A pathological tissue slice water bath detection method and device based on blue-violet light fluorescence difference

Abstract

The application discloses a pathological tissue slice water bath detection method and device based on blue-violet light fluorescence difference, and the core process of the method is as follows: blue-violet light of a preset wavelength is obliquely irradiated on a paraffin slice floating on the surface of water bath water, a band-pass filter is used to collect the self-fluorescence signal of the tissue and generate a fluorescence image; the region of interest is extracted through pretreatment, the fluorescence ratio and the fluorescence area ratio are calculated based on the mean value, the maximum value and the standard deviation of the fluorescence intensity; the determination threshold group corresponding to the current water bath temperature is loaded, and the two features are combined to determine whether the tissue paraffin slice contains the tissue. The essence of the self-fluorescence difference between the tissue and the paraffin is used to improve the contrast, the oblique incidence light path is matched with the band-pass filter to inhibit the water surface reflection and stray light interference and improve the image signal-to-noise ratio; the multi-dimensional feature joint determination is matched with the dynamic gray threshold, and the batch and environmental brightness fluctuation can be resisted, and the algorithm robustness is stronger.

Description

Technical Field

[0001] This invention relates to the fields of optical fluorescence detection, image processing, and pathological slide preparation, and particularly to a method and apparatus for water bath detection of pathological tissue slides based on the difference in blue-violet fluorescence. Background Technology

[0002] Tissue slide preparation is a core pre-procedure step in clinical pathological diagnosis and digital pathological analysis, and the quality of the slides directly determines the accuracy and reliability of subsequent pathological diagnoses. A typical paraffin-embedded tissue slide preparation process includes paraffin embedding, tissue trimming, tissue sectioning, water bath spreading, slide mounting and retrieval, staining, and mounting. Both the trimming and spreading stations require slide inspection in a water bath environment: the trimming station uses a low-temperature water bath (2-8℃) to collect the trimmed paraffin sections to prevent overheating of the blades and subsequent melting of the paraffin, thus affecting the quality of the cut surface. Inspection of the paraffin sections is necessary to determine whether the tissue surface is sufficiently exposed, thereby determining the trimming endpoint. The spreading station uses a high-temperature water bath (40-45℃) to naturally flatten curled sections using the surface tension and thermal effect of water. Inspection of sections floating on the water surface is necessary to determine whether they contain tissue, guiding subsequent slide mounting and retrieval. The core of inspection in both stages is to distinguish between "tissue-containing areas" and "pure paraffin areas," serving as a critical quality control point connecting the cutting process with subsequent steps.

[0003] Currently, in the pathological slide preparation industry, the two water bath testing steps mentioned above generally rely on manual visual judgment. Operators need to be on duty in front of the water bath, observing the differences in translucency of the slides with the naked eye, distinguishing between tissue-containing slides and pure paraffin slides based on experience, and judging the progress of trimming. However, the difference in transparency between unstained tissue slides and pure paraffin slides under visible light is extremely small, making it difficult to distinguish with the naked eye. This manual testing method has many inherent defects: First, the judgment is highly subjective, and the test results are significantly affected by the operator's experience level, fatigue level, and ambient lighting conditions. The consistency of judgments among different operators is poor, and it is easy to miss or misjudge. Second, the testing efficiency is low. Operators need to be on duty throughout the process and judge each slide individually, resulting in high labor costs and being one of the core bottlenecks restricting the automation of the entire slide preparation process. Third, the test results cannot be digitally recorded and traced, making it difficult to meet the traceability requirements of quality control systems such as pathology laboratories for the slide preparation process. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of the prior art by providing a method and apparatus for water bath detection of pathological tissue sections based on the difference in blue-violet fluorescence.

[0005] The objective of this invention is achieved through the following technical solution: a water bath detection method for pathological tissue sections based on the difference in blue-violet fluorescence, comprising the following specific steps: S1: Blue-violet light within a preset center wavelength range is used as the detection light, and the detection light is irradiated at a preset angle relative to the water surface normal onto the paraffin slice to be tested floating on the water surface of the water bath. S2: The autofluorescence signal generated by the blue-violet light excitation of the paraffin slice to be tested is collected in the imaging optical path with a bandpass filter to generate a fluorescence image; S3: Preprocess the fluorescence image to extract the region of interest corresponding to the slice to be detected; S4: Calculate fluorescence characteristic parameters within the region of interest. These parameters include at least the mean fluorescence intensity, the maximum fluorescence intensity, and the standard deviation of fluorescence intensity. Calculate the fluorescence ratio and fluorescence area ratio based on these parameters. The fluorescence ratio is the ratio of the maximum fluorescence intensity to the mean fluorescence intensity. The fluorescence area ratio is the ratio of the area of ​​pixels within the region of interest whose grayscale value exceeds a grayscale value threshold T to the total area of ​​pixels within the region of interest. The grayscale value threshold T is calculated based on the mean fluorescence intensity and the standard deviation of fluorescence intensity. S5: Load the corresponding judgment threshold group based on the current water bath temperature scenario. The judgment threshold group includes the fluorescence ratio threshold and fluorescence area ratio threshold determined based on the current water bath temperature. If the fluorescence ratio and fluorescence area ratio meet the threshold judgment conditions, it is judged as a paraffin section containing tissue; otherwise, it is judged as a pure paraffin section.

[0006] Preferably, the gray value threshold T is the product of the mean fluorescence intensity plus the standard deviation of fluorescence intensity and the confidence coefficient; the threshold determination condition is that the fluorescence ratio is greater than the fluorescence ratio threshold and the fluorescence area ratio is greater than the fluorescence area ratio threshold.

[0007] Preferably, the water bath temperature scenarios include a high-temperature water bath temperature scenario within a first water bath temperature range and a low-temperature water bath temperature scenario within a second temperature range; for the low-temperature water bath temperature scenario, when it is determined to be a paraffin section containing tissue, a "refinement complete" signal is output; when it is determined to be a pure paraffin section, a "continue refinement" signal is output.

[0008] Preferably, in step S5, the judgment threshold groups corresponding to the high-temperature water bath temperature scenario and the low-temperature water bath temperature scenario are obtained by calibration within the corresponding water bath temperature range. The specific calibration method is as follows: N1. Sample Collection: For high-temperature water bath scenarios, at least 50 paraffin sections containing different types of tissue and at least 50 pure paraffin sections were collected as calibration sets. For low-temperature water bath scenarios, paraffin blocks containing different types of tissue were collected, and at least 50 paraffin sections containing tissue were collected in both the incomplete and completed trimming states to form calibration sets. Water bath temperature ranges were set for each water bath scenario. All samples were manually labeled with gold standard tags. The gold standard tag for pure paraffin sections and paraffin sections containing tissue in the incomplete trimming state was "pure paraffin", and the gold standard tag for paraffin sections containing tissue and those in the completed trimming state was "containing tissue". N2. Data acquisition: Under the corresponding water bath temperature scenario, the fluorescence images of each sample are acquired using the same detection conditions as in step S1, and the fluorescence ratio and fluorescence area ratio of each sample are calculated. N3. ROC curves were plotted with fluorescence ratio and fluorescence area ratio as discriminant variables and gold standard label as response variable, respectively. The value corresponding to the maximum value of Youden index was selected as the fluorescence ratio threshold and fluorescence area ratio threshold for this scene.

