A method for assessing nerve fiber density in different epidermal layers

Abstract

This invention discloses a method for evaluating the density of nerve fibers in different epidermal layers, comprising: acquiring immunofluorescence images of skin sections, obtaining image units after preprocessing and standardization; splitting the image units into multiple color channels, and obtaining segmentation mask images of the background, dermal collagen, epidermis, cell nuclei, and epidermal nerve fibers based on the fluorescence signals of different channels; performing logical reasoning based on each segmentation mask image to extract the epidermal-dermal junction contour, the lower epidermal region, the lower epidermal-stratum corneum junction contour, and the stratum corneum region; and fusing the epidermal nerve fiber mask image with the segmentation mask image of each cortex after skeletonization processing to extract the nerve fibers in each cortex; calculating the total length of nerve fibers in each cortex and the area or length of each cortex, and calculating the nerve fiber density in each epidermal layer; this invention achieves automatic segmentation and length density quantification of the four epidermal layers, with low cost and high automation.

Description

Technical Field

[0001] This invention relates to the interdisciplinary fields of biomedicine and artificial intelligence, specifically to a method for assessing the density of nerve fibers in different epidermal layers. Background Technology

[0002] The epidermis is composed of multiple layers of stratified epithelium, consisting of the basal layer, spinous layer, granular layer, and stratum corneum from the inside out. The basal layer, spinous layer, and granular layer are composed of active keratinocytes and are collectively referred to as the lower epidermal layer; the stratum corneum is composed of dead, flattened keratinocytes and acts as a physical barrier. The distribution of nerve fibers (mainly C-fibers and Aδ-fibers) within the epidermis exhibits a distinct hierarchical distribution, primarily located in the basal layer under normal conditions, gradually decreasing towards the surface. In pathological states such as chronic pain or itching, the distribution of nerve fibers shows hierarchical-specific changes: for example, in brachioradialis pruritus, the number of nerve fibers at the basement membrane is reduced while the density in the spinous layer is increased; in chronic nodular pruritus, the number of nerve fibers at the basement membrane and spinous layer is reduced, while abnormal proliferation occurs near the granular layer. Therefore, accurately quantifying the nerve fiber density of different epidermal layers is crucial for the early diagnosis, efficacy evaluation, and mechanistic research of sensory disorders.

[0003] However, existing assessment methods have obvious shortcomings: traditional methods only count the total number of fibers in the entire epidermis or those passing through the basement membrane, which cannot achieve layered quantification and loses important spatial distribution information; manual counting is highly subjective and time-consuming, making it difficult to meet the needs of standardized, high-throughput analysis; professional image analysis software is expensive and has a high operating threshold, while open-source software lacks a dedicated analysis module for epidermal nerve fibers. Summary of the Invention

[0004] Purpose of the invention: The purpose of this invention is to provide a method for assessing the density of nerve fibers in different epidermal layers, thereby solving the problems of existing technologies that cannot achieve quantitative analysis of different epidermal layers and the high subjectivity and low efficiency of manual counting.

[0005] Technical solution: The present invention provides a method for evaluating the density of nerve fibers in different epidermal layers, comprising the following steps: Step 1: Acquire immunofluorescence images of skin sections, and obtain image units after preprocessing and standardization; Step 2: Divide the image unit into multiple color channels, and obtain segmentation mask maps of the background, dermal collagen, epidermis, cell nucleus and epidermal nerve fibers based on the fluorescence signals of different channels. Step 3: Based on the segmentation mask images, perform logical reasoning to extract the epidermal-dermal junction contour, the lower epidermal region, the lower epidermal-stratum corneum junction contour, and the stratum corneum region. After skeletalization processing, the epidermal nerve fiber mask image is fused with each cortical segmentation mask image to extract the nerve fibers in each cortical layer. Step 4: Calculate the total length of nerve fibers in each cortical layer and the area or length of each cortical layer, and calculate the nerve fiber density in each epidermal layer.

[0006] Furthermore, in step 1, the skin section is a frozen section cut vertically into the epidermis, and triple immunofluorescence staining is used to label collagen, nerve fibers and cell nuclei respectively.

[0007] Furthermore, in step 1, preprocessing and standardization include removing non-target structures from the image and performing non-overlapping image block processing.

