Wound microenvironment adaptive recognition method and system based on raman spectrum

Abstract

The application provides a wound microenvironment adaptive recognition method and system based on Raman spectrum, comprising: acquiring spatial partition information of a wound center area, a transition area, an edge area and a periwound tissue area; collecting spontaneous Raman spectrum in each area; performing multi-dimensional quality evaluation on each spectrum; performing adaptive parameter selection operation according to a quality state index, fitting a target pretreatment parameter combination meeting a preset quality requirement for each spectrum from a candidate parameter library and performing pretreatment; extracting a candidate Raman feature; calculating a position mean value of each area and a spatial gradient index between areas; and outputting the position mean value and / or the spatial gradient index. The application solves the spectrum distortion problem caused by fixed parameters through adaptive pretreatment driven by each spectrum quality, and realizes quantitative description of directional changes of the wound microenvironment from the center to the periphery through the spatial gradient index.

Description

Technical Field

[0001] This invention relates to the fields of biomedical spectral detection, medical data processing, and wound assessment, and particularly to an adaptive identification method and system for wound microenvironment based on Raman spectroscopy. Background Technology

[0002] Diabetic foot ulcers are one of the most serious chronic complications of diabetes. Their healing process involves the dynamic interaction of multiple biological events, including inflammatory response, oxidative stress, extracellular matrix remodeling, abnormal lipid metabolism, and local blood perfusion status. In this complex pathophysiological process, significant spatial heterogeneity exists in the wound center, the transition zone from the center to the wound edge, the wound edge, and the surrounding tissues in terms of exudate composition, blood infiltration, residual necrotic tissue, granulation tissue growth, and new epithelial formation. Accurately obtaining information on the tissue molecular composition and spatial distribution of these different regions is crucial for objectively determining the healing stage of the wound, developing individualized clinical intervention strategies, and evaluating treatment efficacy.

[0003] Currently, the main methods for clinically assessing diabetic foot wounds include gross observation, area measurement, histological biopsy, and biochemical testing. While gross observation and area measurement are simple and non-invasive, they primarily reflect macroscopic morphological changes in the wound and cannot provide information at the tissue molecular composition level, nor can they reflect the molecular differences between different regions within the wound. Although histological biopsy and biochemical testing can provide relatively accurate cellular and molecular information, they are both invasive procedures requiring tissue sampling from the wound. This not only causes pain for the patient and increases the risk of infection and delayed healing, but also makes it difficult to perform repeated measurements at multiple time points in a fixed location on the same wound, limiting their application in dynamic follow-up and longitudinal studies.

[0004] Spontaneous Raman spectroscopy, a label-free and non-destructive molecular vibrational spectroscopy technique, can record the vibrational fingerprints of various biomolecules such as proteins, lipids, and nucleic acids within tissues without the addition of external labels. In recent years, it has been increasingly applied to in-situ or in vitro analysis of wound tissues. However, the practical application of Raman spectroscopy in the clinical assessment of diabetic foot wounds still faces several technical obstacles. First, the Raman spectral quality of different regions of the wound is highly unstable—the central region often exhibits a strong fluorescence background due to abundant exudate and blood, while the spectral signal-to-noise ratio of the wound edges and surrounding tissues is relatively better. Using the same set of preprocessing parameters for all spectra inevitably leads to under-correction in some spectra and over-correction in others. Second, subjectively specifying different preprocessing intensities based solely on sampling location may erroneously attribute differences caused by algorithm parameters to actual biological differences, resulting in a loss of objectivity and verifiability. Third, existing studies mostly focus on comparing single-point peak intensities or classifying healing stages based on overall average spectra, lacking comparable indicators that can quantitatively describe directional changes from the wound center to surrounding tissues, the degree of heterogeneity within the same region, and the rate of spatial pattern evolution during follow-up. Furthermore, directly using the absolute intensity of Raman characteristic peaks to represent the concentration of a component ignores the multiplicative effects of tissue optical properties, instrument conditions, and blood interference on signal intensity, potentially leading to misjudgments of the wound microenvironment. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide an adaptive identification method and system for wound microenvironment based on Raman spectroscopy, in order to solve the problems of existing technologies, such as distortion of spectral processing of different quality caused by fixed preprocessing parameters, deviation introduced by subjective setting of processing intensity according to region, lack of spatial directionality quantification index from the center to the periphery, and lack of calibration basis for the direct equivalent of single-point peak intensity to component concentration.

