Terahertz near-field imaging rapid unmarked detection method for aspergillus conidia

Abstract

The invention discloses a method for rapidly detecting aspergillus conidia in a label-free manner through terahertz near-field imaging, and belongs to the technical field of microbiological detection. The method comprises the following steps: synchronously acquiring atomic force microscope morphology images and terahertz near-field signals of spores through a scattering type terahertz time-domain spectrum near-field scanning system, and jointly extracting morphology features, multi-order harmonic signal intensity and a local near-field amplitude spectrum peak value after phase-locked amplification and demodulation processing; and combining into a feature set for spore type identification. The method provided by the invention solves the problem that high-resolution, label-free and rapid detection cannot be realized at the same time in the prior art, and is mainly used for rapid label-free accurate detection and type identification of the aspergillus conidia.

Description

Technical Field

[0001] This invention relates to the field of microbial detection technology. More specifically, this invention relates to a method for rapid, label-free detection of Aspergillus conidia using terahertz near-field imaging. Background Technology

[0002] Aspergillus conidia, as common environmental fungal spores, have significant impacts in the medical, food, and agricultural fields, and their rapid and accurate identification is crucial for disease diagnosis and contamination control. However, existing spore detection methods have many limitations in practical applications, particularly in terms of speed, label-free characteristics, and ability to differentiate between different types.

[0003] Traditional detection methods primarily rely on optical microscopy and microbial culture techniques. While optical microscopy is easy to operate, its spatial resolution is limited to the micrometer level due to diffraction limits, making it difficult to discern the nanoscale morphological details and internal structural features of spores. Furthermore, relying solely on morphological characteristics for type differentiation is highly subjective and prone to misjudgment due to overlaps in size and shape between different Aspergillus conidia. While microbial culture methods can provide relatively reliable type information, they are time-consuming, typically requiring several days to weeks, failing to meet the needs of real-time or rapid detection, and may lead to missed detections due to differences in spore viability during the culture process.

[0004] With the development of molecular biology techniques, nucleic acid amplification-based detection methods, such as PCR, have been applied to spore identification, improving detection specificity. However, these methods typically require complex sample pretreatment, including DNA extraction and purification steps, and rely on fluorescent labeling or probe hybridization. This is not only cumbersome and costly, but the introduction of markers may also alter the original state of the spores, failing to achieve truly interference-free detection. Furthermore, these methods struggle to provide physical and structural information about the spores themselves, limiting their application in morphological and compositional correlation analysis.

[0005] Spectroscopic techniques such as Fourier transform infrared spectroscopy or Raman spectroscopy have been explored for microbial detection, providing information on chemical composition. However, when detecting individual spores, they often face problems such as weak signals and insufficient spatial resolution. Due to the small size of spores (usually less than a few micrometers), far-field spectroscopy is constrained by the diffraction limit and cannot effectively capture the local dielectric properties or nanoscale inhomogeneities on and near the spore surface, resulting in blurred or overlapping spectral features and making it difficult to distinguish closely related Aspergillus species.

[0006] In summary, the limitations of existing spore detection methods mean that they are still significantly insufficient in rapidly, label-free, and accurately identifying Aspergillus conidia types. Summary of the Invention

[0007] One object of the present invention is to provide a method for rapid label-free detection of Aspergillus conidia using terahertz near-field imaging, so as to at least solve the above-mentioned problems.

[0008] To achieve the objectives and other advantages of this invention, a method for rapid label-free detection of Aspergillus conidia using terahertz near-field imaging is provided, comprising: Step 1, preparing Aspergillus conidia into a spore suspension and dropping it onto a low-resistivity silicon substrate, drying it, and then fixing it on the sample stage of a scattering-type terahertz time-domain spectroscopy near-field scanning system; Step 2, using the scattering-type terahertz time-domain spectroscopy near-field scanning system to scan a single spore fixed on the low-resistivity silicon substrate, simultaneously acquiring an atomic force microscope morphology image of the spore and a terahertz near-field time-domain signal for each pixel; Step 3, performing phase-locked amplification and demodulation processing on the terahertz near-field time-domain signal to extract the first-order harmonic near-field image, the second-order harmonic near-field image, and the third-order harmonic near-field image; extracting the local near-field amplitude spectrum at the spore center position from the terahertz near-field time-domain signal; Step 4, extracting the local near-field amplitude spectrum from the atomic force microscope morphology image... The morphological features of each spore are extracted from the image, including height values ​​ranging from 200 to 1500 nm. First-order harmonic signal intensity values ​​ranging from 0.20 to 2.00 au are extracted from the first-order harmonic near-field image, second-order harmonic signal intensity values ​​ranging from 0.00 to 0.80 au are extracted from the second-order harmonic near-field image, and third-order harmonic signal intensity values ​​ranging from 0.00 to 0.50 au are extracted from the third-order harmonic near-field image. Peak amplitude values ​​ranging from 1.25 to 2.50 au are extracted from the local near-field amplitude spectrum. Step five involves combining the height value, first-order harmonic signal intensity value, second-order harmonic signal intensity value, third-order harmonic signal intensity value, and peak amplitude value of each spore into a feature set. By comparing the differences in various parameters of Aspergillus conidia in the feature set, the type of Aspergillus conidia can be identified.