[0008] Furthermore, in step 2, a segmentation mask map of the background, dermal collagen, and epidermal layer is obtained based on the fluorescence signal of the red channel, as follows: the background area is determined by the threshold of the red channel, the collagen contour is located by the high threshold fluorescence signal, and the overall contour of the epidermal layer is identified by the low threshold background fluorescence.

[0009] Furthermore, in step 2, a segmentation mask map of the cell nucleus is obtained based on the fluorescence signal of the blue channel, as follows: the cell nucleus is identified by the fluorescence signal of the blue channel, cell nucleus connectivity is achieved through morphological expansion, and region fusion is completed by gap filling.

[0010] Furthermore, in step 2, the fluorescent signal of the green channel is used to obtain the epidermal nerve fiber mask map. Specifically, the nerve fibers are identified by the fluorescent signal of the green channel, and the nerve fibers in the epidermal layer are extracted by feature fusion combined with the epidermal layer segmentation mask map. The nerve fiber mask map is generated by skeleton extraction, filtering and threshold segmentation.

[0011] Furthermore, in step 3, the extraction of the epidermal-dermal junction contour is as follows: the intersection of the dermal collagen contour and the epidermal contour is used as a spatial constraint to extract the junction contour; the extraction of the lower epidermal region is as follows: the nucleus region is differentially processed with the epidermal-dermal junction contour to obtain the lower epidermal region composed of active keratinocytes; the extraction of the stratum corneum region is as follows: the lower epidermal region is subtracted from the epidermal contour to obtain the stratum corneum region.

[0012] Furthermore, the contour of the lower epidermal-stratum corneum boundary is extracted as follows: the boundary contour is extracted by finding the intersection of the stratum corneum region and the lower epidermal region.

[0013] Furthermore, nerve fiber density is characterized by dividing the total length of nerve fibers in each cortex by the area of ​​the cortex or the length of the junctional line, where the area is used as the denominator for regional cortex and the length is used as the denominator for junctional cortex.

[0014] Furthermore, the formula for calculating nerve fiber density is as follows: ; in, Choose 1, 2, 3, 4. Pick , , , , For each cropping image Total length of nerve fibers in the layer, For each cropping image The area or length of the layer; when When taking 2 or 4, For area, when When taking 1 or 3, Let be the length; where, , , , These represent the outline of the epidermal-dermal junction, the lower epidermal region, the outline of the epidermal-stratum corneum junction, and the stratum corneum region E, respectively.

[0015] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: (1) Automatic segmentation of the epidermis: The red channel is used to identify the epidermal-dermal junction E1, the blue channel is used to identify the lower epidermal layer E2, and set operations are used to identify E3 and the stratum corneum E4. For the first time, automatic and accurate segmentation of four different epidermal regions has been achieved.

[0016] (2) Using length density index: The total length of nerve fibers divided by the cortical area (or the length of the junction) is used as the density index. Compared with the traditional manual counting method (which only counts nerve fibers that cross the dermal-epidermal junction), it more directly reflects the distribution abundance of nerve fibers and is more accurate in assessing long and thin fibers.

[0017] (3) High-throughput automated analysis: This invention automatically completes image segmentation, neural fiber skeletonization, length extraction and density calculation, reducing the analysis time of a single slice from tens of minutes to a few minutes and eliminating the subjective bias of manual counting.

[0018] (4) The present invention is simple to operate, highly reproducible, and easy to promote and popularize in ordinary laboratories. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the process of the present invention; Figure 2 This is the skin slice image unit of the present invention; Figure 3 This is a schematic diagram illustrating the differentiation of different cortical layers in this invention; Figure 4 This is a schematic diagram of the background outline of the present invention; Figure 5This is a schematic diagram of the dermal collagen outline of the present invention; Figure 6 This is a schematic diagram of the epidermal layer outline of the present invention; Figure 7 This is a schematic diagram of the outline of the cell nucleus segmentation mask of the present invention; Figure 8 This is a mask diagram of the epidermal nerve fibers of the present invention; Figure 9 This is a schematic diagram of the epidermal-dermal junction contour E1 identified by the present invention; Figure 10 This is a schematic diagram of the outline of the lower epidermal region E2 of the present invention; Figure 11 This is a schematic diagram of the outline of the stratum corneum region E4 of the present invention; Figure 12 This is a schematic diagram of the contour E3 of the lower epidermal-stratum corneum junction line of the present invention; Figure 13 This is a schematic diagram of the topological framework of the central axis of each cortical nerve fiber in this invention. Detailed Implementation

[0020] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0021] See appendix Figure 1 In a preferred embodiment, a method for assessing the density of nerve fibers in different epidermal layers is provided, comprising the following steps: Step S1: Image acquisition and standardization.