[0009] Preferably, in step S3, the image preprocessing includes dark field subtraction, white field correction, noise suppression, adaptive background subtraction, and artifact filtering in sequence. Noise suppression is achieved by combining adaptive median filtering with Gaussian low-pass filtering. Adaptive background subtraction is used to eliminate water surface ripple interference. Adaptive background subtraction freezes water surface displacement by using a global shutter short exposure. After detection triggering, a dwell time is set, or the water surface stability is judged by inter-frame difference. If the water surface is not stable, it is resampled after a delay. The water surface ring zone inward from the image boundary is extracted as the background region. The low-frequency background components are estimated and subtracted by mean filtering or two-dimensional quadratic polynomial fitting to eliminate low-frequency water surface ripples. Artifact filtering is used to remove interference from bubbles and debris in the water. The artifact filtering generates a foreground mask with the background grayscale mean plus 3 times the background standard deviation as the threshold. Morphological opening operation is performed on the foreground mask to remove small debris, and morphological closing operation is performed to fill holes. Connectivity analysis is performed, and connected components with an area not less than the preset minimum area are retained for subsequent extraction of regions of interest. At the same time, bubble regions are identified by combining the region grayscale mean and roundness features, and bubble regions are excluded from the effective regions.

[0010] As a preferred embodiment, the method for extracting the region of interest in step S3 is as follows: The median gray level of the boundary ring of the fluorescence image is used as the background level. The foreground threshold is calculated by superimposing the background level by 3 times the background standard deviation or by the Otsu method. A binary mask is generated based on the foreground threshold. After performing morphological opening and closing operations on the binary mask, connected component analysis is performed to retain connected components whose area is within a preset area range. The union, circumscribed rectangle, or convex hull of the connected components is used as the region of interest. The preset area range has a preset minimum area as the lower limit and 60% of the effective area of ​​the fluorescence image field of view as the upper limit.

[0011] Preferably, the center wavelength of the detection light is 380-420nm, and the angle of the detection light relative to the normal of the water surface is 20°-60°. In step S2, the center wavelength of the bandpass filter is 440-520nm, the bandwidth is 40-80nm, and the out-of-band cutoff depth is ≥OD4, so as to block specular reflection and scattered stray light of the excitation light.

[0012] Preferably, the detection light is generated by driving a blue-violet LED array. The driving current of the blue-violet LED in the low-temperature water bath temperature scenario is greater than that in the high-temperature water bath temperature scenario. The exposure time for fluorescence image acquisition in the low-temperature water bath temperature scenario is greater than that in the high-temperature water bath temperature scenario. The angle of the detection light relative to the normal of the water surface in the low-temperature water bath temperature scenario is greater than that in the high-temperature water bath temperature scenario.

[0013] A water bath detection device for pathological tissue sections based on the difference in blue-violet fluorescence, comprising: The blue-violet light illumination module is used to emit blue-violet light with a center wavelength of 380-420nm as the detection light, and to irradiate the paraffin slice to be tested on the surface of the water bath at an incident angle of 20°-60° relative to the normal of the water surface. An imaging acquisition module is used to acquire fluorescence images; the imaging acquisition module includes a camera, an imaging lens and a bandpass filter; the bandpass filter is disposed in the imaging optical path to selectively transmit tissue autofluorescence and block stray or reflected light from the excitation. The image processing module is used to perform preprocessing, region of interest extraction, and multi-dimensional fluorescence feature calculation on the acquired fluorescence images; the fluorescence features include at least the mean fluorescence intensity, the maximum fluorescence intensity, the standard deviation of fluorescence intensity, the fluorescence ratio, and the fluorescence area ratio. The judgment output module has built-in judgment threshold groups corresponding to different water bath temperature scenarios. It is used to compare the calculated fluorescence ratio and fluorescence area ratio with the corresponding thresholds and output the judgment result.

[0014] Preferably, the image processing module is further configured with a water surface ripple suppression unit and an artifact filtering unit: The water surface ripple suppression unit freezes water surface movement through global shutter short exposure and judges water surface stability through inter-frame difference. If the water surface is unstable, it performs delayed resampling. The artifact filtering unit removes artifacts caused by tiny tissue debris and air bubbles through morphological operations and connected component analysis.

[0015] The beneficial effects of this invention are: 1. This invention utilizes the autofluorescence properties of endogenous substances in biological tissues. After excitation with blue-violet light, the tissue-containing area produces a significant fluorescence signal, while the pure paraffin area exhibits extremely weak fluorescence response. The signal contrast between the two is far greater than the difference in transparency under visible light, fundamentally reducing the difficulty of differentiation. By using an obliquely incident excitation light path in conjunction with a bandpass filter for imaging, the specular reflection from the water surface and interference from ambient stray light can be suppressed, significantly improving the signal-to-noise ratio of the fluorescence image and further amplifying the signal boundary between the tissue and paraffin. The invention employs a multi-dimensional feature-based determination using fluorescence ratio and fluorescence area ratio, and the grayscale threshold is dynamically calculated based on the mean and standard deviation of fluorescence intensity. This avoids the shortcomings of a single absolute intensity threshold being susceptible to batch variations and fluctuations in ambient brightness, resulting in a more robust algorithm.

[0016] 2. This invention achieves full automation of the detection process. From light signal acquisition and image processing to final judgment, no human intervention is required. The detection speed can match the production line pace, significantly reducing the labor cost at the workstation. 3. In view of the differences in paraffin state and fluorescence background between low-temperature water bath (2-8℃) for surface finishing and high-temperature water bath (40-45℃) for film spreading, this invention adopts the design of "loading corresponding judgment threshold groups according to water bath temperature". Without modifying the hardware structure, the threshold parameters can be switched at the algorithm level to adapt to the two water bath detection stations with different temperatures, ensuring that the optimal judgment accuracy can be achieved in both working conditions.

[0017] 4. This invention achieves full digitization of detection data. The original fluorescence image, fluorescence characteristic parameters, and judgment results of each slide can be stored in association with the sample number, forming a complete quality control record of the slide preparation process, which meets the quality management requirements of pathology laboratories for traceability of the slide preparation process. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the detection device of the present invention.

[0019] Figure 2 This is a flowchart of the method of the present invention.

[0020] Figure 3 This is a fluorescence image of a paraffin section from the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the scope of protection of the present invention.

[0022] Those skilled in the art should understand that, in the disclosure of this invention, the terms "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the above terms should not be construed as limiting this invention.

[0023] It is understood that the term "a" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element can be one, while in another embodiment, the number of the element can be multiple, and the term "a" should not be understood as a limitation on the number.