[0009] Preferably, the scattering-type terahertz time-domain spectroscopy near-field scanning system integrates an atomic force microscope module and a terahertz time-domain spectroscopy module; the terahertz time-domain spectroscopy module uses a femtosecond laser as the light source, with a working repetition frequency of 100 MHz, an average output power of 100 mW, a pulse width of 45 fs, and generates terahertz radiation in the frequency range of 0.2-2.0 THz; the atomic force microscope module operates in knocking mode, with a microcantilever beam oscillation frequency of 50 kHz, a probe tip radius of 20 nm, a knocking amplitude of approximately 150 nm, and a cantilever beam length of 285 μm.

[0010] Preferably, a scattering-type terahertz time-domain spectroscopy near-field scanning system is used to scan individual spores fixed on a low-resistivity silicon substrate. Specifically, this includes: first, performing a preliminary scan in a 50×50μm area to locate the spores, and then performing a fine scan in a 10×10μm or smaller area.

[0011] Preferably, the terahertz near-field time-domain signal is subjected to phase-locked amplification and demodulation processing, specifically including: performing phase-locked demodulation on the terahertz near-field time-domain signal of each pixel, using the oscillation frequency of the micro-cantilever beam and its integer multiples of harmonics as reference signals, and extracting the nth-order near-field signal component. Sn(τ) Where n is the harmonic order, including first, second, and third order. τ The optical delay time is used to separate the modulation signal related to spore morphology and internal structure; the nth order near-field signal component of each extracted pixel is... Sn(τ) Perform a Fast Fourier Transform to convert the time-domain signal into a frequency-domain signal, and obtain the near-field spectrum. Sn(w) ,in, w This is a terahertz frequency, and this conversion allows the signal to be used to analyze the spectral characteristics of the terahertz band; based on the near-field spectrum. Sn(w) Amplitude information is extracted for each harmonic order n. For the first harmonic, the near-field spectrum is first normalized to eliminate the influence of system response and floor. The normalization formula is: E 1,norm ( w )= E 1,sam ( w ) / E 1,ref ( w ),in, E 1,sam ( w () is the first harmonic spectrum of the spore. E 1,ref ( w ) is the first harmonic spectrum of the reference substrate, and then from E 1,norm ( w The amplitude s1 is extracted, while for higher-order harmonics, it is directly extracted from... Sn(w) Extracting amplitude s n This allows us to obtain quantized signal strength data; the amplitude s of each harmonic order n is then extracted. n The pixels are combined into a two-dimensional matrix to form a terahertz near-field image, including first-order, second-order, and third-order harmonic near-field images. The first-order harmonic near-field image reflects the overall morphology of the spore and the near-field enhancement effect, while the second-order and third-order harmonic near-field images show stronger sensitivity to subtle changes in the near-surface dielectric properties and nanoscale compositional inhomogeneities within the spore.

[0012] Preferably, step five specifically includes: constructing a multidimensional feature space, where the feature set of each spore corresponds to a data point in the space; based on a predefined Aspergillus conidia type reference database, which stores typical range values ​​of various parameters for different Aspergillus conidia types in the feature set; and by calculating the Euclidean distance between the feature set of the spore to be tested and the typical range values ​​of various parameters in the reference database, performing parameter matching and classification decisions to identify the specific type of the spore to be tested.

