCR cycling parameters to improve PCR detection results.

[0043] Preferably, in one embodiment of the present invention, the method for obtaining the matrix inhibitory factor includes: Please see Figure 2The diagram illustrates a flowchart of a method for obtaining a matrix inhibitory factor according to an embodiment of the present invention, specifically including: Step S201: Take a pre-set sixth dose of vitreous fluid subsample from the vitreous fluid sample and perform a drop test to obtain a diffusion image of the corresponding subsample after it has been dropped; determine the sample area in the diffusion image, and obtain the gel parameters of the corresponding subsample based on the boundary ruggedness of the sample area.

[0044] Considering that normal vitreous fluid samples should spread in a spherical shape after dripping, their boundaries are relatively smooth and rounded; conversely, when vitreous fluid contains a lot of protein, resulting in high viscosity, it will exhibit stringing when dripping, and the boundaries of the dripping area will become more rugged due to the presence of fine filaments. Based on this, the gel parameters of vitreous fluid samples can be obtained; the gel parameters reflect the viscosity of vitreous fluid, and indirectly reflect whether there are proteins in it that affect PCR detection.

[0045] In one embodiment of the present invention, the preset sixth dose is set to 5 μl. A 5 μl vitreous fluid subsample is taken from the vitreous fluid sample using a 10 μl pipette for a drop experiment. The vitreous fluid subsample is suspended vertically 1 cm above the glass slide, and the sample is slowly released to collect a diffusion image of the drop onto the glass slide. This is an existing technical method and will not be described in detail here.

[0046] In a preferred embodiment of the present invention, considering that the slope between adjacent boundary pixels reflects the local changes of the boundary, the more discrete the slope, the more severe the abrupt change of the local boundary, the less it conforms to the characteristic of the boundary being spherically rounded, and the more likely there is local stringing, the larger the gel parameter; therefore, the method for obtaining the gel parameter includes: performing edge detection on the diffusion image, taking the area enclosed by the edge as the sample area; and taking the variance of the slope between all adjacent edge pixels in the sample area as the gel parameter.

[0047] It should be noted that Canny edge detection is specifically used. The area enclosed by the closed edge is used as the sample area (the area where the vitreous liquid sample falls). Then, within the sample area, starting from any boundary pixel, the system is traversed in any direction. A coordinate system is constructed with each pixel as the origin, and the slope between each pixel and the next adjacent pixel is calculated. The variance of the slope is used as the gel parameter. The more discrete all the slopes are, the more fluctuations there are in the local changes at the boundary, that is, there are filamentous abrupt changes, which do not conform to the smooth changes of a sphere, and the larger the gel parameter is.

[0048] In another embodiment of the present invention, the viscosity parameters obtained above can be used directly instead of the gel parameters.

[0049] Step S202: Take a subsample of a preset seventh dose from the vitreous fluid sample and perform a pipetting experiment to obtain the liquid surface image of the corresponding subsample during the pipetting process; determine the reflective patches in the liquid surface image based on grayscale features, and obtain the contamination parameters of the corresponding subsample based on the boundary ruggedness features and grayscale values ​​of the reflective patches.

[0050] Considering that it is difficult to observe minute contamination features in vitreous fluid samples when they are completely still, the slow pipetting action causes slight ripples on the surface. This dynamic change instantly generates many tiny specular reflections at various angles, allowing observation of irregular reflective spots formed by contaminants such as lipids or interface inhibitors on the surface. These reflective spots have rugged boundaries and relatively high brightness. Based on this, contamination parameters of the vitreous fluid sample can be obtained. These parameters reflect the degree of contamination of the vitreous fluid by lipids or interface inhibitors, indirectly reflecting the extent of its impact on PCR detection.