[0022] The prepared skin tissue was longitudinally sectioned perpendicular to the epidermis using a cryostat, resulting in 50 μm thick frozen sections. Triple immunofluorescence staining was performed on the skin sections: COL IV antibody was used for the red channel to label type IV collagen, PGP9.5 antibody for the green channel to label nerve fibers, and DAPI antibody for the blue channel to label cell nuclei. Images of the entire skin section were acquired along the Z-axis at 2 μm steps under a 20x objective lens using a laser confocal microscope. Non-target structures such as hair follicles and sweat glands were removed from the acquired images. The preprocessed images were then divided into non-overlapping image blocks to obtain multiple image units. (See attached image) Figure 2 This is the original image of one of the typical image units, where image A is a combined image of the red, blue and green channels, image R is a red channel image (showing type IV collagen), image B is a blue channel image (showing the cell nucleus), and image G is a green channel image (showing nerve fibers).

[0023] Conventional skin cryosections are typically 5-20 μm thick, but this protocol uses 50 μm thick sections. This is because nerve fibers within the epidermis are long, thin, and have complex branching; excessively thin sections can easily sever nerve fibers, leading to underestimated fiber length or false negatives. The 50 μm thickness balances dye penetration efficiency with the thickness of the sections, avoiding staining blurring and poor contrast caused by excessively thick sections. It also improves the integrity of nerve fibers in three-dimensional space, ensuring the accuracy of subsequent statistical analysis. Three-dimensional stacked images can record the direction of nerve fibers at different depths. When generating a two-dimensional characterization map through maximum intensity projection, fiber signals at all depth levels are preserved, avoiding missed fiber detections caused by single-focal-plane imaging.

[0024] When using a 20x objective lens, the confocal optical section thickness is approximately 1.8-2.2 μm. A 2 μm step size for layer-by-layer scanning allows for precise resolution of the basement membrane boundary, tracing of delicate epidermal nerve fibers, and complete reconstruction of the three-dimensional spatial relationship between nerve fibers and the type IV collagen basement membrane. If the step size is too large, thin basement membrane layers and short-branched nerve endings are prone to sampling omissions, leading to distortion in subsequent semi-automatic segmentation and fiber density quantization results. If the step size is too small (e.g., 0.5 μm, 1 μm), there is significant overlap of fluorescence signals in adjacent Z-layer images, resulting in excessive interlayer signal redundancy. This significantly increases image storage, prolongs image acquisition time, and also increases the risk of fluorescence quenching.

[0025] Step S2: Image segmentation model construction.

[0026] S21. Divide the image unit into red channel, green channel and blue channel.

[0027] S22. In the red channel, the background area is determined by a threshold. See Appendix. Figure 4 The outline of COL IV-labeled dermal collagen was located using the high-threshold fluorescence signal in the red channel; see appendix. Figure 5 The overall contour of the epidermis is identified by low-threshold background fluorescence in the red channel; see appendix. Figure 6 .

[0028] In the red channel, in addition to the strong fluorescent signal produced by COLⅣ-labeled dermal collagen, the epidermal tissue itself produces a weaker non-specific fluorescence, which is referred to as "background fluorescence" in this application. This background fluorescence can be captured by setting a low threshold (5-60), thereby identifying the outline of the entire epidermal layer.

[0029] S23. DAPI-labeled cell nuclei are identified using the blue channel fluorescence signal; cell nuclei are connected through morphological dilation, and then region fusion is completed through gap filling to obtain a cell nucleus segmentation mask image (see appendix). Figure 7 .

[0030] S24. Identify nerve fibers using the PGP9.5 fluorescence signal from the green channel; perform feature fusion by combining the epidermal layer segmentation mask to extract nerve fibers within the epidermal layer; generate an epidermal layer nerve fiber mask through skeleton extraction, Gaussian filtering, and threshold segmentation (see appendix). Figure 8 .