[0024] like Figures 1 to 3 As shown, a water bath detection method for pathological tissue sections based on the difference in blue-violet fluorescence includes the following specific steps: S1: Using blue-violet light within a preset center wavelength range as the detection light, the detection light illuminates the paraffin slice to be tested floating on the surface of the water bath at a preset angle relative to the water surface normal.

[0025] In this step, the center wavelength of the detection light is 380-420nm, and the angle of the detection light relative to the normal of the water surface is 20°-60°. The blue-violet light is generated by a blue-violet light illumination module (blue-violet LED array). The blue-violet LED array has 4-16 LEDs (6-8 LEDs in this embodiment), with a power of 0.5-3W per LED, a driving current range of 50-500mA, a PWM dimming resolution of ≥10bit, and a PWM frequency of ≥1kHz.

[0026] The paraffin sections to be tested float on the surface of the water in the water bath. This invention targets two water bath temperature scenarios: a high-temperature scenario within a first temperature range and a low-temperature scenario within a second temperature range. In the high-temperature scenario, the first water bath temperature range is 40-45°C. The test object is a tissue section (3-5 μm thick, approximately 15mm × 15mm to 25mm × 25mm) produced by a microtome, which floats on the surface of the constant-temperature water bath. The water temperature in the bath is controlled at 40-45°C, utilizing the surface tension and heat of the water to partially melt and soften the paraffin, allowing the sections to naturally flatten. The water bath is made of a transparent or light-colored material (such as transparent acrylic or glass) to facilitate optical imaging. The water depth in the bath is 10-30mm. After the paraffin sections to be tested are transferred from the microtome to the water bath, a single test is triggered directly. The test is completed while the sections are still floating on the water surface and have not yet been retrieved. The binary classification signal, which determines whether the sample contains tissue or is pure paraffin, is used for downstream automatic slide picking guidance.

[0027] The object of the low-temperature water bath testing scenario is a paraffin section cut from a wax block after rough or fine cutting. This section floats naturally on the surface of a constant-temperature water bath at 2-8°C. The paraffin is in a solid state (melting point approximately 56-60°C), and the surface of the section is rough due to the cutting, with obvious tool marks. The water temperature in the bath is controlled at 2-8°C. This low-temperature setting is to cool the cutting blade and prevent the heat from the blade from melting the paraffin on the surface of the wax block, which would cause the cut surface to become sticky and deformed due to tissue compression. The water depth in the bath should be such that the section floats on the surface, facilitating optical imaging.

[0028] After each roughing or finishing cut and before the next cut, a detection is triggered to determine whether the section contains tissue (i.e., whether the tissue surface within the paraffin block has been fully exposed, manifested as whether the tissue-containing area is visible), providing an objective basis for determining the end point of the finishing process. If a tissue-containing area is detected and the fluorescence characteristics meet the finishing completion criteria, a "finishing complete" signal is output; otherwise, a "continue finishing" signal is output.

[0029] This invention utilizes 380-420nm blue-violet light as the detection light. The core principle of this invention is that paraffin and human tissue exhibit fundamentally different fluorescence responses when excited by 380-420nm blue-violet light. This difference stems from the fundamental differences in their molecular structures. This will be explained in detail below from a chemical and physical perspective.

[0030] Chemical nature of paraffin: Paraffin is a mixture of C25-C35 n-alkanes, with the general molecular formula CnH2n+2. Its molecular skeleton is composed of sp3 hybridized carbon atoms connected by C-C single bonds (σ bonds), and further connected by CH single bonds (also σ bonds). These σ bonds have antibonding orbitals (σ... The band gap between the σ and bonding orbitals (σ) is in the vacuum ultraviolet band (typically <180nm, corresponding to energies >7eV). The photon energy of blue-violet light at 380-420nm is approximately 2.95-3.26eV, much smaller than that of the σ→σ band gap in paraffin molecules. The minimum energy required for electron transition. Therefore, paraffin molecules absorb almost no blue-violet light in the 380-420 nm range, meaning photons penetrate the paraffin molecular system without electronic transition. Furthermore, the paraffin molecule structure lacks a conjugated double bond system (no alternating single and double bonds), aromatic rings, and chromophores such as carbonyl groups (C=O), thus eliminating the possibility of π→π transitions. or n→π The transition requires a lower energy level excited state. Therefore, paraffin does not produce autofluorescence when excited by blue-violet light in the 380-420 nm range; instead, it produces Rayleigh scattering at the same wavelength as the incident light and extremely weak Raman scattering. Rayleigh scattering occurs at the same wavelength as the incident light (405 nm) and can be effectively filtered out by a bandpass filter.

[0031] Sources of fluorophores in human tissues: Tissues fixed in formalin and embedded in paraffin contain various endogenous fluorophores, mainly including the following components: 1. Collagen: Collagen is a major component of the extracellular matrix. Its molecules contain pyridine cross-links, forming a conjugated π-electron system. The excitation wavelength range is 370-405nm, and the emission peak is about 420-480nm. 2. Elastin: Elastin is the main component of blood vessel walls and elastic fibers. Its cross-linked structure (desmosine and isodesmosine) forms a conjugated system. The excitation wavelength is 320-400nm, and the emission peak is about 400-490nm. 3. Reduced nicotinamide adenine dinucleotide (NADH), a coenzyme for cell metabolism, with an excitation wavelength of approximately 340-370 nm and an emission peak of approximately 440-480 nm; 4. Flavin adenine dinucleotide (FAD): This substance is a coenzyme for mitochondrial metabolism. Its excitation wavelength is about 450 nm, and its emission peak is about 520-540 nm (its contribution is smaller when excited at 405 nm).

[0032] The common characteristic of the aforementioned endogenous fluorophores is the presence of conjugated π-electron systems in their molecular structures (alternating single and double bonds, aromatic rings, etc.). The band gap between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) of these π-electron systems is in the range of 2.5-3.5 eV, which corresponds precisely to the photon energy of 380-420 nm blue-violet light (2.95-3.26 eV), allowing for effective π→π transitions. Electronic transition. After losing some energy through nonradiative relaxation (vibrational relaxation, internal conversion), a molecule in an excited state transitions from the lowest vibrational energy level of the excited state back to various vibrational energy levels of the ground state, releasing fluorescent photons with a longer wavelength than the excitation light (Stokes shift), i.e., blue-green fluorescence of 440-520 nm.

[0033] In this application, the wavelength of the detection light is chosen to be 405nm. The basis for choosing the 405nm wavelength is: 405nm blue-violet light simultaneously meets the following three conditions: (1) it is located within the main excitation peak range of collagen and elastin (370-405nm), with high excitation efficiency; (2) it has sufficient Stokes shift (>15nm) between it and the emission peak of the above-mentioned fluorophores (420-520nm), which facilitates the separation of excitation light and fluorescence with filters; (3) LED devices with a wavelength of 405nm are industrially mature, low in cost, long in life (>50,000 hours), and have high light output power density, making them suitable for industrial online detection.