[0013] Preferably, parameter matching and classification decisions are performed by calculating the Euclidean distance between the feature set of the spores to be tested and the typical range values ​​of each parameter in the reference database. Specifically, a weighted Euclidean distance algorithm is used to assign a weight coefficient based on its classification ability to each parameter in the feature set. The weighted Euclidean distance between the spores to be tested and the center value of the typical range value of each type of spore in the reference database is calculated, and the spores to be tested are classified into the category with the smallest weighted Euclidean distance. The weight coefficient is adaptively determined by analyzing the distribution dispersion of various types of spores on different parameters in historical data. For parameters that show greater differences and more concentrated distribution among different types of spores, a higher weight coefficient is assigned.

[0014] Preferably, the method for determining the weighting coefficients includes: collecting and storing feature set data of multiple samples of known types of Aspergillus conidia to form a historical database, wherein the feature set of each sample includes five parameters: height value, first harmonic signal intensity value, second harmonic signal intensity value, third harmonic signal intensity value, and amplitude peak value; for each parameter in the feature set, calculating the Fisher discriminant ratio of that parameter in distinguishing different spore types, wherein the Fisher discriminant ratio is calculated as the ratio of the variance of the parameter among different spore types to the sum of the variances of the parameter within the same spore type, the higher the ratio, the stronger the ability of the parameter to distinguish different spore types; and normalizing the Fisher discriminant ratios of all five parameters to obtain the weighting coefficients of the five parameters.

[0015] Preferably, the Aspergillus conidia include Aspergillus niger spores, Aspergillus flavus spores, and Aspergillus terreus spores.

[0016] Preferably, according to the reference database, the height value of *Aspergillus niger* spores is 380-420 nm, and the intensity values ​​of the first to third harmonic signals are 0.87-1.98 au, 0-0.679 au, and 0-0.43 au, respectively, with a peak amplitude of 1.40-1.50 au in the frequency range of 1.5-2.0 THz; the height value of *Aspergillus flavus* spores is 610-650 nm, and the intensity values ​​of the first to third harmonic signals are 0.73-1.89 au, 0-0.671 au, and 0-0.37 au, respectively, with a peak amplitude of 2.32-2.42 au in the frequency range of 1.5-2.0 THz; the height value of *Aspergillus terreus* spores is 1080-1120 nm, and the intensity values ​​of the first to third harmonic signals are 0.32-1.82 au, 0-0.5 au, and 0-0.33 au, respectively, with a peak amplitude of 1.40-1.50 au in the frequency range of 1.5-2.0 THz. The peak amplitude in the THz frequency range is 1.85-1.95au.

[0017] The present invention has at least the following beneficial effects:

[0018] First, by integrating atomic force microscopy morphological imaging with terahertz near-field spectroscopy, multimodal, label-free detection of individual Aspergillus conidia was achieved. Furthermore, nanometer-precise physical morphology information and terahertz spectral information reflecting internal dielectric properties could be simultaneously acquired from the same spore sample, thus constructing a comprehensive feature set including height, multi-harmonic signal intensity, and local spectral peaks. This multi-parameter fusion analysis strategy overcomes the limitations of traditional single detection methods (such as relying solely on morphological or chemical markers), significantly enhancing the ability to distinguish between different spore types. Since the entire process requires no fluorescent labeling or complex sample pretreatment, it not only preserves the original state of the spores and avoids interference that may be introduced by markers, but also greatly simplifies the operation steps, providing a reliable technical foundation for achieving rapid and accurate spore type identification.

[0019] Secondly, by specifically defining the key components and parameters of the scattering-type terahertz time-domain spectroscopy near-field scanning system, hardware guarantees were provided for achieving high signal-to-noise ratio and high spatial resolution detection. The integrated atomic force microscopy module ensured the precise acquisition of nanoscale morphology, while its tapping mode and probe with specific parameters effectively reduced damage to fragile spore samples. The terahertz time-domain spectroscopy module, employing a femtosecond laser and limiting its operating parameters, could generate broadband, short-pulse terahertz radiation, which is crucial for exciting rich spectral responses in spores and acquiring high-quality time-domain signals. The optimized combination of system parameters enabled the terahertz signal to be localized and enhanced by the probe tip, thereby breaking the diffraction limit and sensitively detecting subtle changes in dielectric properties on and near the spore surface, laying a solid technical foundation for subsequent accurate extraction of feature information.