[0031] S3, Parse the object segmentation steps.

[0032] S31. Using the intersection of the dermal collagen contour and the epidermal contour as a spatial constraint, the epidermal-dermal junction contour E1 is extracted. (See Appendix) Figure 9 The nuclear region was differentiated from the epidermal-dermal junction contour E1 to obtain the lower epidermal region E2, which consists of active keratinocytes. (See Appendix) Figure 10 The spatial distribution of this region strictly corresponds to the keratinocyte differentiation gradient, with clear boundaries and continuous contours. Subtracting the lower epidermal region E2 from the epidermal contour yields the stratum corneum region E4. (See appendix) Figure 11 Determine the stratum corneum region and the lower epidermal region. The intersection of the epidermis and stratum corneum is used to extract the contour E3 of the lower epidermal-stratum corneum boundary. (See Appendix) Figure 12 .

[0033] S32. Based on the epidermal neural fiber mask image, perform skeletonization processing, and fuse the images with the segmentation masks of each epidermal layer (E1, E2, E3, E4) to extract neural fibers from each epidermal layer (E1, E2, E3, E4). (See Appendix) Figure 13 .

[0034] Appendix Figure 3 This is a schematic diagram illustrating the differentiation of different cortical layers in this application, clearly showing... (epidermal-dermal junction) (Lower epidermal region) (Lower epidermis-stratum corneum junction) and (Cuticle area).

[0035] The following section provides a detailed explanation of S2 and S3 in conjunction with specific image processing procedures. The specific processing procedures are as follows: Step S2: Image segmentation model construction.

[0036] S21. Based on the objective lens used for image acquisition, map the image pixel size to a uniform physical resolution. Split the RGB channels: Split the original image to obtain: Original Image R, Original Image G, and Original Image B.

[0037] S22. The original image R is segmented using a fixed threshold, with the threshold range set to (0, 5). Connected component filtering is performed using area selection rules, retaining only the largest connected component to generate a binary mask for the background region. Pixel zeroing is then performed. The blank areas of the slide lack specific fluorescence of COL IV, containing only microscope dark current and camera noise forming extremely low grayscale pixels. The pure background pixels have stable grayscale values ​​between 0 and 5. Using an upper limit of 5 as the cutoff allows for complete removal of the signal-free background without mistakenly removing weak COL IV signals from the epidermal edges. If the cutoff value is >5, weakly fluorescent epidermal edges will be mistakenly classified as background and lost. If the cutoff value is <5, a large amount of noise background will remain, interfering with subsequent mask generation. The original image R is segmented using a fixed threshold, with the threshold range set to (5, 60). A morphological dilation operation is performed on the obtained contour at a scale of 0.75 μm, and holes are filled to obtain a binary mask for the epidermal region. This achieves accurate extraction of the epidermal contour through the background fluorescence of the red channel. The epidermal layer exhibits weak fluorescence expression, with pixel grayscale concentrated in the 5–60 range. This range contains only weak fluorescence signals from the epidermis and excludes high-brightness collagen signals from the dermis and background noise. If the upper threshold is below 60, the relatively brighter fluorescent segments in the epidermis will be truncated, resulting in incomplete epidermal contours. If the upper threshold is above 60, high-brightness collagen pixels from the dermis will be mixed in, leading to excessive expansion of the epidermal mask and an inability to accurately distinguish between the anatomical regions of the epidermis and dermis. The epidermal boundary is delineated by the COL Ⅳ signal of a thin basement membrane. The natural thickness of the basement membrane is only 1–3 μm, and the fluorescence signal is delicate. Locally, it is prone to tiny discontinuities and gaps due to section compression and uneven fluorescence. The 0.75 μm microscale expansion only fills the tiny discontinuities and micropores of the basement membrane and does not significantly expand the contour outward, strictly adhering to the true anatomical boundary of the epidermis. If the expansion scale is significantly greater than 0.75 μm, the epidermal mask will expand excessively towards the dermis, encroaching on the dermal region; if the expansion scale is less than 0.75 μm, the basement membrane breakpoints cannot be closed, resulting in gaps and discontinuous contours in the epidermal mask, failing to form a complete closed epidermal region. Therefore, 0.75 μm is the optimal scale for repairing minor basement membrane breaks without compromising the precision of the epidermal boundary. A fixed threshold segmentation was performed on the original image R, with the threshold range set to (60, 255). Connected component filtering was performed using area selection rules, retaining only the largest connected components. A morphological expansion operation was performed on the resulting contour at a scale of 2.1 μm to fill the holes, obtaining a binary mask for the dermal collagen region. The dermis is rich in densely packed type IV collagen fibers, exhibiting strong fluorescence signals and high grayscale. All dermal collagen-positive signal pixels have a grayscale greater than 60, with the upper grayscale limit of 255 being the maximum grayscale value of the imaging device, completely covering all high-brightness fluorescence signals in the dermis. If the lower threshold is below 60, epidermal pixels will be included in the dermal mask, causing confusion between the epidermal and dermal regions, which will directly affect the subsequent quantitative zoning of epidermal nerve fibers.Dermal collagen consists of reticular fiber bundles, with collagen fibers appearing as dispersed strands and naturally occurring micro-gaps between them. COL IV fluorescence shows a discontinuous reticular distribution, requiring larger-scale expansion to connect the dispersed collagen and form a complete, continuous dermal region. A 2.1 μm expansion scale can effectively bridge the dispersed collagen fibers within the dermis, filling the micro-gaps between the fibers and resulting in a fully connected dermal mask after pore filling. Simultaneously, this scale does not breach the epidermal-dermal junction basement membrane, preventing the dermal mask from intruding into epidermal regions. If the expansion scale is below 2.1 μm, the reticular collagen cannot connect with each other, and the dermal region will be fragmented into numerous small connected domains, failing to form a complete dermal mask. If the expansion scale is much larger than 2.1 μm, the mask will cross the basement membrane and intrude into the epidermal layer, causing confusion between the epidermal and dermal regions and complete distortion of zoning and quantification.