[0034] The effect of water temperature on fluorescence mechanism: The absorption coefficient of water molecules (H2O) is less than 0.001 cm⁻¹ in both the 405 nm and 440-520 nm wavelength ranges. - ¹ It is almost completely transparent to both excitation light and fluorescence. Water temperature changes from 2°C to 45°C have no substantial impact on the optical properties of water, the chemical structure of paraffin (σ-bond band gap), or the conjugated structure of fluorophores in tissues (π-electron system). Therefore, the chemical-physical nature of fluorescence differences remains consistent across the entire temperature range of 2-45°C. Water temperature changes only affect the physical state of paraffin: below its melting point, paraffin is solid (rough surface, strong scattering, corresponding to solid sections in cold water (rough surface)); near its softening point, paraffin gradually softens (smooth surface, weak scattering, corresponding to flattened sections in hot water). These two differences in physical state correspond to different background scattering levels, thus requiring separate threshold settings, but without altering the fundamental distinguishing criterion of "paraffin does not fluoresce, tissue fluoresces."

[0035] S2: The autofluorescence signal generated by the blue-violet light excitation of the paraffin slice to be tested is collected in the imaging optical path with a bandpass filter to generate a fluorescence image.

[0036] In this step, the imaging acquisition module acquires fluorescence images. The imaging acquisition module includes a camera (global shutter CMOS, resolution ≥ 5 million pixels, dynamic range ≥ 12 bits, quantum efficiency ≥ 40% in the 440-550nm band), an imaging lens, and a bandpass filter; the bandpass filter is disposed in the imaging optical path to selectively transmit tissue autofluorescence and block stray or reflected light from the excitation.

[0037] The imaging lens has its working distance and magnification selected according to the size of the water bath. The camera and bandpass filter are positioned directly above or diagonally above the water bath, and the bandpass filter can be placed in front of the imaging lens.

[0038] In this embodiment, the center wavelength of the bandpass filter is 440-520nm, the bandwidth is 40-80nm, and the out-of-band cutoff depth is ≥OD4. The function of the bandpass filter is to selectively transmit the 440-520nm fluorescence signal generated by the tissue, while blocking the following unwanted light: 1. Specular reflection of the detection light on the water surface (the direction follows the law of specular reflection, but part of it enters the camera's field of view); (2) Scattered light of the detection light in the water (Rayleigh scattering, whose wavelength is consistent with the detection light); 3. Visible light components in the ambient light (to reduce crosstalk to the fluorescence signal).

[0039] In this application, the out-of-band cutoff depth of the filter is ≥OD4 (i.e., transmittance ≤0.01%), which can attenuate the detection light reflection by more than 10,000 times, while the transmittance of tissue fluorescence signals in the 440-520nm range is greater than 90%.

[0040] When the paraffin section to be tested floats on the water surface, the optical system faces the following unique problems introduced by the water bath environment: 1. Specular Reflection on Water Surface: When the detection light illuminates the water surface at a certain angle, according to Fresnel's reflection formula, the reflectivity of the air-water interface for unpolarized light at an incident angle of 30°-45° is approximately 2%-5%. The reflected detection light (405nm wavelength) may enter the camera's field of view, causing glare. A possible countermeasure is to select an incident angle of 30°-45°, deflecting the principal direction of the specularly reflected light away from the camera's line of sight (the camera is positioned directly above the water bath), ensuring only a small amount of scattered light enters the camera. Simultaneously, a bandpass filter further filters out the reflected components of the 405nm detection light entering the camera.

[0041] 2. Water absorption: Liquid water has extremely low absorption coefficients (less than 0.001 cm⁻¹) in the 405 nm and 440-520 nm wavelength ranges. - ¹), which is approximately transparent to both excitation light and fluorescence, and the signal attenuation caused by the thickness of the water layer (10-30 mm) in the fluorescence transmission path is negligible (less than 0.3%).

[0042] 3. Refractive index matching: The refractive index n of paraffin is approximately 1.44, while that of water is approximately 1.33. When the slice floats on water, it forms a three-layer medium structure of air (n=1.0), water (n=1.33), and paraffin (n=1.44). The gradual transition of refractive index reduces the high reflection that may occur at the paraffin-air interface to some extent, which is beneficial to the effective emission of fluorescence signals.

[0043] 4. Background Scattering from a Semi-Melted Paraffin Surface in a High-Temperature Water Bath: In a 40-45°C high-temperature water bath, paraffin approaches its softening point. Partial melting and re-solidification of the paraffin on the slice surface creates a relatively smooth surface. This smooth surface scatters less 405nm wavelength detection light, resulting in less stray light entering the camera's field of view and a better fluorescence signal-to-noise ratio than in a cold water scenario. Therefore, in a high-temperature water bath scenario, the fluorescence ratio threshold can be set higher (e.g., 1.65).

[0044] 5. Scattering Background of Rough Solid Paraffin Surface in Low-Temperature Water Bath Scenarios: In a low-temperature water bath of 2-8°C, paraffin is solid, and the surface of the slices formed by cutting is rough, with irregular textures ranging from micrometers to tens of micrometers. When 405nm detection light illuminates this rough surface, in addition to fluorescence emission from the tissue region, a strong diffuse reflection background is also generated. Although this scattered light is significantly attenuated by the bandpass filter, the noise floor in the 440-520nm passband is still slightly higher than that of the smooth surface in a hot water scenario. Possible countermeasures include: in low-temperature water bath scenarios, appropriately increasing the driving current of the blue-violet LED (250-350mA can be selected) to enhance the fluorescence excitation intensity, and using a larger incident angle (50°±5° can be selected) to deflect specular reflection away from the camera's line of sight. Simultaneously, using a lower fluorescence ratio threshold can compensate for the change in effective signal-to-noise ratio caused by the increased scattering background. Also, increasing the exposure time for fluorescence image acquisition in low-temperature water bath scenarios is recommended. In other words, the driving current of the blue-violet LED in the low-temperature water bath scenario is greater than that in the high-temperature water bath scenario; the exposure time for acquiring fluorescence images in the low-temperature water bath scenario is greater than that in the high-temperature water bath scenario; and the angle of the detection light relative to the normal of the water surface in the low-temperature water bath scenario is greater than that in the high-temperature water bath scenario.

[0045] Among them, the fluorescence images of paraffin sections are as follows: Figure 3 As shown.

[0046] S3: Preprocess the fluorescence image to extract the region of interest corresponding to the slice to be detected.

[0047] The image preprocessing includes, in sequence, dark field subtraction, white field correction, noise suppression, adaptive background subtraction, and artifact filtering; Dark field subtraction eliminates the inherent grayscale background caused by camera dark current and stray ambient light, while white field correction corrects systematic grayscale deviations caused by uneven distribution of blue-violet excitation light field and differences in the response of camera pixels. This unifies the grayscale benchmark of the image at the imaging system level, ensuring the comparability of fluorescence grayscale values ​​at different acquisition times and field-of-view locations. It avoids fluctuations in characteristic parameters of the same tissue sample across different workstations and batches due to systematic biases. Simultaneously, it provides an accurate zero-value baseline for subsequent quantitative calculation of fluorescence intensity, preventing the dark background from overwhelming weak fluorescence tissue signals and improving the ability to distinguish samples with low fluorescence intensity.