[0006] On the one hand, the present invention provides an adaptive identification method for the wound microenvironment based on Raman spectroscopy, the method comprising: The spatial partitioning information of the target wound is obtained. The spatial partitioning information is used to indicate the region category of multiple preset location areas on the target wound. The multiple preset location areas include the wound center area, the transition area from the center to the wound edge, the wound edge area, and the peri-wound tissue area. Based on the spatial partitioning information, at least one spontaneous Raman spectrum is collected in each preset location region, and the acquisition parameters and location attribution information corresponding to each spontaneous Raman spectrum are recorded. For each spontaneous Raman spectrum, a quality evaluation operation is performed to obtain the quality status index of the spontaneous Raman spectrum in multiple quality dimensions. Based on the quality status index of each spontaneous Raman spectrum, an adaptive parameter selection operation is performed to adapt a target preprocessing parameter combination for the spontaneous Raman spectrum from a pre-constructed candidate parameter library. The adaptive parameter selection operation includes: preprocessing the spontaneous Raman spectrum using multiple sets of candidate preprocessing parameter combinations from the candidate parameter library, evaluating the processing effect of each set of candidate preprocessing parameter combinations, and determining the set of candidate preprocessing parameter combinations whose processing effect meets the preset quality requirements as the target preprocessing parameter combination for the spontaneous Raman spectrum. The corresponding spontaneous Raman spectra are preprocessed using the adapted target preprocessing parameter combination; Candidate Raman features were extracted from the preprocessed individual Raman spectra, and the candidate Raman features included feature information of molecular vibrational spectral bands related to the wound microenvironment. Based on the candidate Raman features of all valid sampling points within each preset location area, calculate the location mean of each preset location area; Based on the average position of each preset location region, calculate at least one set of spatial gradient indices between preset location region combinations; wherein, the spatial gradient indices include gradient difference and / or gradient ratio, and the gradient difference and gradient ratio are used to characterize the difference in the change of feature values ​​of each effective sampling point in the same direction; Output the average location value and / or spatial gradient index of each preset location area.

[0007] In some embodiments of the present invention, the central region of the wound is the geometric center point or central area of ​​the wound; the transition region from the center to the wound edge is the midpoint or layer transition position between the central region of the wound and the edge region of the wound in the same direction; the edge region of the wound is the boundary of the wound and its adjacent area; and the peri-wound tissue region is the adjacent tissue region outside the boundary of the wound.

[0008] In some embodiments of the present invention, when acquiring the spontaneous Raman spectrum, a cross-shaped radial sampling template with the central region of the wound as the intersection point is used. The cross-shaped radial sampling template has four axes with the central region of the wound as the intersection point, forming at least eight radial directions. Sampling points for the transition region, the edge region of the wound, and the peri-wound tissue region are respectively set in each of the radial directions.

[0009] In some embodiments of the present invention, the quality status indicators include signal-to-noise ratio, fluorescence background intensity, baseline slope, saturated pixel ratio, cosmic ray anomaly, characteristic peak integrity, peak position shift, hemoglobin reference spectrum matching degree, repeat spectrum consistency, and instrument drift. The quality threshold corresponding to each quality status indicator is predetermined by the statistical distribution of quality control samples, qualified spectra in pre-experiments, quantiles of training samples, or histological and biochemical calibration results.

[0010] In some embodiments of the present invention, the adaptive parameter selection operation includes a combination of candidate preprocessing parameters, a smoothing window parameter, and a multi-peak fitting constraint parameter; the preset quality requirements include a baseline residual condition, a post-processing signal-to-noise ratio condition, a condition for maintaining the position and area of ​​the target peak, and a condition for consistent repeating spectra.

[0011] In some embodiments of the present invention, during the adaptive parameter selection operation: For spectra with fluorescence background intensity exceeding a preset threshold, the candidate preprocessing parameter combination with the smallest baseline residual is selected from the combinations of candidate preprocessing parameters that satisfy the baseline residual condition, the signal-to-noise ratio condition after processing, the condition for maintaining the peak position and peak area of ​​the target peak, and the condition for consistency of repeated spectra. For spectra with a signal-to-noise ratio lower than a preset threshold, re-acquisition is performed first; if the signal-to-noise ratio is still lower than the preset threshold after re-acquisition, a smoothing window parameter is selected, provided that the peak position and peak area of ​​the target peak are maintained to the extent that the preset quality requirements are met. For spectra with peak overlap exceeding a preset threshold, multi-peak fitting with constraints on peak position, peak width, and non-negative peak area is invoked.

[0012] In some embodiments of the present invention, the method further includes: Based on the candidate Raman features of each valid sampling point within each preset location area, a spatial heterogeneity index for each preset location area is calculated. The spatial heterogeneity index includes the coefficient of variation and / or the range. The coefficient of variation is used to characterize the dispersion of the feature values ​​of each valid sampling point relative to the mean of the area, and the range is used to characterize the maximum fluctuation range of the feature values ​​of each valid sampling point.

[0013] In some embodiments of the present invention, when data from at least two data collection points exist for the target wound, the method further includes: Calculate the rate of change of the same spatial gradient index between adjacent acquisition time points. The rate of change is used to characterize how fast and in what direction the spatial gradient index changes over time.

[0014] In some embodiments of the present invention, the candidate Raman features include a wavenumber of 498 cm⁻¹. -1 1094cm -1 1250cm -1 1307cm -1 1064cm -1 1000cm -1 and 1122cm -1 The candidate Raman features are represented by peak height, peak area, fitted peak area, spectral decoupling coefficient, or model output calibrated by histological and biochemical indicators.

[0015] On the other hand, the present invention provides an adaptive identification system for wound microenvironment based on Raman spectroscopy, including a processor, a memory, and a computer program or instructions stored in the memory. The processor is used to execute the computer program or instructions, and when the computer program or instructions are executed, the system implements the steps of any of the methods mentioned above.