[0038] S23. Using a binary mask of the epidermal region as a spatial constraint, perform fixed threshold segmentation on the original image B, with the threshold range set to (40, 255). Run the process to generate a binary image and determine the cell nucleus outline. Perform morphological dilation on the obtained outline at a scale of 0.75 μm and fill the holes. Combine this with area filtering rules to filter connected components, obtaining the cell nucleus segmentation mask image. The grayscale of pixels with complete cell nuclei and weak fluorescence at the cell nucleus edges is ≥40, while 255 is the maximum grayscale value of an 8-bit image from the imaging device, which can completely cover all high-brightness cell nucleus fluorescence pixels without signal truncation or loss. If the segmentation lower limit is >40: weak DAPI fluorescence at the cell nucleus edges will be removed, resulting in incomplete cell nucleus outlines and missed detection of small epidermal cell nuclei, leading to an under-count of nuclei. If the segmentation lower limit is <40: low-grayscale noise in the stratum corneum will be misjudged as nuclear signals, generating pseudo-nucleus connected components, interfering with subsequent analysis. Epidermal cell nuclei are micron-sized structures. Inconsistent slice thickness, laser scattering, and fluorescence staining can easily cause minute breaks and gaps in the nuclear outline. Direct thresholding results in discontinuous nuclear outlines that cannot form complete, closed nuclear regions. A micro-expansion scale of 0.75 μm fills only the minute breaks at the nuclear edges and tiny internal pores, without significantly expanding the nuclear outline outwards, thus conforming to the true anatomical boundaries of the nucleus.

[0039] S24. Using the epidermal region as a spatial constraint, coarse thresholding is performed on the original image G, with the threshold range set to (100, 255), to obtain the core region of strong fluorescent signals in the green channel. Gaussian smoothing and fine segmentation with a low threshold of (10, 255) are then used to connect broken weak signal segments. A binary mask image of the epidermal nerve fibers is obtained using a dual-threshold iterative framework. In the green channel, PGP9.5 nerve fibers exhibit significant gray-level gradient differences: the nerve fiber trunk and bulging areas are fluorescently enriched, with pixel gray levels ≥100, forming a complete and continuous strong fluorescent core; while nerve endings, fine branches, and superficial epidermal fibers generally have gray levels below 100 due to low fluorescence loading and light scattering loss; background stray signals such as epidermal cytoplasm, intercellular spaces, and imaging noise all have gray levels below 10. A coarse threshold of 100 preserves only the fiber backbone regions with high signal-to-noise ratio and no noise interference, quickly removing large areas of low grayscale background and weak specks from the epidermis. This first locks the main location of the nerve fiber, constructing a complete fiber core framework and eliminating the computational interference of numerous irrelevant pixels for subsequent fine segmentation, reducing iterative computational pressure. After Gaussian smoothing to suppress random noise, a low threshold (10, 255) is used for fine segmentation, which can completely capture broken weak fiber fragments within the grayscale range of 10-100, connecting scattered fine branches with the backbone obtained from coarse segmentation to restore the complete neural fiber course.