[0048] Noise suppression is achieved by combining adaptive median filtering with Gaussian low-pass filtering. This combined filtering scheme is designed for the noise characteristics of weak fluorescence images, and the two types of filtering complement each other: adaptive median filtering can dynamically adjust the filtering window according to local gray-level fluctuations, effectively removing pulsed salt-and-pepper noise caused by sensor thermal noise and electromagnetic interference, while avoiding excessive blurring of tissue edges caused by fixed window filtering; Gaussian low-pass filtering can suppress high-frequency random white noise in the image, smooth the background graininess, and reduce the interference of random fluctuations in background gray-level on the calculation of intensity standard deviation. The combination of the two significantly reduces the image noise level while preserving the fluorescence intensity distribution characteristics and edge contours of the tissue region to the greatest extent, avoiding distortion of key parameters such as the mean, maximum, and standard deviation of fluorescence intensity caused by filtering, and improving the reliability of subsequent feature statistics.

[0049] Adaptive background subtraction is used to eliminate water surface ripple interference. Adaptive background subtraction freezes water surface displacement by using a global shutter short exposure. After detection triggering, a dwell time is set, or the water surface stability is judged by inter-frame difference. If the water surface is not stable, it is resampled after a delay. The water surface ring zone inward from the image boundary is extracted as the background region. The low-frequency background components are estimated and subtracted by mean filtering or two-dimensional quadratic polynomial fitting to eliminate low-frequency water surface ripples.

[0050] Specifically, assume the water surface velocity v ≈ 0.1–1 mm / s (still bath). Motion freeze: the exposure t_exp under global shutter speed is included within the LED illumination window t_led, t_led = 50–200 ms, and the displacement within a single frame Δd = v·t_exp < 0.1 mm, corresponding to an image shift < 1.5 px (camera calibration scale approximately 12.3 px / mm). After triggering, wait 100–300 ms for the ripples to decay, or calculate the absolute difference D between the two preview frames: D=Σ|I_t I_{t 1}|, I_t is the grayscale value of a pixel in the image of frame t (the current frame). 1} is the t-th The grayscale value of a pixel at the same coordinate position in one frame (previous frame) of an image; If D > T_motion, it is determined to be an unstable state, and the sample is resampled after a 50ms delay, with a maximum of 3 retries.

[0051] T_motion is the inter-frame differential motion threshold (water surface stability determination threshold).

[0052] Low-frequency fluctuation elimination: The low-frequency component B(x,y) is estimated using the water surface ring outside the slice (the 10% area inward from the image boundary, which is pure background water). The fitted background component B_fit is obtained using a 31×31 mean filter or a two-dimensional quadratic polynomial fitting (degree 2). The corrected image I_corr is: I_corr=I_raw B_fit; Where I_raw represents the original captured image.

[0053] Adaptive background subtraction freezes instantaneous water surface displacement through global short-exposure shutters, avoiding motion blur. Combined with pause-and-wait or inter-frame difference analysis for water surface stability assessment, acquisition is triggered only when the water surface is relatively stable, reducing the amplitude of fluctuation interference at the source and avoiding the acquisition of invalid images with violent fluctuations. By extracting the water surface ring inward from the image boundary as a pure background region (without slice interference), the background components corresponding to low-frequency water surface undulations are estimated and subtracted as a whole through mean filtering or two-dimensional quadratic polynomial fitting. This eliminates global background grayscale deviations caused by water surface curvature and gradual changes in illumination, making the background baseline of the entire image more uniform and smooth. This feature effectively solves the pain point of water surface fluctuations unique to water bathing scenes, avoiding misidentification of water surface reflections and undulations as tissue fluorescence signals, ensuring a unified reference benchmark for fluorescence intensity at different locations within the same image, and significantly improving the stability and repeatability of quantitative fluorescence parameters.

[0054] Artifact filtering is used to remove interference from bubbles and debris in the water. The artifact filtering generates a foreground mask with the background grayscale mean plus 3 times the background standard deviation as the threshold. Morphological opening operation is performed on the foreground mask to remove small debris, and morphological closing operation is performed to fill holes. Connectivity analysis is performed, and connected components with an area not less than the preset minimum area are retained for subsequent extraction of regions of interest. At the same time, bubble regions are identified by combining the region grayscale mean and roundness features, and bubble regions are excluded from the effective regions.

[0055] Specifically, the threshold T_fg of the foreground mask is: T_fg = I_bg + 3·I_bg_std; I_bg is the mean gray level of the background region, and I_bg_std is the standard deviation of the gray level of the background region. Connected component filtering: Mark the 8-neighbor connected components c, and only retain connected components with an area not less than the preset minimum area A_min to participate in the subsequent extraction of regions of interest. A_min is determined by calibration (A_min≥50px≈0.33mm² for high temperature water bath scenarios). Bubble Removal: If the average grayscale value of the current area is less than 0.5 times the average grayscale value of the background area and the roundness is >0.7, it is identified as a bubble and removed from the current area.

[0056] The method for extracting the region of interest is as follows: The median gray level of the boundary ring of the fluorescence image is used as the background level. The foreground threshold is calculated by superimposing the background level by 3 times the background standard deviation or by the Otsu method. A binary mask is generated based on the foreground threshold. After performing morphological opening and closing operations on the binary mask, connected component analysis is performed to retain connected components whose area is within a preset area range. The union, circumscribed rectangle, or convex hull of the connected components is used as the region of interest. The preset area range has a preset minimum area as the lower limit and 60% of the effective area of ​​the fluorescence image field of view as the upper limit.

[0057] When extracting regions of interest (ROIs), the image boundary ring is selected as the background statistical region. This region is naturally unobstructed by the slices and is a pure water surface background, avoiding interference from the fluorescence signal of the slices themselves in the background level statistics and ensuring the purity of the background baseline. The median gray level is used as the background level, which is more resistant to local gray level anomalies caused by sporadic bubbles and debris within the ring than the arithmetic mean. The background estimate is more stable and less susceptible to extreme value shifts. Connected component analysis is used to decompose and mark independent foreground regions in the binary mask, providing a basis for unitized processing in subsequent screening. Using a preset minimum area as the lower limit, artifact connected components such as tiny dust, debris, and microbubbles that are much smaller than normal paraffin slices can be removed in batches, eliminating interference from invalid small targets. Using 60% of the effective area of ​​the field of view as the upper limit, abnormal situations such as large-area water surface reflection, overexposure of the entire image, and missegmentation of the background across the entire field of view can be excluded. This conforms to the actual size range of pathological paraffin slices (normal slice size is much smaller than 60% of the field of view). Filtering abnormal connected components that do not conform to physical common sense from the size dimension significantly reduces the probability of ROI mis-extraction.