[0016] This invention establishes a spatial partitioning system for the wound center, transition zone, edge zone, and peri-wound tissue zone, and combines it with multi-directional radial sampling. This upgrades wound assessment from single-point readings to a distribution analysis covering the entire spatial axis from the center to the peri-wound, enabling the spatial heterogeneity of the wound microenvironment to be systematically captured and quantified.

[0017] Furthermore, this invention achieves independent adaptation of preprocessing parameters for each spectrum through multi-dimensional quality evaluation of each spectrum and an adaptive parameter selection mechanism based on quality status indicators. This transforms the decision-making basis for preprocessing parameters from the operator's subjective experience to the objective quality data of the spectrum itself. This ensures that all types of spectra receive processing suitable for their own quality status and avoids algorithmic biases introduced by manually setting different processing intensities due to different sampling locations. The entire data processing process is traceable and verifiable.

[0018] Furthermore, this invention utilizes a quantitative system of location mean and inter-regional gradient difference and gradient ratio to transform the mean characteristics of each region into comparative results with spatial directional significance, thereby achieving a numerical description of the direction and magnitude of changes in molecular components of the wound microenvironment from the center to the periphery. The coefficient of variation and range quantitatively describe the dispersion of each sampling point within the same region, and the temporal rate of change quantifies the direction and speed of spatial pattern evolution as the healing process progresses, forming a complete characterization system from intra-regional uniformity to inter-regional variation trends, and from static spatial distribution to dynamic evolution tracking.

[0019] Furthermore, this invention employs spontaneous Raman spectroscopy, eliminating the need for external markers and wound sampling. It allows for repeated measurements at multiple time points on the same wound site, avoiding the pain and infection risks associated with invasive procedures. Candidate Raman characteristics are represented by peak height, peak area, fitted peak area, spectral decoupling coefficient, or calibrated model output, rather than directly representing concentration with wavenumber alone. This avoids misjudgments caused by neglecting the influence of tissue optical properties and instrument status on signal intensity. This invention can be deployed as an integrated acquisition-analysis system or integrated as a standalone software module into existing Raman equipment, exhibiting good compatibility and scalability.

[0020] Additional advantages, objects, and features of the invention will be set forth in part in the description which follows, and will also become apparent in part to those skilled in the art upon studying the description, or may be learned by practice of the invention. The objects and other advantages of the invention can be realized and obtained by means of the structures specifically pointed out in the specification and drawings.

[0021] Those skilled in the art will understand that the objectives and advantages achievable with this invention are not limited to those specifically described above, and that the above and other objectives achievable with this invention will become clearer from the following detailed description. Attached Figure Description

[0022] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, are not intended to limit the scope of the invention. In the drawings: Figure 1 This is a schematic diagram illustrating the steps of an adaptive identification method for wound microenvironment based on Raman spectroscopy in one embodiment of the present invention.

[0023] Figure 2 This is a flowchart illustrating an adaptive identification method for wound microenvironment based on Raman spectroscopy in one embodiment of the present invention.

[0024] Figure 3 This is a schematic diagram of a cross-shaped spatial sampling template in one embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this invention are used to explain the invention, but are not intended to limit the invention.

[0026] It should also be noted that, in order to avoid obscuring the invention with unnecessary details, only the structures and / or processing steps closely related to the solution according to the invention are shown in the accompanying drawings, while other details that are not closely related to the invention are omitted.

[0027] It should be emphasized that the term "including / comprises" as used herein refers to the presence of a feature, element, step, or component, but does not exclude the presence or addition of one or more other features, elements, steps, or components.

[0028] It should also be noted that, unless otherwise specified, the term "connection" in this article can refer not only to a direct connection, but also to an indirect connection involving an intermediary.

[0029] In the following description, embodiments of the invention will be illustrated with reference to the accompanying drawings. In the drawings, the same reference numerals represent the same or similar parts, or the same or similar steps.

[0030] It should be emphasized here that the step markers mentioned below are not a limitation on the order of the steps, but should be understood as meaning that the steps can be executed in the order mentioned in the embodiments, or in a different order than in the embodiments, or several steps can be executed simultaneously.

[0031] To address the problems in existing technologies, such as distortion in spectral processing due to fixed preprocessing parameters leading to varying quality, human bias introduced by subjectively setting processing intensity by region, lack of spatial directionality quantification indicators from the center to the periphery, and the lack of calibration basis for directly equating single-point peak intensity with component concentration, this invention provides an adaptive identification method for the wound microenvironment based on Raman spectroscopy. Figure 1 As shown, the method includes the following steps S101~S109: Step S101: Obtain spatial zoning information of the target wound. The spatial zoning information is used to indicate the region categories of multiple preset location areas on the target wound, including the wound center area, the transition area from the center to the wound edge, the wound edge area, and the perianal tissue area.

[0032] Step S102: Based on the spatial partitioning information, collect at least one spontaneous Raman spectrum in each preset location area, and record the acquisition parameters and location information corresponding to each spontaneous Raman spectrum.

[0033] Step S103: For each spontaneous Raman spectrum, perform a quality evaluation operation to obtain the quality status index of the spontaneous Raman spectrum in multiple quality dimensions.