[0020] Third, the two-step scanning strategy of "preliminary scanning for localization followed by detailed scanning for analysis" effectively improves overall detection efficiency while ensuring detection accuracy and detailed information acquisition. The initial large-area rapid scan quickly determines the approximate location of spores on the substrate, avoiding the time wasted on unnecessary detailed scanning in areas without samples. Subsequently, a high-resolution detailed scan is performed within the located micro-region, concentrating resources to acquire the most crucial morphological and terahertz signal details of the spores. This strategy cleverly balances the trade-off between scanning range and resolution, ensuring the spatial accuracy required for single-spore level analysis while preventing excessively long detection times due to global high-resolution scanning. This makes the method more suitable for practical applications requiring rapid screening of multiple spores.

[0021] Fourth, by employing specific signal processing procedures such as phase-locked amplification, demodulation, and spectral analysis on terahertz near-field time-domain signals, it is possible to effectively separate and quantify the characteristic information of different physical origins from complex raw signals. Phase-locked demodulation, using the harmonics of the probe oscillation frequency as a reference, can greatly suppress environmental noise and extract weak signals closely related to the interaction between spores and probes. Harmonic signals of different orders are extracted and images are formed. The first-order harmonic image can reliably reflect the overall morphology of the spore and the near-field enhancement effect, while higher-order harmonic (second- and third-order) images show higher sensitivity to local characteristics such as the dielectric constant gradient and nanoscale compositional inhomogeneities in the near-surface region of the spore. This hierarchical information extraction enriches the analytical dimensions and provides the possibility of revealing more subtle internal structural differences between different types of spores.

[0022] Fifth, by constructing a multi-dimensional feature space and introducing a predefined reference database, the spore identification process is transformed from subjective judgment to objective, quantitative data matching. By establishing a standardized classification framework, the type of the spore to be tested can be determined by calculating its Euclidean distance to typical values ​​of known types in the database. This method reduces the risk of misjudgment due to differences in operator experience and improves the reliability and repeatability of the identification results. Utilizing multi-dimensional features rather than a single parameter for comprehensive evaluation fully leverages the complementarity between different feature parameters. Even if some parameters partially overlap between different spore types, multi-parameter joint decision-making can effectively improve the accuracy and robustness of classification, laying the foundation for automated spore classification and identification.

[0023] Sixth, the use of a weighted Euclidean distance algorithm for parameter matching significantly improves the scientific rigor and accuracy of classification decisions. By assigning adaptive weights based on the classification ability of each parameter, this algorithm amplifies the influence of key parameters with high discriminative power and low dispersion among different spore types, while weakening the influence of parameters with weaker discriminative power. This data-driven approach makes the classification model more flexible and intelligent, better adapting to the inherent differences in feature performance among different spore types, thereby achieving more accurate and reliable type identification in complex real-world data and effectively reducing misclassification.

[0024] Seventh, the Fisher discriminant ratio can quantitatively evaluate the performance of each feature parameter in terms of "between-class differences" and "within-class differences." A higher ratio indicates that the parameter can more clearly separate different categories. Calculating this ratio using historical data and normalizing it to weights ensures that weight allocation is entirely determined by the inherent patterns of the data itself, avoiding subjective assumptions. Classification models built using the Fisher discriminant ratio-based weight coefficient determination method have strong statistical significance and persuasiveness. Furthermore, they can continuously optimize and adaptively adjust weights as the historical database expands, enabling the entire recognition system to continuously learn and improve, thus maintaining high classification performance over the long term.

[0025] Eighth, the application targets are specified as conidia of Aspergillus niger, Aspergillus flavus, and Aspergillus terreus. These Aspergillus species are important representatives in environmental, industrial, and medical fields, and their rapid identification is urgently needed. By limiting the specific types, the proposed multi-parameter feature set extraction and analysis method can be optimized and validated for the biological characteristics of these specific spores.

[0026] Ninth, by providing clear and typical numerical ranges for various characteristic parameters of three specific Aspergillus spores in the reference database, a direct and operable criterion for rapid and accurate classification and identification is provided. These experimentally verified parameter ranges (such as height, intensity of each harmonic signal, and peak amplitude) constitute the "characteristic fingerprint" of each spore, allowing users to intuitively and efficiently compare their measured values ​​with these established ranges when classifying the spores to be tested. This greatly simplifies the classification decision-making process, improves the identification speed, and ensures the consistency of results between different operators or different batches of experiments.