[0051] In one embodiment of the present invention, the preset seventh dose is set to 50 μl. A 50 μl vitreous fluid subsample is taken from the vitreous fluid sample using a 200 μl pipette. During the low-speed (<500 μL / min) pipetting process, liquid surface images are acquired in real time. One liquid surface image is randomly selected for analysis to calculate the contamination coefficient. Alternatively, all liquid surface images can be analyzed, and the average contamination coefficient is taken. This is an existing technical method and will not be described in detail here.

[0052] In a preferred embodiment of the present invention, considering that irregular light spots are usually bright, the reflective patches can be determined based on threshold segmentation, and the contamination parameters can be evaluated by combining the boundary ruggedness features. The method for obtaining the contamination parameters includes: obtaining the foreground region in the liquid surface image based on the threshold segmentation algorithm, and taking the foreground region as the reflective patch; taking the variance of the slope between adjacent boundary points in the reflective patch as the boundary ruggedness parameter; and taking the product of the boundary ruggedness parameter and the gray value mean of the pixels in the reflective patch as the contamination parameter.

[0053] It should be noted that the foreground region (with a high gray value) is divided using Otsu threshold segmentation, and the foreground region is used as a reflected light spot (a well-known technique, which will not be described in detail here). Then, the variance of the slope between adjacent boundary points in the reflected spot is calculated to obtain the boundary ruggedness parameter (the slope calculation method is the same as the slope analysis process in step S201 when calculating the gel parameters, which will not be described in detail here).

[0054] Step S203: Combine the colloidal parameters and contamination parameters, and use the normalized value of the fusion result as the matrix inhibition factor.

[0055] Since a larger gel and contamination parameters indicate a greater likelihood of the presence of contaminants such as proteins, lipids, and interface inhibitors in the vitreous fluid sample, which have a greater impact on PCR detection, the gel and contamination parameters are multiplied and fused together. The product is then mapped onto a tanh function for normalization to obtain the matrix inhibition factor.

[0056] Step S3: Input the physical interference factors, nucleic acid quality factors and matrix inhibition factors of the intraocular fluid sample into the pre-trained cyclic parameter recommendation model, and perform rapid detection of biomarkers in the intraocular fluid sample according to the recommended cyclic parameters output by the model.

[0057] Since physical interference factors, nucleic acid quality factors, and matrix inhibition factors reflect the influence of intraocular fluid samples on PCR detection from different perspectives, and neural network models can establish nonlinear mapping relationships between physical interference factors, nucleic acid quality factors, matrix inhibition factors, and optimal PCR cycling parameters, after obtaining the physical interference factors, nucleic acid quality factors, and matrix inhibition factors of intraocular fluid samples, they can be further input into a pre-trained cycling parameter recommendation model, and the model will output recommended cycling parameters for rapid detection.

[0058] In a preferred embodiment of the present invention, the method for obtaining the cyclic parameter recommendation model includes: Physical interference factors, nucleic acid quality factors, and matrix inhibition factors were obtained from different historical intraocular fluid samples. The optimal cycling parameters for each historical intraocular fluid sample were labeled to construct a training sample set. A feedforward multilayer perceptron model was trained based on the training sample set, and the trained model was used as a cycling parameter recommendation model.

[0059] The training process of the feedforward multilayer perceptron model is a well-known technique, and its general process is briefly described here: A large number of historical intraocular fluid samples were obtained, and the physical interference factor, nucleic acid quality factor, and matrix inhibition factor for each historical intraocular fluid sample were calculated based on the method described in step S2 above. Then, the optimal cycling parameters for each historical intraocular fluid sample were labeled based on human experience to obtain sample data for model training. The cycling parameters include at least the following: total number of cycles, annealing temperature, extension time, and initial denaturation time. A simple, high-generalization-capability feedforward multilayer perceptron (MLP) model is used to establish a nonlinear mapping between input sample features and optimal PCR cyclic parameters: the input layer contains three nodes, corresponding to physical interference factors, nucleic acid quality factors, and matrix inhibition factors, respectively; the hidden layers are designed as three layers, containing 32, 16, and 8 nodes, respectively, using the ReLU activation function to increase the model's nonlinear expression capability; the output layer contains four nodes, corresponding to four cyclic parameters; the output uses a linear activation function to adapt to the regression task; and the loss function adopts a multi-objective mean squared error form. The sample dataset used for model training is divided into a training set and a validation set in a 7:3 ratio. After the training set is input into the model for training, the validation set is used for validation analysis to complete the model training and obtain the trained cyclic parameter recommendation model. The physical interference factors, nucleic acid quality factors, and matrix inhibition factors of the intraocular fluid sample to be tested are input into the trained cyclic parameter recommendation model, and the model automatically outputs the cyclic parameters corresponding to the intraocular fluid sample to be tested.