[0034] Step S104: Based on the quality status index of each spontaneous Raman spectrum, perform an adaptive parameter selection operation to adapt a target preprocessing parameter combination for the spontaneous Raman spectrum from a pre-constructed candidate parameter library. The adaptive parameter selection operation includes: preprocessing the spontaneous Raman spectrum using multiple sets of candidate preprocessing parameter combinations from the candidate parameter library, evaluating the processing effect of each set of candidate preprocessing parameter combinations, and determining the set of candidate preprocessing parameter combinations whose processing effect meets the preset quality requirements as the target preprocessing parameter combination for the spontaneous Raman spectrum.

[0035] Step S105: Preprocess the corresponding spontaneous Raman spectrum using the adapted target preprocessing parameter combination.

[0036] Step S106: Extract candidate Raman features from the preprocessed individual Raman spectra. The candidate Raman features include characteristic information of molecular vibrational spectral bands related to the wound microenvironment.

[0037] Step S107: Calculate the position mean of each preset location region based on the candidate Raman features of all valid sampling points within each preset location region.

[0038] Step S108: Calculate the spatial gradient index between at least one set of preset location region combinations based on the average position value of each preset location region. The spatial gradient index includes gradient difference and / or gradient ratio, which characterize the differences in feature values ​​of each valid sampling point along the same direction.

[0039] Step S109: Output the average position value and / or spatial gradient index of each preset location area.

[0040] Figure 2 The overall flow diagram of the method is further illustrated.

[0041] In step S101, spatial partitioning information of the target wound is obtained. This spatial partitioning information is used to indicate the region categories of multiple preset location areas on the target wound. These preset location areas include the wound center area, the transition area from the center to the wound edge, the wound edge area, and the perianal tissue area. Specifically, the wound center area is the geometric center point or central region of the wound; the transition area from the center to the wound edge is the midpoint or layered transition point between the wound center area and the wound edge area in the same direction; the wound edge area is the wound boundary and its adjacent area; and the perianal tissue area is the adjacent tissue area outside the wound boundary. The wound center M, the transition area from the center to the wound edge MP, the wound edge P, and the perianal tissue S are merely internal location markers set for recording and calculation convenience in this application and are not common partitioning names used in academia or industry.

[0042] In practice, the wound image or real-time view is first acquired, then the wound boundary and center are determined, and normalized coordinates are established. The operator can visually locate the wound based on its appearance and record the points in the image or real-time view, or use manual marking, semi-automatic contour recognition, or automatic image segmentation to determine the wound boundary. The wound center is determined by the geometric center of the boundary, the center of the largest inscribed circle, or the center confirmed by the operator. The transition point is determined by the midpoint between the center and the wound edge point in the same direction, or by the layered transition point. The perigrained tissue zone is determined by extending a preset distance radially outward from the wound edge point. When using visual point selection, coordinates must be marked on the concurrent image, and the radial direction and distance from the boundary must be recorded. Standardized transition points are then generated using geometric rules.

[0043] Step S101 transforms the originally planar wound surface into a structured spatial labeling system. Each subsequent sampling point will carry the category of its region and spatial coordinates, providing data dimensions for the subsequent calculation of the location mean and spatial gradient indicators.

[0044] In step S102, at least one spontaneous Raman spectrum is collected in each preset location area according to the spatial partitioning information, and the acquisition parameters and location information corresponding to each spontaneous Raman spectrum are recorded.

[0045] In some embodiments, equipment quality control is performed before data acquisition. Wavenumber calibration, resolution verification, system signal-to-noise ratio measurement, and drift checks are performed using silicon wafers, dimethyl sulfoxide, or polystyrene standards before or between batches. If a preset threshold is exceeded, recalibration, refocusing, or adjustment of acquisition parameters is performed before reacquiring. During acquisition, the Raman spectroscopy probe or microscope objective is aimed at each preset sampling point on the target wound surface. At least two spectra are acquired for each sampling point, and the region category, spatial coordinates, acquisition time, excitation wavelength, tissue surface power or power density, spot diameter, working distance, integration time, exposure parameters, spectral range, spectral resolution, and number of repetitions are recorded.

[0046] In some embodiments, a star-shaped radial sampling template with the central region of the wound as the intersection point is used when acquiring spontaneous Raman spectra, such as... Figure 3 As shown, the star-shaped radial sampling template has four axes intersecting at the center of the wound, forming at least eight radial directions. Sampling points for the transition zone, wound edge zone, and perianal tissue zone are set in each radial direction. The direction of increase or decrease, and the stratification points, are adjusted according to the wound area, perimeter, boundary complexity, spot diameter, and allowable acquisition time. When the wound area is large, the boundary is irregular, or there are significant differences within the region, the number of radial directions is increased, or transition stratification points are added between the center and the edge.

[0047] Step S102 captures the spatial heterogeneity of the wound in different directions through multi-directional and multi-level system sampling, avoiding sampling bias caused by the asymmetry of the wound shape due to single-directional sampling. At the same time, complete parameter recording makes the acquisition conditions of each spectrum completely traceable.

[0048] In step S103, a quality assessment operation is performed on each spontaneous Raman spectrum to obtain quality status indicators of the spontaneous Raman spectrum in multiple quality dimensions. The purpose of this step is to quantitatively evaluate the signal quality of each spectrum and form a quality profile of the spectrum. Due to the significant differences in tissue state in different regions of the wound, different spectra show significant differences in performance in different dimensions such as signal-to-noise ratio, fluorescence background intensity, and peak overlap. Multi-dimensional quality assessment can comprehensively quantify the quality status of each spectrum, providing a data-driven decision-making basis for subsequent adaptive parameter selection.