[0027] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the scattering-type terahertz time-domain spectroscopy near-field scanning system used in this invention; Figure 2 These are AFM morphology images and THz near-field images of three Aspergillus spores according to an embodiment of the present invention; Figure 3 This is an AFM morphology image and cross-sectional analysis diagram of three Aspergillus spores according to an embodiment of the present invention; Figure 4 This is a schematic diagram of THz near-field images and cross-sectional profile analysis of three Aspergillus spores according to an embodiment of the present invention; Figure 5 This is a near-field image of Aspergillus niger spores at levels 1-4 THz and the corresponding near-field intensity profile, according to an embodiment of the present invention. Figure 6 This is an amplitude curve of the spore center of three Aspergillus spores in an embodiment of the present invention within a frequency range of 1.5-2.0 THz. Detailed Implementation

[0029] The present invention will now be described in further detail with reference to the embodiments and accompanying drawings, so that those skilled in the art can implement it based on the description.

[0030] It should be understood that terms such as “having,” “comprising,” and “including” as used herein do not exclude the presence or addition of one or more other elements or combinations thereof.

[0031] It should be noted that, unless otherwise specified, the experimental methods described in the following implementation plan are all conventional methods, and the reagents and materials described are all commercially available unless otherwise specified.

[0032] In one embodiment of the present invention, a method for rapid label-free detection of Aspergillus conidia using terahertz near-field imaging includes four aspects: sample preparation, scanning and signal acquisition, signal processing and feature extraction, and type identification.

[0033] In the preparation and fixation of Aspergillus conidia samples, the Aspergillus conidia used can be obtained from samples isolated from the natural environment or from standardized laboratory culture systems. The suspension can be prepared using sterile water, physiological saline, or phosphate buffer, with a concentration sufficient to ensure uniform dispersion of individual spores without aggregation. A common low-resistivity silica substrate is selected. The volume of the added spore suspension is flexibly adjusted according to the substrate size and spore concentration. After addition, the substrate can be treated by natural air drying, vacuum drying, or low-temperature drying. After drying, the substrate is fixed on the sample stage of a scattering terahertz time-domain spectroscopy near-field scanning system using physical adsorption or non-destructive fixation methods to ensure stable sample position and prevent shifting during scanning.

[0034] During the scanning process, the scattering-type terahertz time-domain spectroscopy near-field scanning system must ensure the synchronous and coordinated operation of the atomic force microscopy module and the terahertz time-domain spectroscopy module. Before scanning, the height and position of the sample stage should be adjusted to ensure that the target spore is in the center of the scanning field of view. The scanning parameters can be fine-tuned according to the actual size of the spore to ensure that the scanning range can completely cover a single spore. The atomic force microscopy morphology image and the terahertz near-field time-domain signal of each pixel are acquired simultaneously. During the acquisition process, environmental interference factors are controlled to ensure the clarity and stability of the image and signal, and to avoid the influence of external noise on the data.

[0035] In terahertz near-field time-domain signal processing, the lock-in amplification stage can select an appropriate amplification factor based on the signal strength to improve the signal-to-noise ratio. The demodulation process uses the oscillation frequency of the microcantilever beam and its integer harmonics as reference signals to accurately separate the modulation signal related to the spore, thereby extracting first-, second-, and third-order harmonic near-field images. Abnormal interference signal points are removed during extraction to ensure image quality. Extraction of local near-field amplitude spectra requires precise location of the spore center to avoid the intrusion of background signals. The frequency range of the extracted spectrum must cover the terahertz band that reflects the characteristic response of the spore to ensure effective capture of amplitude peaks.

[0036] In the feature parameter extraction stage, height values ​​were measured at key locations such as the apex and middle of the spores from atomic force microscopy (AFM) images to ensure representativeness. Valid height values ​​were recorded within the range of 200-1500 nm. Corresponding signal intensity values ​​were extracted from first-, second-, and third-order harmonic near-field images. Valid values ​​were selected within the ranges of 0.20-2.00 au for first-order harmonics, 0.00-0.80 au for second-order, and 0.00-0.50 au for third-order. Amplitude peak values ​​within the range of 1.25-2.50 au were extracted from local near-field amplitude spectra. After combining these parameters to form a feature set, a reasonable parameter comparison standard was established. By comparing the inherent differences in various parameters among different Aspergillus conidia, distinguishable feature parameters were selected, ultimately achieving type identification of Aspergillus conidia.