[0058] S4: Calculate fluorescence characteristic parameters within the region of interest. The fluorescence characteristic parameters include at least the mean fluorescence intensity, the maximum fluorescence intensity, and the standard deviation of fluorescence intensity. Calculate the fluorescence ratio and fluorescence area ratio based on the mean fluorescence intensity, the maximum fluorescence intensity, and the standard deviation of fluorescence intensity. The fluorescence ratio is the ratio of the maximum fluorescence intensity to the mean fluorescence intensity. The fluorescence area ratio is the ratio of the area of ​​pixels within the region of interest whose grayscale value exceeds the grayscale value threshold T to the total pixel area of ​​the region of interest. The grayscale value threshold T is calculated based on the mean fluorescence intensity and the standard deviation of fluorescence intensity.

[0059] Wherein, the fluorescence intensity mean I_mean is the arithmetic mean of the gray values ​​of all pixels in the region of interest.

[0060] The maximum fluorescence intensity I_max is the pixel value with the highest grayscale value within the region of interest.

[0061] The standard deviation of fluorescence intensity, I_std, is the standard deviation of pixel gray values ​​within the region of interest, reflecting the non-uniformity of fluorescence distribution.

[0062] The fluorescence ratio R_f = I_max / I_mean; this ratio amplifies the local bright features of the fluorescence signal; if the fluorescence distribution in the tissue section is uneven (strong fluorescence in the tissue area and weak fluorescence in the paraffin area), the R_f value is larger; the fluorescence intensity of pure paraffin sections is uniform, and the R_f value is close to 1.0.

[0063] The fluorescence area ratio A_f is the ratio of the area of ​​pixels within the region of interest whose grayscale values ​​exceed a grayscale threshold T to the total area of ​​pixels within the region of interest. The grayscale threshold T is calculated as the product of the mean fluorescence intensity, the standard deviation of fluorescence intensity, and the confidence coefficient. The specific formula is as follows: T = I_mean + k·I_std; Where k is the confidence coefficient, and in this application, the value of k is 3.

[0064] S5: Load the corresponding judgment threshold group based on the current water bath temperature scenario. The judgment threshold group includes the fluorescence ratio threshold and fluorescence area ratio threshold determined based on the current water bath temperature. If the fluorescence ratio and fluorescence area ratio meet the threshold judgment conditions, it is judged as a paraffin section containing tissue; otherwise, it is judged as a pure paraffin section.

[0065] In this step, the threshold determination condition is: the fluorescence ratio R_f is greater than the fluorescence ratio threshold R_threshold and the fluorescence area ratio A_f is greater than the fluorescence area ratio threshold A_threshold.

[0066] For low-temperature water bath scenarios, when the section is determined to be a paraffin section containing tissue, a "refinement complete" signal is output; when the section is determined to be a pure paraffin section, a "continue refinement" signal is output.

[0067] The judgment threshold groups corresponding to the high-temperature water bath temperature scenario and the low-temperature water bath temperature scenario are obtained by calibration within the corresponding water bath temperature range. The specific calibration method is as follows: N1. Sample Collection: For high-temperature water bath scenarios, at least 50 paraffin sections containing different types of tissue and at least 50 pure paraffin sections were collected as calibration sets. For low-temperature water bath scenarios, paraffin blocks containing different types of tissue were collected, and at least 50 paraffin sections containing tissue were collected in both the incomplete and completed trimming states to form calibration sets. Water bath temperature ranges were set for each water bath scenario. All samples were manually labeled with gold standard tags. The gold standard tag for pure paraffin sections and paraffin sections containing tissue in the incomplete trimming state was "pure paraffin", and the gold standard tag for paraffin sections containing tissue and those in the completed trimming state was "containing tissue". N2. Data acquisition: Under the corresponding water bath temperature scenario, the fluorescence images of each sample are acquired using the same detection conditions as in step S1, and the fluorescence ratio and fluorescence area ratio of each sample are calculated. N3. Receiver-Operating Characteristic (ROC) curves were plotted with fluorescence ratio and fluorescence area ratio as discriminant variables and the gold standard label as the response variable, respectively. The value corresponding to the maximum value of the Youden index (sensitivity + specificity - 1) was selected as the threshold for fluorescence ratio and fluorescence area ratio for this scenario. Selecting the value corresponding to the maximum value of the Youden index as the threshold essentially achieves an optimal balance between sensitivity (true positive rate, corresponding to avoiding false negatives) and specificity (true negative rate, corresponding to avoiding false positives). This method avoids both excessively pursuing a low false negative rate leading to a large number of pure paraffin sections being misclassified as containing tissue and excessively pursuing a low false positive rate leading to the underdetection of tissue-containing sections.

[0068] After obtaining the fluorescence ratio threshold and the fluorescence area ratio threshold, the generalization performance of the threshold is verified using an independent validation set (no less than 40 samples per scenario), requiring both sensitivity and specificity to be greater than or equal to 95%.

[0069] Calibration result reference values: In the low temperature water bath scenario, R_threshold=1.55, A_threshold=4%; in the high temperature water bath scenario, R_threshold=1.65, A_threshold=5%.

[0070] A water bath detection device for pathological tissue sections based on the difference in blue-violet fluorescence includes: The blue-violet light illumination module is used to emit blue-violet light with a center wavelength of 380-420nm as the detection light, and to irradiate the paraffin slice to be tested on the surface of the water bath at an incident angle of 20°-60° relative to the normal of the water surface. An imaging acquisition module is used to acquire fluorescence images; the imaging acquisition module includes a camera, an imaging lens and a bandpass filter; the bandpass filter is disposed in the imaging optical path to selectively transmit tissue autofluorescence and block stray or reflected light from the excitation. The image processing module is used to perform preprocessing, region of interest extraction, and multi-dimensional fluorescence feature calculation on the acquired fluorescence images; the fluorescence features include at least the mean fluorescence intensity, the maximum fluorescence intensity, the standard deviation of fluorescence intensity, the fluorescence ratio, and the fluorescence area ratio. The judgment output module has built-in judgment threshold groups corresponding to different water bath temperature scenarios. It is used to compare the calculated fluorescence ratio and fluorescence area ratio with the corresponding thresholds and output the judgment result.

[0071] The image processing module is also equipped with a water surface ripple suppression unit and an artifact filtering unit. The water surface ripple suppression unit freezes water surface movement through global shutter short exposure and judges water surface stability through inter-frame difference. If the water surface is unstable, it performs delayed resampling. The artifact filtering unit removes artifacts caused by tiny tissue debris and air bubbles through morphological operations and connected component analysis.

[0072] Example 1: Water bath fluorescence detection of routine HE-stained gastric tissue sections in a high-temperature water bath (42°C) setting.