[0049] In some embodiments, the following quality status indicators are calculated for each spectrum: signal-to-noise ratio, fluorescence background intensity, baseline slope, saturated pixel ratio, cosmic ray anomaly, characteristic peak integrity, peak position shift, hemoglobin reference spectrum matching degree, repeatability of spectra, and instrument drift. Unqualified spectra are preferentially reacquired; if still unqualified, they are discarded and the reason is recorded. The quality thresholds corresponding to each quality status indicator are predetermined by the statistical distribution of quality control samples, qualified spectra from pre-experiments, quantiles of training samples, or histological and biochemical calibration results, and are solidified before formal analysis.

[0050] In step S103, each spectrum obtains a quality status index vector, which contains the quantitative score of the spectrum in each quality dimension, providing objective data input for adaptive parameter selection rather than the subjective judgment of the operator.

[0051] In step S104, an adaptive parameter selection operation is performed based on the quality status index. This step is one of the core innovative steps of this invention, aiming to enable each spectrum to be adapted to its own exclusive preprocessing parameters according to its own quality status, thereby solving the problems of partial spectral distortion caused by fixed parameters and human bias introduced by subjectively setting parameters according to regions. The principle is that the decision-making basis for the preprocessing parameters is the quality status index of the spectrum itself, rather than the sampling location.

[0052] In the specific implementation, a candidate parameter library is pre-constructed, containing multiple sets of candidate preprocessing parameter combinations that have been experimentally verified. For each spectrum that has passed quality evaluation, the spectrum is preprocessed using multiple sets of candidate preprocessing parameter combinations from the candidate parameter library, and the processing effect of each set of parameters is calculated. The processing effect is compared with the preset quality requirements item by item, and the set of parameters that simultaneously meets all preset quality requirements is selected as the target preprocessing parameter combination.

[0053] In some embodiments, the candidate preprocessing parameter combination includes baseline correction parameters, smoothing window parameters, and multi-peak fitting constraint parameters. Preset quality requirements include baseline residual conditions, post-processed signal-to-noise ratio conditions, target peak position and area preservation conditions, and repeatability spectrum consistency conditions.

[0054] In some embodiments, for spectra with fluorescence background intensity exceeding a preset threshold, the candidate preprocessing parameter combination with the smallest baseline residual is selected from those satisfying the baseline residual condition, post-processed signal-to-noise ratio condition, target peak position and area preservation condition, and repeatability spectrum consistency condition. For spectra with signal-to-noise ratio below a preset threshold, re-acquisition is prioritized. If the re-acquisition result is still below the preset threshold, a smoothing window parameter is selected provided that the target peak position and area preservation meets the preset quality requirements. For spectra with peak overlap exceeding a preset threshold, multi-peak fitting with peak position, peak width, and non-negative peak area constraints is invoked.

[0055] Different spectra at the same location can use different parameters due to differences in quality, while spectra at different locations with similar quality can use the same parameters. If all candidate parameter combinations fail to meet the preset quality requirements, the spectrum is marked as unusable and eliminated, and all information, including its quality index, candidate parameters and evaluation indicators, and the reason for elimination, is recorded.

[0056] In step S104, the allocation of preprocessing parameters is determined entirely by the quality state of the spectrum itself rather than by the location label. This ensures that each spectrum can receive processing that is suitable for its own quality, while avoiding human bias caused by prior location. The entire processing process is traceable and verifiable.

[0057] In step S105, the spectrum is preprocessed using a combination of target preprocessing parameters. The purpose of this step is to apply the preprocessing parameters selected in step S104 to the spectral data, remove various interfering factors, and extract effective spectral signals that can be used for subsequent feature extraction and quantitative analysis.

[0058] In practice, preprocessing includes at least one of the following: instrument background subtraction, environmental background subtraction, cosmic ray removal, band clipping, baseline correction, smoothing and noise reduction, peak position calibration, peak extraction, overlapping peak separation, spectral line decoupling, and normalization. The specific operations performed and the parameters for each operation are determined based on the target preprocessing parameter combination adapted to the target spectrum in step S104. For high fluorescence spectroscopy, the baseline correction model and parameters are compared; for low signal-to-noise ratio spectroscopy, smoothing parameters are compared while ensuring the target peaks remain within acceptable limits. When peaks overlap, multi-peak fitting with constraints on peak position, peak width, and non-negative peak area is used.

[0059] Through the preprocessing operation in step S105, various interfering factors in each spectrum are subtracted or corrected one by one, and finally high-quality spectral data that can be used for subsequent feature extraction and quantitative analysis are obtained.

[0060] In step S106, candidate Raman features are extracted from the preprocessed spectrum. The purpose of this step is to transform the preprocessed spectral curve into a numerical feature vector, so that subsequent spatial quantitative analysis can be performed at the numerical level, rather than relying solely on visual observation of the spectral curve.