[0037] In this embodiment, there is no need to label Aspergillus conidia, which avoids the complexity of the process and the interference with the original state of the spores caused by the labeling operation. By simultaneously acquiring multi-dimensional feature parameters related to morphology and terahertz, the differences between different types of spores are fully captured. This simplifies the detection process and improves the reliability of spore type identification, achieving the goal of rapid, label-free detection of Aspergillus conidia and differentiation of their types.

[0038] In another embodiment of the invention, such as Figure 1As shown, the scattering-type terahertz time-domain spectroscopy near-field scanning system adopts a modular integrated design, physically integrating the atomic force microscope module and the terahertz time-domain spectroscopy module on the same working platform. The two modules work together through internal signal transmission lines and control systems, ensuring that the acquisition of morphological images and the detection of terahertz signals can be carried out synchronously during the scanning process, avoiding signal delay or misalignment.

[0039] The terahertz time-domain spectroscopy module uses a commercial standard femtosecond laser as the light source, which can be obtained from conventional optical instrument suppliers. Its operating repetition frequency can be selected in the range of 90-110MHz, the average output power can be adjusted to 90-110mW, the pulse width can be selected to 40-50fs, and the generated terahertz radiation frequency range can cover 0.1-2.1THz. These parameters need to be set through the system's built-in debugging software, combined with the signal strength and response sensitivity required for spore detection, to gradually calibrate and ensure that the radiation output is stable and can effectively excite the characteristic spectral response of Aspergillus conidia.

[0040] The atomic force microscope module is fixed to operate in tapping mode. This mode is chosen based on the fragile structure of Aspergillus conidia, which can reduce mechanical damage to the spores during scanning. Before operation, the tapping mode must be selected in the system software and the corresponding feedback parameters must be set to ensure that the probe and the sample surface maintain an appropriate interaction force during scanning, so as to avoid damaging the spores or detaching them from the sample surface, resulting in signal loss.

[0041] The microcantilever beam oscillation frequency of the atomic force microscope module can be finely adjusted between 45-55kHz, the probe tip radius can be selected from 15-25nm, the tapping amplitude can be set to 140-160nm, and the cantilever beam length can be selected from 280-290μm. The determination of these parameters needs to be carried out through preliminary experiments, scanning tests on Aspergillus conidia of different sizes, observing the clarity and integrity of the morphological images under different parameters, and screening out suitable parameter combinations. The initial parameter values ​​refer to the standard configuration of the equipment at the factory, and then fine-tuned according to the actual detection requirements.

[0042] In this embodiment, by clearly defining the system's module composition and key operating parameters, stable hardware support is provided for the detection process. The integrated modules ensure the synchronization of morphological and spectral data, the parameter settings of the terahertz light source can generate suitable radiation signals, and the working mode and component parameters of the atomic force microscope are adapted to the needs of spore detection. This reduces sample damage while improving the quality and spatial resolution of data acquisition, laying a reliable foundation for subsequent feature extraction and type identification.

[0043] In another embodiment of the invention, the initial scanning area can be selected between 45×45μm and 55×55μm. The parameter settings are mainly based on the expected dispersion density of spores on the low-resistivity silicon substrate, ensuring that the selected area can cover the main range where spores may exist, and avoiding the omission of target spores due to an area that is too small. During operation, the initial scanning mode is first selected in the scattering terahertz time-domain spectroscopy near-field scanning system, the set area parameters are input, and a moderate scanning speed is maintained to quickly traverse the area. The target locations with typical morphological characteristics of spores are observed and marked through the real-time feedback images from the atomic force microscope module, providing clear directions for subsequent fine scanning.

[0044] The fine scanning area can be selected in smaller sizes such as 10×10μm, 8×8μm, or 5×5μm. The specific values ​​are adjusted according to the actual size of the initially located spore to ensure that the scanning area completely encapsulates a single spore without including excessive irrelevant basal areas. After determining the location of the target spore, adjust the sample stage coordinates to center the spore in the fine scanning area. Switch the system to fine scanning mode to increase the scanning resolution and slowly scan the area to ensure that the synchronously acquired atomic force microscopy morphology image and terahertz near-field time-domain signal clearly reflect the detailed features of the spore.

[0045] Before setting parameters, a preliminary experiment is required. Select spore samples with different dispersion densities and test the effect of different preliminary scanning areas on positioning efficiency. Parameters that can quickly cover potential spore areas without increasing invalid scanning time are selected. For fine scanning areas, observe the diameter of spores in the preliminary scanning image and select an area that is 2-3 times larger than the spore diameter to ensure that the spore as a whole and the surrounding key areas can be scanned.