[0073] (1) Samples: One hundred paraffin-embedded gastric mucosal tissue blocks (25mm×25mm) provided by the pathology department of a tertiary hospital were selected. After standard trimming, the blocks were sectioned (4μm thickness) and transferred to a 42°C constant temperature water bath to float on the water surface. Among them, 60 tissue-containing sections and 50 pure paraffin sections produced at the pre-trimming stage were used as the test set. All samples were independently labeled with the gold standard label of "containing tissue" or "pure paraffin" by two senior technicians.

[0074] (2) System Configuration: The blue-violet LED array uses eight 405nm LEDs (1W power per LED, center wavelength 405nm±5nm), symmetrically distributed at a 45° incident angle on the four sides of the water bath (two LEDs on each side). The LED driving current is 200mA, the PWM frequency is 2kHz, and the illumination time is 50ms. The bandpass filter has a center wavelength of 480nm, a full width at half maximum (FWHM) of 80nm (440-520nm), and an out-of-band cutoff depth of OD4 (measured cutoff depth > OD5 in the 450-410nm wavelength range). The industrial camera uses a Sony IMX253 chip (12.3 megapixels, 3.45μm pixel size, 14-bit dynamic range, global shutter), with a lens working distance of 180mm and a field of view covering the effective area of ​​the water bath (approximately 100mm×100mm). The camera exposure time is 50ms, and the gain is 1×. The image processing module is an ARM Cortex-A72 embedded processor running a Linux real-time operating system. Ambient light is controlled to <50 lux by the light shield on the equipment housing.

[0075] (3) Calibration: 60 slides (50 tissue slides + 50 pure paraffin slides) were randomly selected as the calibration set. Fluorescence images were acquired for each slide in the calibration set, and R_f and A_f values ​​were extracted. ROC analysis was used to determine the optimal thresholds: R_threshold = 1.65, A_threshold = 5%.

[0076] (4) Operating steps: The slice falls into a 42°C water bath, directly triggering the detection. The PLC sends a trigger signal; The blue-violet LED array is lit for 50ms, and the camera simultaneously exposes for 50ms to acquire fluorescence images (440-520nm channel); the images are then transmitted to the embedded processor (USB 3.0, taking about 15ms). Image preprocessing (dark field subtraction, 5ms) + ROI extraction (threshold segmentation, 8ms) + fluorescence feature calculation (I_mean, I_max, I_std, R_f, A_f) + threshold comparison and judgment; Output a judgment signal. The total time for a single detection is approximately 95ms.

[0077] Example 2: Fluorescence detection of tissue sections of different tissue types in a high-temperature water bath (42°C).

[0078] (1) Samples: 100 paraffin sections (pathological tissues including liver, kidney, breast, lymph nodes, and colonic mucosa) were selected, 20 sections of each of the 5 common pathological tissue types. Each tissue type was sectioned after trimming (section thickness 4μm). The first 3 sections of each paraffin block (60 sections in total) were taken for testing. Ten pure paraffin sections (sections before trimming) of each tissue type were prepared simultaneously as controls.

[0079] (2) System configuration: Same as in Example 1. Additional test of the effect under different LED driving current conditions: For each tissue type, the driving current was tested at 100mA, 200mA and 300mA respectively.

[0080] (3) Detection results (taking a 480nm bandpass filter with a driving current of 200mA as an example): The mean fluorescence characteristics of sections from each tissue type (sample size 60) are as follows: liver: I_mean=3120±580, I_max=8950±1120, R_f=3.4±0.6, A_f=14.8%±3.5%; kidney: I_mean=2780±520, I_max=7930±980, R_f=3.1±0.5, A_f=13.2%±3.1%; Breast: I_mean=2410±480, I_max=6520±860, R_f=2.7±0.5, A_f=10.8%±2.8%; Lymph nodes: I_mean=1980±420, I_max=4860±680, R_f=2.5±0.4, A_f=9.5%±2.5%; Colonic mucosa: I_mean=2250±450, I_max=5980±780, R_f=2.7±0.4, A_f=11.2%±2.9%; Mean fluorescence characteristics of pure paraffin slides of various tissue types (sample size 10): I_mean=405-418, R_f=1.04-1.09, A_f=0.8%-1.5%.

[0081] The effect of different driving currents on the R_f value of various tissue types: the average R_f value is 2.4-3.0 at 100mA, 2.5-3.4 at 200mA, and 2.6-3.5 at 300mA. The increase in R_f from 200mA to 300mA is less than 5%, but the blue-violet LED generates more heat; therefore, 200mA is recommended as the standard driving current.

[0082] (4) Analysis: The R_f values ​​of the above five tissue types were all >2.0 (range 2.5-3.4) in tissue sections, while the R_f values ​​of pure paraffin sections were all <1.1. Using the R_threshold=1.6 obtained in Example 1, R_threshold=1.65 can effectively cover all five tissue types. The A_f values ​​of tissue sections were all >5% (range 9.5%-14.8%), while the A_f values ​​of pure paraffin sections were all <2%. Using A_threshold=5% can effectively distinguish them.

[0083] Overall accuracy rates: Liver 98.3%, Kidney 97.5%, Breast 96.7%, Lymph Nodes 95.8%, Colon 97.5%. The relatively low fluorescence intensity of lymph nodes is related to the uniformity of fluorescence in the sections during water bath (lymph node tissue is dense, and the fluorescence signal is relatively concentrated but covers a small area).

[0084] (5) Conclusion: The accuracy of this method in detecting tissue in water bath sections of common pathological tissue types such as liver, kidney, breast, lymph nodes and colonic mucosa is ≥95%. R_threshold=1.65 can be used as a uniform judgment threshold, without the need to set it separately according to tissue type.

[0085] This invention is not limited to the preferred embodiments described above. Anyone can derive other products in various forms under the guidance of this invention. However, regardless of any changes in shape or structure, any technical solution that is the same as or similar to this application falls within the protection scope of this invention.