[0061] In some embodiments, candidate Raman features are extracted from the preprocessed spectrum. Candidate Raman features include those with a wavenumber of 498 cm⁻¹. -1 1094cm -1 1250cm -1 1307cm -1 1064cm -1 1000cm -1 and 1122cm -1Features of the location. 498cm -1 Attributable to the SS stretching vibration of protein disulfide bonds, it is associated with structural changes in sulfur-containing proteins and serves as a candidate signal for substances related to oxidative stress. (1094 cm⁻¹) -1 Belongs to PO2 - Symmetric stretching vibrations, associated with the nucleic acid phosphate backbone, and possibly contributed by phospholipids in tissues, serve as a composite feature related to the inflammatory microenvironment. (1250cm) -1 This band belongs to the amide III band, reflects protein secondary structure, and may include collagen contributions; it is a collagen I-related band. (1307 cm⁻¹) -1 Attributable to CH2 torsional or oscillatory vibrations, commonly found in lipids, and can be contributed by superimposed proteins in complex tissues, representing a collagen III-related spectral band. 1064 cm⁻¹ -1 Attributable to lipid chain CC skeletal stretching vibrations, this may overlap with other tissue components and serve as a candidate feature band for lipid-related applications. 1000cm -1 This is attributed to the phenylalanine aromatic ring respiratory vibration and is considered a candidate feature for proteins or phenylalanine-related characteristics. (1122cm) -1 CC stretching vibrations attributable to lipid, protein, or glycosyl skeletons are referred to in this application as signals related to blood components. If used to evaluate blood interference, they should be verified separately using hemoglobin reference spectra or histological and biochemical tests. Each wavenumber is only used to identify candidate spectral segments, and the correspondence between each feature and biological indicator is calibrated by histological, biochemical tests, or standard samples.

[0062] In some embodiments, candidate Raman features are represented as peak height, peak area, fitted peak area, spectral decoupling coefficient, or model output calibrated with histological and biochemical indicators.

[0063] In some embodiments, after extracting candidate Raman features, at 1447 cm⁻¹ -1 Using the characteristic peak at 1447 cm⁻¹ as a unified reference characteristic peak, the Raman characteristics of each candidate were calculated and compared with the characteristic peak at 1447 cm⁻¹. -1 The relative ratio of characteristic peaks is used to make relative comparisons between characteristics.

[0064] In step S106, each spectrum is converted into a feature vector, and each feature vector dimension represents the relative abundance of a molecular category, providing a numerical data basis for subsequent spatial quantitative analysis.

[0065] In step S107, the location mean of each region is calculated based on the candidate Raman features of the valid sampling points. The purpose of this step is to aggregate the feature values ​​of each sampling point within the same region to obtain a quantized value representing the average level of the region, thus eliminating random fluctuations between sampling points.

[0066] In some embodiments, for each preset location region, the feature values ​​of a candidate Raman feature in the spectral data of all valid sampling points within that region are collected, and the location mean is calculated. The location mean is calculated using formula (1): ; (1) in, It indicates any location in the central area of ​​the wound, the transition area from the center to the edge of the wound, the edge area of ​​the wound, or the peri-wound tissue area; Indicates position The Middle A Raman candidate feature value for a valid sampling point; Indicates position The number of valid sampling points; Indicates position The mean value of the position of the candidate Raman feature is used to represent the average spectral feature level at that position.

[0067] In step S107, each region obtains a quantitative value representing the overall level of that region, providing a basis for the subsequent calculation of spatial gradient indicators.

[0068] In step S108, the spatial gradient index is calculated based on the location mean. The purpose of this step is to convert the single-point mean of each region into spatial directional information, and to quantitatively describe the direction and magnitude of changes in molecular composition from the wound center to the surrounding tissue in a numerical manner.

[0069] In some embodiments, the spatial gradient difference and spatial gradient ratio are calculated for each candidate Raman feature. The spatial gradient difference represents the difference in the absolute amount of feature content between two locations, and the calculation formula is shown in formula (2): ; (2) in, and It refers to two different locations within the wound center area, the transition area from the center to the wound edge, the wound edge area, or the peri-wound tissue area; Indicates position The mean position of the middle Raman feature; Indicates position The mean position of the middle Raman feature; Indicates position relative position The gradient difference, greater than 0 indicates The content of characteristic components at the location is higher than Position, less than 0 indicates The content of characteristic components at the location is lower than Location, this metric represents the absolute difference in features between two locations, and is not equivalent to the gradient slope after dividing by the physical distance.

[0070] In some embodiments, the spatial gradient ratio represents the relative multiple relationship of feature content between two locations, and the calculation formula is shown in formula (3): ; (3) in, This represents the spatial gradient ratio; a value greater than 1 indicates... The feature content of the location is higher than Position, equal to 1 indicates that two positions are close, less than 1 indicates that... The feature content of the location is lower than Location. Gradient ratio only applies to locations. The calculation is performed when the eigenvalues ​​are not zero and the eigenvalues ​​are calculated using a positive scale that allows for multiple comparisons.

[0071] Step S108 gives the originally isolated mean data of each region a comparative meaning in the spatial dimension, so that the center-periphery directional change of the wound microenvironment can be quantitatively described mathematically.

[0072] In some embodiments, the location combinations include combinations between the wound center area and the surrounding tissue area, between the wound center area and the wound edge area, between the wound center area and the transition area, between the transition area and the wound edge area, and between the wound edge area and the surrounding tissue area.