[0040] S3, Parse the object segmentation steps.

[0041] S31. Fuse the binary mask of the epidermal region with the binary mask of the dermal collagen region, extract the intersection area of ​​the two, and obtain image E1, which is the contour of the epidermal-dermal junction. Perform a pixel zeroing operation with the E1 binary mask as a constraint, and remove the corresponding area of ​​the epidermal-dermal junction contour E1 in the nucleated cell layer region to obtain image E2, which is the lower epidermal region. Fuse the binary mask of the epidermal region with the binary mask of the E2 region, perform a set difference operation, and perform connected component filtering based on the area filtering rule to obtain image E4, which is the binary mask of the stratum corneum region. Perform a second set difference operation on the binary masks of the E2 region and the E4 region to obtain image E3, which is the contour of the lower epidermal-stratum corneum junction.

[0042] S32. Through morphological iterative refinement, the two-dimensional planar region of epidermal nerve fibers is reduced to a single-pixel-wide central axis topological skeleton, fully preserving the topological connectivity and branching features of the signal. The topological skeleton is fused with binary masks of target cortical layers E1, E2, E3, and E4, respectively, and nerve fiber segments with a length ≥1 μm are selected and retained. The total length of independent nerve fiber segments in each region is then statistically calculated. When the center line of a single pixel is obtained through morphological iteration, local dotted fluorescent noise and residual background noise will be refined to generate a large number of isolated, extremely short pseudo skeletons; setting a length threshold of 1μm can filter out all meaningless fragmented pseudo line segments at once and eliminate false positive counting interference.

[0043] S4, Computational Model.

[0044] S41. Quantitative analysis of layered morphology.

[0045] A solid binary region is generated using E1 as the spatial constraint. After dimensionality reduction, a single-pixel-width centerline topological skeleton is obtained, and its length is statistically analyzed to obtain the total length of the epidermal-dermal junction contour E1. The same processing method is used to obtain the total length of the lower epidermal-stratum corneum junction contour E3. A solid binary region is generated using E2 as the spatial constraint, and the region area is statistically analyzed. The same processing method is used to obtain the area of ​​the stratum corneum E4.

[0046] S42. Input the above parameters and calculate the nerve fiber density in each epidermal layer of E1, E2, E3, and E4.

[0047] Calculate using the following formula: ; in, Choose 1, 2, 3, 4. Pick , , , , For each cropping image Total length of nerve fibers in the layer, For each cropping image The area or length of the layer; when When taking 2 or 4, For area, when When taking 1 or 3, Let be the length; where, , , , These represent the outline of the epidermal-dermal junction, the lower epidermal region, the outline of the epidermal-stratum corneum junction, and the stratum corneum region E, respectively.

[0048] When compiling statistics, for and (Boundary line) The length is expressed in μm, and the density unit is μm / μm (fiber length / junction length); for and (area), Area (unit: square μm), density unit is density unit μm / μm² (fiber length / area).

[0049] Table 1. Examples of nerve fiber density calculations for different epidermal layers. ; The results above show that the image segmentation model successfully identified [the image]. , , , Each cortex is analyzed, and the total length of nerve fibers is automatically calculated. and Nerve fibers were detected in the area. and No area was detected, consistent with normal skin physiological distribution. The entire analysis process took approximately 3 minutes, significantly better than the 30 minutes or more required for manual counting.

[0050] It should be noted that channel splitting, thresholding, morphological dilation and filling, skeleton extraction, Gaussian filtering, and set operations all employ well-known techniques in the fields of digital image processing and computer vision. Those skilled in the art can implement these techniques based on the descriptions in this manual, combined with common knowledge and conventional experimental skills.