[0046] The experimental subjects were Aspergillus niger spores, Aspergillus flavus spores, and Aspergillus terreus spores fixed on a low-resistivity silicon substrate. The experimental method was to perform preliminary scanning and fine scanning on each sample in sequence, record the total time and data quality of each scan, and verify the rationality of parameter selection by comparing the results under different combinations of scanning parameters, so as to ensure that the detection efficiency is improved without affecting the integrity of the data.

[0047] This implementation method effectively balances detection efficiency and data accuracy through step-by-step scanning. The initial scan quickly locks onto the target, avoiding the time wasted by blind fine scanning. The fine scan focuses on a single spore, ensuring the integrity and clarity of the feature data. This allows the entire detection process to proceed efficiently while providing high-quality raw data for subsequent feature extraction and type identification, thus enhancing the practicality and operability of the method.

[0048] In another embodiment of the present invention, when performing phase-locked demodulation on the terahertz near-field time-domain signal of each pixel, the reference signal is selected from the oscillation frequency of the microcantilever beam and its integer multiples of harmonics. The microcantilever beam oscillation frequency can be selected within the range of 45-55kHz, and the integer multiples of harmonics correspond to the first, second, and third orders, i.e., 1, 2, and 3 times the oscillation frequency. During operation, the reference signal frequency is first set in the signal processing system to be consistent with the oscillation frequency of the microcantilever beam in the atomic force microscope module. Then, the time-domain signal of each pixel acquired by scanning is sequentially demodulated using phase-locked demodulation. Through filtering and phase matching, the modulation signal related to the spore morphology and internal structure is separated to obtain the nth-order near-field signal component. Sn(τ) τ is the optical delay time, which can be finely adjusted according to the propagation path of the terahertz pulse to ensure signal synchronization.

[0049] For the extracted nth order near-field signal component Sn(τ) Terahertz frequency during Fast Fourier Transform w The frequency range can be set from 0.1 to 2.1 THz to match the frequency band requirements of spore characteristic responses. During operation, conventional signal processing software is used to perform transformation operations on each order of signal components of each pixel, converting the time-domain signal in the time dimension into a frequency-domain signal in the frequency dimension. Sn(w) During the transformation, an appropriate number of sampling points is set to ensure spectral resolution, so that the transformed signal can clearly present the spectral characteristics of spores in the terahertz band, providing stable frequency domain data for subsequent amplitude extraction.

[0050] Based on near-field spectrum Sn(w) When extracting amplitude information, the first harmonic needs to be normalized first, referring to the first harmonic spectrum of the substrate. E 1,ref ( w This data was obtained by scanning blank areas on a low-resistivity silicon substrate to ensure consistency with the substrate environment where the spores were located. The normalization process strictly followed the formula. E 1,norm ( w )= E 1,sam ( w ) / E 1,ref ( w After calculation, eliminating interference from the system's own response and the background background, E 1,norm ( w The amplitude s1 is extracted from the data, with a value ranging from 0.20 to 2.00 au; the second and third harmonics are directly extracted from... Sn(w) Extracting amplitude s nThe second-order values ​​range from 0.00 to 0.80 au, and the third-order values ​​range from 0.00 to 0.50 au. Values ​​from the stable segment of the signal in the spectrum are selected during extraction to avoid abnormal peak values ​​affecting the results.

[0051] When the extracted harmonic amplitudes sn are combined into a two-dimensional matrix according to the spatial position of the pixels, the size of the matrix matches the fine scanning area. If the fine scanning area is 10×10μm, the scanning step size can be set to 50-100nm, corresponding to a matrix pixel count of 100×100 to 200×200. During operation, the amplitude data of each order are sequentially filled into the two-dimensional matrix according to the x and y coordinates of the pixels during the scanning process, forming first-order, second-order, and third-order harmonic near-field images. The first-order image focuses on reflecting the overall morphology of the spore and the near-field enhancement effect, while the second- and third-order images are more likely to capture subtle changes in the near-surface dielectric properties and nanoscale compositional inhomogeneities of the spore.