[0073] In step S109, the location mean and / or spatial gradient index are output. The purpose of this step is to provide the calculated quantitative analysis results to clinicians or researchers in a readable and usable form.

[0074] In some embodiments, the output formats include tables, bar charts, spatial heatmaps of raw or normalized features, regional trend maps, gradient spatial maps, heterogeneity maps, time-varying rates of change, and comprehensive quantitative reports. The output also includes the normalization method used, parameter versions, number of valid points, and quality control information. The quality control information includes all processing information such as the quality status indicators for each spectrum, preprocessing parameter combinations, number of valid points, and reasons for exclusion, ensuring complete traceability of the entire analysis process.

[0075] Step S109 realizes a complete closed loop from raw spectral acquisition to clinical application output. The output quantitative indicators can be directly used for horizontal comparison between different wounds, longitudinal tracking of the same wound, and efficacy evaluation of different intervention measures.

[0076] In some embodiments, the method of the present invention further includes calculating a spatial heterogeneity index for each preset location region based on the candidate Raman features of each effective sampling point within each preset location region. The spatial heterogeneity index includes the coefficient of variation and / or the range. The coefficient of variation is used to characterize the degree of dispersion of the feature values ​​of each effective sampling point relative to the mean of the region, and the calculation formula is shown in formula (4): ; (4) in, Indicates position The standard deviation of the samples at each valid sampling point within the range; Indicates position Raman characteristic mean; Position expressed as a percentage Internal relative dispersion.

[0077] In some embodiments, the range is used to characterize the maximum fluctuation range of the feature values ​​of each effective sampling point, and the calculation formula is shown in formula (5): ; (5) in, and Positions The maximum and minimum values ​​of the internal eigenvalues. The range is susceptible to individual extreme values ​​and should be reported together with the number of valid points, the original distribution, and the quality control results. This method is only used for indicators with a mean greater than 0, a true zero point, and suitable for proportional comparisons. Indicators with a mean close to 0, 0, or negative are not included in the calculation. Instead, sample standard deviation or range should be used to describe heterogeneity.

[0078] In some embodiments, when data from at least two acquisition time points exist for the target wound, the method of the present invention further includes calculating the rate of change of the same spatial gradient index between adjacent acquisition time points. This rate of change is used to characterize the speed and direction of change of the spatial gradient index over time. The formula for calculating the rate of change is shown in formula (6): ; (6) in, The selected spatial gradient index can be either the gradient difference or the gradient ratio. , For two collection times and A positive result indicates that the indicator increases over time, while a negative result indicates that the indicator decreases over time. When time is measured in days, the corresponding indicator is calculated per day.

[0079] In some embodiments, normalization is consistently applied in the same comparison task using one of the following: full-spectrum scale normalization, stable reference signal normalization, or tissue reference normalization. Full-spectrum scale normalization uses the vector norm of the entire spectrum or the integral area of ​​a selected band as the scale. Stable reference signal normalization uses a reference signal that has undergone stability testing as the scale. Tissue reference normalization uses normal skin of the same subject, contralateral skin, or a pre-defined perilesional reference tissue as a reference. A single wavenumber must not be consistently used as a universal internal standard without verification.

[0080] Corresponding to the above method, the present invention also provides an adaptive identification system for wound microenvironment based on Raman spectroscopy, including a processor, a memory, and a computer program or instructions stored in the memory. The processor is used to execute the computer program or instructions, and when the computer program or instructions are executed, the system implements the aforementioned method steps.

[0081] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the aforementioned method. The computer-readable storage medium may be a tangible storage medium, such as random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, register, floppy disk, hard disk, removable storage disk, CD-ROM, or any other form of storage medium known in the art.

[0082] Those skilled in the art will understand that the exemplary components, systems, and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Whether implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention. When implemented in hardware, it can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this invention are programs or code segments used to perform the desired tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried in a carrier wave.

[0083] It should be clarified that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of the present invention.

[0084] In this invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or in place of features of other embodiments.

[0085] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations of the embodiments of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An adaptive identification method for wound microenvironment based on Raman spectroscopy, characterized in that, The method includes: The spatial partitioning information of the target wound is obtained. The spatial partitioning information is used to indicate the region category of multiple preset location areas on the target wound. The multiple preset location areas include the wound center area, the transition area from the center to the wound edge, the wound edge area, and the peri-wound tissue area. Based on the spatial partitioning information, at least one spontaneous Raman spectrum is collected in each preset location region, and the acquisition parameters and location attribution information corresponding to each spontaneous Raman spectrum are recorded. For each spontaneous Raman spectrum, a quality evaluation operation is performed to obtain the quality status index of the spontaneous Raman spectrum in multiple quality dimensions. Based on the quality status index of each spontaneous Raman spectrum, an adaptive parameter selection operation is performed to adapt a target preprocessing parameter combination for the spontaneous Raman spectrum from a pre-constructed candidate parameter library. The adaptive parameter selection operation includes: preprocessing the spontaneous Raman spectrum using multiple sets of candidate preprocessing parameter combinations from the candidate parameter library, evaluating the processing effect of each set of candidate preprocessing parameter combinations, and determining the set of candidate preprocessing parameter combinations whose processing effect meets the preset quality requirements as the target preprocessing parameter combination for the spontaneous Raman spectrum. The corresponding spontaneous Raman spectra are preprocessed using the adapted target preprocessing parameter combination; Candidate Raman features were extracted from the preprocessed individual Raman spectra, and the candidate Raman features included feature information of molecular vibrational spectral bands related to the wound microenvironment. Based on the candidate Raman features of all valid sampling points within each preset location area, calculate the location mean of each preset location area; Based on the average position of each preset location region, calculate at least one set of spatial gradient indices between preset location region combinations; wherein, the spatial gradient indices include gradient difference and / or gradient ratio, and the gradient difference and gradient ratio are used to characterize the difference in the change of feature values ​​of each effective sampling point in the same direction; Output the average location value and / or spatial gradient index of each preset location area.