[0051] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for assessing the density of nerve fibers in different epidermal layers, characterized in that, Includes the following steps: Step 1: Acquire immunofluorescence images of skin sections, and obtain image units after preprocessing and standardization; Step 2: Divide the image unit into multiple color channels, and obtain segmentation mask maps of the background, dermal collagen, epidermis, cell nucleus and epidermal nerve fibers based on the fluorescence signals of different channels. Step 3: Based on the segmentation mask images, perform logical reasoning to extract the epidermal-dermal junction contour, the lower epidermal region, the lower epidermal-stratum corneum junction contour, and the stratum corneum region. After skeletalization processing, the epidermal nerve fiber mask image is fused with each cortical segmentation mask image to extract the nerve fibers in each cortical layer. Step 4: Calculate the total length of nerve fibers in each cortical layer and the area or length of each cortical layer, and calculate the nerve fiber density in each epidermal layer.

2. The method for evaluating the density of nerve fibers in different epidermal layers according to claim 1, characterized in that, In step 1, the skin sections are frozen sections cut vertically into the epidermis and are stained with triple immunofluorescence to label collagen, nerve fibers and cell nuclei respectively.

3. The method for evaluating the density of nerve fibers in different epidermal layers according to claim 1, characterized in that, In step 1, preprocessing and normalization include removing non-target structures from the image and performing non-overlapping image block processing.

4. The method for evaluating the density of nerve fibers in different epidermal layers according to claim 1, characterized in that, In step 2, a segmentation mask map of the background, dermal collagen, and epidermal layer is obtained based on the fluorescence signal of the red channel, as follows: the background area is determined by the threshold of the red channel, the collagen contour is located by the high threshold fluorescence signal, and the overall contour of the epidermal layer is identified by the low threshold background fluorescence.

5. The method for evaluating the density of nerve fibers in different epidermal layers according to claim 1, characterized in that, In step 2, a segmentation mask map of the cell nucleus is obtained based on the fluorescence signal of the blue channel, as follows: the cell nucleus is identified by the fluorescence signal of the blue channel, cell nucleus connectivity is achieved through morphological expansion, and region fusion is completed by gap filling.

6. The method for evaluating the density of nerve fibers in different epidermal layers according to claim 1, characterized in that, In step 2, the fluorescent signal of the green channel is used to obtain the epidermal nerve fiber mask map. Specifically, the nerve fibers are identified by the fluorescent signal of the green channel, and the nerve fibers in the epidermal layer are extracted by feature fusion combined with the epidermal layer segmentation mask map. The nerve fiber mask map is generated by skeleton extraction, filtering and threshold segmentation.

7. The method for evaluating the density of nerve fibers in different epidermal layers according to claim 1, characterized in that, In step 3, the extraction of the epidermal-dermal junction contour is as follows: the intersection of the dermal collagen contour and the epidermal contour is used as a spatial constraint to extract the junction contour; the extraction of the lower epidermal region is as follows: the nucleus region is differentially processed with the epidermal-dermal junction contour to obtain the lower epidermal region composed of active keratinocytes; the extraction of the stratum corneum region is as follows: the lower epidermal region is subtracted from the epidermal contour to obtain the stratum corneum region.

8. The method for evaluating the density of nerve fibers in different epidermal layers according to claim 1, characterized in that, The specific steps for extracting the contour of the lower epidermal-stratum corneum boundary are as follows: The contour of the boundary is extracted by finding the intersection of the stratum corneum region and the lower epidermal region.

9. The method for evaluating the density of nerve fibers in different epidermal layers according to claim 1, characterized in that, Nerve fiber density is characterized by dividing the total length of nerve fibers in each cortex by the area of ​​the cortex or the length of the junctional line, where the area is used as the denominator for regional cortex and the length is used as the denominator for junctional cortex.

10. The method for evaluating the density of nerve fibers in different epidermal layers according to claim 9, characterized in that, The formula for calculating nerve fiber density is: ； in, Choose 1, 2, 3, 4. Pick , , , , For each cropping image Total length of nerve fibers in the layer, For each cropping image The area or length of the layer; when When taking 2 or 4, For area, when When taking 1 or 3, Let be the length; where, , , , These represent the outline of the epidermal-dermal junction, the lower epidermal region, the outline of the epidermal-stratum corneum junction, and the stratum corneum region E, respectively.

Images