[0060] Finally, the cyclic parameters output by the model are used to perform PCR testing on the intraocular fluid samples to be tested. This is a well-known technique, and the specific testing process will not be described in detail here.

[0061] In summary, this invention collects intraocular fluid samples for testing; obtains physical interference factors based on the color, turbidity, and flow characteristics of the intraocular fluid samples; obtains nucleic acid quality factors based on the concentration, purity, and dominant nucleic acid characteristics of the nucleic acids in the intraocular fluid samples; and obtains matrix inhibition factors based on the drop morphology and pipetting surface characteristics of the intraocular fluid samples. The physical interference factors, nucleic acid quality factors, and matrix inhibition factors of the intraocular fluid samples are input into a pre-trained cyclic parameter recommendation model, and the biomarkers of the intraocular fluid samples are rapidly detected based on the recommended cyclic parameters output by the model. This invention quantifies the sample characteristic factors that reflect the influence of intraocular fluid samples on PCR detection from different perspectives, and then uses a pre-trained model to recommend cyclic parameters for rapid detection, thereby improving the detection efficiency and effectiveness of intraocular fluid biomarkers.

[0062] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0063] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A rapid detection method for intraocular fluid biomarkers based on artificial intelligence, characterized in that, The method includes: Collect intraocular fluid samples to be tested; Based on the color, turbidity, and flow characteristics of intraocular fluid samples, physical interference factors were obtained; based on the concentration, purity, and dominant characteristics of nucleic acids in intraocular fluid samples, nucleic acid quality factors were obtained; and based on the drop morphology and pipetting surface characteristics of intraocular fluid samples, matrix inhibition factors were obtained. The physical interference factor, the nucleic acid quality factor, and the matrix inhibition factor of the intraocular fluid sample are input into a pre-trained cyclic parameter recommendation model, and the biomarkers of the intraocular fluid sample are rapidly detected based on the recommended cyclic parameters output by the model.

2. The rapid detection method for intraocular fluid biomarkers based on artificial intelligence according to claim 1, characterized in that, The intraocular fluid sample includes at least an aqueous humor sample and a vitreous humor sample.

3. The rapid detection method for intraocular fluid biomarkers based on artificial intelligence according to claim 2, characterized in that, The method for obtaining the physical interference factor includes: A preset first dose of aqueous humor subsample is taken from the aqueous humor sample, and the RGB image of the aqueous humor subsample in a static state is acquired. The light transmittance is obtained, and the explicit parameters of the aqueous humor subsample are obtained based on the light transmittance and the R channel value of the pixel in the RGB image. A second dose of vitreous fluid was taken from the vitreous fluid sample and a rheological test was performed. The viscosity parameters of the vitreous fluid sample were obtained based on the flow pressure and flow velocity of the vitreous fluid sample at each moment. The explicit parameters and the viscosity parameters are fused, and the normalized value of the fusion result is used as the physical interference factor.

4. The rapid detection method for intraocular fluid biomarkers based on artificial intelligence according to claim 3, characterized in that, The methods for obtaining the explicit parameters include: In an RGB image, the sum of the R channel values ​​of all pixels is used as the numerator, and the sum of the B and G channel values ​​of all pixels is used as the denominator. The ratio of the fractions is used as the bloodiness index. The negative correlation mapping result of the light transmittance is used as the turbidity index. The bloodiness index and the turbidity index are fused, and the fusion result is used as an explicit parameter.