Claims

1. A water bath detection method for pathological tissue sections based on the difference in blue-violet fluorescence, characterized in that, The specific steps include the following: S1: Blue-violet light within a preset center wavelength range is used as the detection light, and the detection light is irradiated at a preset angle relative to the water surface normal onto the paraffin slice to be tested floating on the water surface of the water bath; the center wavelength of the detection light is 380-420nm, and the angle of the detection light relative to the water surface normal is 20°-60°. S2: The autofluorescence signal generated by the blue-violet light excitation of the paraffin slice to be tested is collected in the imaging optical path with a bandpass filter to generate a fluorescence image; S3: Preprocess the fluorescence image to extract the region of interest corresponding to the slice to be detected; the image preprocessing includes dark field subtraction, white field correction, noise suppression, adaptive background subtraction, and artifact filtering in sequence; Noise suppression is achieved by combining adaptive median filtering with Gaussian low-pass filtering. Adaptive background subtraction is used to eliminate water surface ripple interference. Adaptive background subtraction freezes water surface displacement by using a global shutter short exposure. After detection triggering, a dwell time is set, or the water surface stability is judged by inter-frame difference. If the water surface is not stable, it is resampled after a delay. The water surface ring zone inward from the image boundary is extracted as the background region. The low-frequency background components are estimated and subtracted by mean filtering or two-dimensional quadratic polynomial fitting to eliminate low-frequency water surface ripples. Artifact filtering is used to remove interference from bubbles and debris in the water. The artifact filtering generates a foreground mask with the background grayscale mean plus 3 times the background standard deviation as the threshold. Morphological opening operation is performed on the foreground mask to remove small debris, and morphological closing operation is performed to fill holes. Connectivity analysis is performed, and connected components with an area not less than the preset minimum area are retained for subsequent extraction of regions of interest. At the same time, bubble regions are identified by combining the region grayscale mean and roundness features, and bubble regions are excluded from the effective regions. S4: Calculate fluorescence characteristic parameters within the region of interest. These parameters include at least the mean fluorescence intensity, the maximum fluorescence intensity, and the standard deviation of fluorescence intensity. Calculate the fluorescence ratio and fluorescence area ratio based on these parameters. The fluorescence ratio is the ratio of the maximum fluorescence intensity to the mean fluorescence intensity. The fluorescence area ratio is the ratio of the area of ​​pixels within the region of interest whose grayscale value exceeds a grayscale value threshold T to the total area of ​​pixels within the region of interest. The grayscale value threshold T is calculated based on the mean fluorescence intensity and the standard deviation of fluorescence intensity. S5: Based on the current water bath temperature scenario, a corresponding threshold group is loaded. The threshold group includes a fluorescence ratio threshold and a fluorescence area ratio threshold determined based on the current water bath temperature. If the fluorescence ratio and fluorescence area ratio meet the threshold determination conditions, the section is determined to contain tissue in paraffin; otherwise, it is determined to be a pure paraffin section. The gray value threshold T is the product of the mean fluorescence intensity plus the standard deviation of fluorescence intensity and the confidence coefficient. The threshold determination conditions are: the fluorescence ratio is greater than the fluorescence ratio threshold and the fluorescence area ratio is greater than the fluorescence area ratio threshold. The water bath temperature scenarios include a high-temperature water bath scenario of 40-45°C and a low-temperature water bath scenario of 2-8°C. For the low-temperature water bath scenario, when the section is determined to contain tissue paraffin, a "refinement complete" signal is output; when the section is determined to be pure paraffin, a "continue refinement" signal is output.

2. The method for water bath detection of pathological tissue sections based on blue-violet fluorescence difference according to claim 1, characterized in that, In step S5, the judgment threshold groups corresponding to the high-temperature water bath temperature scenario and the low-temperature water bath temperature scenario are obtained by calibration within the corresponding water bath temperature range. The specific calibration method is as follows: N1. Sample collection: For high-temperature water bath scenarios, collect no less than 50 tissue-containing paraffin sections with different types of tissues and no less than 50 pure paraffin sections as calibration sets; For low-temperature water bath scenarios, paraffin blocks containing different tissues were collected. At least 50 paraffin sections containing tissues were collected in both the incomplete and completed trimming states to form a calibration set. Water bath temperature ranges were set for the corresponding water bath temperature scenarios. All samples were manually labeled with gold standard tags. The gold standard tag for pure paraffin sections and paraffin sections containing tissues in the incomplete trimming state was "pure paraffin", and the gold standard tag for paraffin sections containing tissues and paraffin sections in the completed trimming state was "containing tissues". N2. Data acquisition: Under the corresponding water bath temperature scenario, the fluorescence images of each sample are acquired using the same detection conditions as in step S1, and the fluorescence ratio and fluorescence area ratio of each sample are calculated. N3. ROC curves were plotted with fluorescence ratio and fluorescence area ratio as discriminant variables and gold standard label as response variable, respectively. The value corresponding to the maximum value of Youden index was selected as the fluorescence ratio threshold and fluorescence area ratio threshold for this scene.

3. The method for water bath detection of pathological tissue sections based on the difference in blue-violet fluorescence according to claim 1, characterized in that, In step S3, the method for extracting the region of interest is as follows: The median gray level of the boundary ring of the fluorescence image is used as the background level. The foreground threshold is calculated by superimposing the background level by 3 times the background standard deviation or by the Otsu method. A binary mask is generated based on the foreground threshold. After performing morphological opening and closing operations on the binary mask, connected component analysis is performed to retain connected components whose area is within a preset area range. The union, circumscribed rectangle, or convex hull of the connected components is taken as the region of interest. The preset area range is limited by a preset minimum area and is limited by 60% of the effective area of ​​the fluorescence image field of view.

4. The method for water bath detection of pathological tissue sections based on blue-violet fluorescence difference according to claim 1, characterized in that, In step S2, the center wavelength of the bandpass filter is 440-520nm, the bandwidth is 40-80nm, and the out-of-band cutoff depth is ≥OD4, so as to block specular reflection and scattered stray light of the excitation light.

5. The method for water bath detection of pathological tissue sections based on blue-violet fluorescence difference according to claim 1, characterized in that, The detection light is generated by driving a blue-violet LED array. The driving current of the blue-violet LED in the low-temperature water bath temperature scenario is greater than that in the high-temperature water bath temperature scenario. The fluorescence image acquisition exposure time in the low-temperature water bath temperature scenario is greater than that in the high-temperature water bath temperature scenario. In a low-temperature water bath scenario, the angle of the detection light relative to the water surface normal is greater than that in a high-temperature water bath scenario.

6. A water bath detection device for pathological tissue sections based on blue-violet fluorescence difference, used to implement the water bath detection method for pathological tissue sections based on blue-violet fluorescence difference as described in any one of claims 1-5, characterized in that, include: The blue-violet light illumination module is used to emit blue-violet light with a center wavelength of 380-420nm as the detection light, and to irradiate the paraffin slice to be tested on the surface of the water bath at an incident angle of 20°-60° relative to the normal of the water surface. An imaging acquisition module is used to acquire fluorescence images; the imaging acquisition module includes a camera, an imaging lens and a bandpass filter; the bandpass filter is disposed in the imaging optical path to selectively transmit tissue autofluorescence and block stray or reflected light from the excitation. The image processing module is used to perform preprocessing, region of interest extraction, and multi-dimensional fluorescence feature calculation on the acquired fluorescence images. The fluorescence features include at least the mean fluorescence intensity, maximum fluorescence intensity, standard deviation of fluorescence intensity, fluorescence ratio, and fluorescence area ratio. The image processing module is also equipped with a water surface ripple suppression unit and an artifact filtering unit: the water surface ripple suppression unit freezes the water surface movement by global shutter short exposure and judges the water surface stability by inter-frame difference. If the water surface is unstable, it performs delayed resampling. The artifact filtering unit removes artifacts caused by tiny tissue debris and air bubbles through morphological operations and connected component analysis. The judgment output module has built-in judgment threshold groups corresponding to different water bath temperature scenarios. It is used to compare the calculated fluorescence ratio and fluorescence area ratio with the corresponding thresholds and output the judgment result.

Images