2. The adaptive identification method for wound microenvironment based on Raman spectroscopy according to claim 1, characterized in that, The central area of ​​the wound is the geometric center point or central region of the wound; the transition area from the center to the wound edge is the midpoint or layer transition point between the central area of ​​the wound and the edge area of ​​the wound in the same direction; the edge area of ​​the wound is the boundary of the wound and its adjacent area; the peri-wound tissue area is the adjacent tissue area outside the boundary of the wound.

3. The adaptive identification method for wound microenvironment based on Raman spectroscopy according to claim 1 or 2, characterized in that, When collecting the spontaneous Raman spectrum, a cross-shaped radial sampling template with the central area of ​​the wound as the intersection point is used. The cross-shaped radial sampling template has four axes with the central area of ​​the wound as the intersection point, forming at least eight radial directions. Sampling points for the transition area, the edge area of ​​the wound, and the peri-wound tissue area are respectively set in each of the radial directions.

4. The adaptive identification method for wound microenvironment based on Raman spectroscopy according to claim 1, characterized in that, The quality status indicators include signal-to-noise ratio, fluorescence background intensity, baseline slope, saturated pixel ratio, cosmic ray anomaly, characteristic peak integrity, peak position shift, hemoglobin reference spectrum matching degree, repeat spectrum consistency, and instrument drift. The quality thresholds corresponding to each quality status indicator are predetermined by the statistical distribution of quality control samples, qualified spectra from pre-experiments, quantiles of training samples, or histological and biochemical calibration results.

5. The adaptive identification method for wound microenvironment based on Raman spectroscopy according to claim 1, characterized in that, In the adaptive parameter selection operation, the candidate preprocessing parameter combination includes baseline correction parameters, smoothing window parameters, and multi-peak fitting constraint parameters; the preset quality requirements include baseline residual conditions, post-processing signal-to-noise ratio conditions, target peak position and peak area preservation conditions, and repeatability spectrum consistency conditions.

6. The adaptive identification method for wound microenvironment based on Raman spectroscopy according to claim 5, characterized in that, In the adaptive parameter selection operation: For spectra with fluorescence background intensity exceeding a preset threshold, the candidate preprocessing parameter combination with the smallest baseline residual is selected from the combinations of candidate preprocessing parameters that satisfy the baseline residual condition, the signal-to-noise ratio condition after processing, the condition for maintaining the peak position and peak area of ​​the target peak, and the condition for consistency of repeated spectra. For spectra with a signal-to-noise ratio lower than a preset threshold, re-acquisition is performed first; if the signal-to-noise ratio is still lower than the preset threshold after re-acquisition, a smoothing window parameter is selected, provided that the peak position and peak area of ​​the target peak are maintained to the extent that the preset quality requirements are met. For spectra with peak overlap exceeding a preset threshold, multi-peak fitting with constraints on peak position, peak width, and non-negative peak area is invoked.

7. The adaptive identification method for wound microenvironment based on Raman spectroscopy according to claim 1, characterized in that, The method further includes: Based on the candidate Raman features of each valid sampling point within each preset location area, a spatial heterogeneity index for each preset location area is calculated. The spatial heterogeneity index includes the coefficient of variation and / or the range. The coefficient of variation is used to characterize the dispersion of the feature values ​​of each valid sampling point relative to the mean of the area, and the range is used to characterize the maximum fluctuation range of the feature values ​​of each valid sampling point.

8. The adaptive identification method for wound microenvironment based on Raman spectroscopy according to claim 1, characterized in that, When data from at least two data collection points are available for the target wound, the method further includes: Calculate the rate of change of the same spatial gradient index between adjacent acquisition time points. The rate of change is used to characterize how fast and in what direction the spatial gradient index changes over time.

9. The adaptive identification method for wound microenvironment based on Raman spectroscopy according to claim 1, characterized in that, The candidate Raman features include wavenumber 498 cm⁻¹. -1 1094cm -1 1250cm -1 1307cm -1 1064cm -1 1000cm -1 and 1122cm -1 The candidate Raman features are represented by peak height, peak area, fitted peak area, spectral decoupling coefficient, or model output calibrated by histological and biochemical indicators.

10. A wound microenvironment adaptive recognition system based on Raman spectroscopy, comprising a processor, a memory, and a computer program or instructions stored in the memory, characterized in that, The processor is configured to execute the computer program or instructions, and when the computer program or instructions are executed, the system implements the steps of the method as described in any one of claims 1 to 9.

Images