5. The rapid detection method for intraocular fluid biomarkers based on artificial intelligence according to claim 2, characterized in that, The method for obtaining the nucleic acid quality factor includes: For each type of intraocular fluid sample, a subsample of the third preset dose was taken for quantitative detection to obtain the nucleic acid concentration parameter of the corresponding subsample; a subsample of the fourth preset dose was taken, and the nucleic acid purity parameter was obtained based on the protein content and organic matter content of the corresponding subsample; a subsample of the fifth preset dose was taken, and the dominant nucleic acid parameter was obtained based on the RNA concentration and DNA concentration of the corresponding subsample. The nucleic acid concentration parameter, nucleic acid purity parameter, and nucleic acid dominant parameter of each intraocular fluid sample are fused, and the normalized value of the fusion result is used as the nucleic acid quality factor.

6. The rapid detection method for intraocular fluid biomarkers based on artificial intelligence according to claim 5, characterized in that, The method for obtaining the nucleic acid purity parameter includes: For each intraocular fluid sample, a subsample of a preset fourth dose was subjected to spectrophotometric detection. The A260 / A280 ratio was used as a protein purity parameter, and the A260 / A230 ratio was used as an organic contaminant purity parameter. The protein purity parameter and the organic contaminant purity parameter were fused together, and the fusion result was used as the nucleic acid purity parameter.

7. The rapid detection method for intraocular fluid biomarkers based on artificial intelligence according to claim 2, characterized in that, The method for obtaining the matrix inhibitory factor includes: A sixth dose of vitreous fluid was taken from a vitreous fluid sample and a drop test was performed to obtain a diffusion image of the corresponding vitreous fluid sub-sample after it was dropped. The sample area in the diffusion image was determined, and the gel parameters of the corresponding vitreous fluid sub-sample were obtained based on the boundary ruggedness of the sample area. A pre-set seventh dose of vitreous fluid subsample was taken from the vitreous fluid sample and a pipetting experiment was performed to obtain the liquid surface image of the corresponding vitreous fluid subsample during the pipetting process; the reflective patches in the liquid surface image were determined based on grayscale features, and the contamination parameters of the corresponding vitreous fluid subsample were obtained based on the boundary ruggedness features and grayscale values ​​of the reflective patches. The gel parameters and the contamination parameters are fused, and the normalized value of the fusion result is used as the matrix inhibition factor.

8. The rapid detection method for intraocular fluid biomarkers based on artificial intelligence according to claim 7, characterized in that, The method for obtaining the gel parameters includes: Edge detection is performed on the diffused image, and the area enclosed by the edges is taken as the sample area; the variance of the slope between all adjacent edge pixels in the sample area is taken as the gel parameter.

9. The rapid detection method for intraocular fluid biomarkers based on artificial intelligence according to claim 7, characterized in that, The methods for obtaining the pollution parameters include: The foreground region in the liquid surface image is obtained based on a threshold segmentation algorithm, and the foreground region is taken as a reflective patch. The variance of the slope between adjacent boundary points in the reflective patch is taken as a boundary ruggedness parameter. The product of the boundary ruggedness parameter and the gray-scale mean of the pixels in the reflective patch is taken as a contamination parameter.

10. The rapid detection method for intraocular fluid biomarkers based on artificial intelligence according to claim 1, characterized in that, The method for obtaining the cyclic parameter recommendation model includes: The physical interference factor, nucleic acid quality factor, and matrix inhibition factor of different historical intraocular fluid samples are obtained, and the optimal cycling parameters are labeled for each historical intraocular fluid sample to construct a training sample set; a feedforward multilayer perceptron model is trained based on the training sample set, and the trained model is used as a cycling parameter recommendation model.

Images