[0052] In this embodiment, background interference and system errors are effectively eliminated through a standardized signal processing procedure, so that the extracted harmonic signals and near-field images can more realistically reflect the characteristic information of spores. Near-field images of different orders present the morphology and internal characteristics of spores from different dimensions, providing rich and reliable data support for subsequent feature parameter extraction and type identification, and improving the accuracy of the entire detection method.

[0053] In another embodiment of the present invention, when constructing the multidimensional feature space, the spatial dimension is consistent with the number of parameters in the feature set, namely, the height value, the intensity values ​​of the first to third harmonic signals, and the peak amplitude, forming a five-dimensional feature space. The feature set of each Aspergillus conidia is a complete set of five-dimensional data, which is sequentially assigned to each coordinate axis in the space according to the parameter type, and transformed into a unique data point in the space. During the transformation process, the original quantization attributes of the parameter values ​​are maintained to ensure that the data point can truly reflect the feature combination of the spore.

[0054] A predefined reference database of Aspergillus conidia types is established, with data derived from the detection of Aspergillus conidia samples of known types. Characteristic parameters are extracted from multiple groups of the same type of spores using the aforementioned method, and statistical analysis is performed to obtain typical range values ​​for each parameter. During database construction, it is necessary to ensure a sufficient number of samples for each type of spore to guarantee the representativeness of the typical range values. During storage, samples are categorized and archived according to spore type, clearly recording the parameter ranges for each type to facilitate rapid retrieval and comparison later.

[0055] When calculating Euclidean distance, the typical range values ​​of parameters for various Aspergillus conidia are first retrieved from the reference database. The center value or statistical average of each range is taken as the comparison benchmark. Then, the characteristic set parameters of the spore to be tested are matched one by one with the benchmark and substituted into the Euclidean distance calculation formula for calculation. During the calculation process, the units of the parameters are kept consistent to avoid distortion of the distance calculation results due to unit differences. The distance of each spore to be tested needs to be calculated separately with all types of spore benchmarks in the database to obtain a set of distance data.

[0056] During parameter matching and classification decisions, the calculated Euclidean distance data is used as the basis for matching the spore to be tested with the spore of the type with the smallest distance, thus determining that the spore to be tested belongs to that type. Before making a decision, a reasonable distance threshold can be set through pre-experiments. If all calculated distances exceed the threshold, it can indicate that the spore type is not included in the reference database, ensuring the rigor of the decision results. Standard samples of Aspergillus niger, Aspergillus flavus, and Aspergillus terreus spores are selected as experimental subjects, and the rationality of the matching logic is verified through multiple comparisons.

[0057] In this embodiment, by constructing a standardized feature space and a data matching process, spore type identification is transformed into objective numerical calculation and comparison, reducing the interference of subjective human judgment. At the same time, the comprehensive consideration of multi-dimensional parameters can make full use of the complementary information of different parameters, improving the reliability and repeatability of type identification, and providing an operable classification basis for the standardized detection of Aspergillus conidia.

[0058] In another embodiment of the present invention, a weighted Euclidean distance algorithm is used for parameter matching and classification decision-making. This algorithm is suitable for comprehensive comparison of five parameters in the feature set: height value, first to third order harmonic signal intensity value, amplitude peak value, etc. The algorithm can be run with the help of conventional data processing software, ensuring the stability and repeatability of the calculation process, without the need for additional complex program development.

[0059] The allocation of weighting coefficients is based on the classification and identification capabilities of the parameters. The weighting coefficients of different parameters can be adjusted between 0.1 and 0.4. For example, the weighting coefficient for amplitude peaks or height values, which are prominent in distinguishing different spore types, can be set between 0.3 and 0.4, while the weighting coefficient for the intensity value of third-order harmonic signals, which has relatively weak identification capabilities, can be set between 0.1 and 0.2, to ensure that the weighting allocation matches the actual identification function of the parameters.

[0060] When calculating the weighted Euclidean distance, the center values ​​of the typical ranges of various parameters for each type of spore in the reference database must first be determined. For example, the center value for the height of Aspergillus niger spores is 400 nm, Aspergillus flavus is 630 nm, and Aspergillus terreus is 1100 nm. The center values ​​for other parameters are all taken as the median values ​​of their corresponding ranges. Then, the difference between each parameter value of the spore to be tested and its corresponding center value is calculated. The difference is squared and multiplied by the weight coefficient of the parameter. Finally, the results of all parameter calculations are summed and the square root is taken to obtain the weighted Euclidean distance between the spore to be tested and the center value of that type of spore. The spore to be tested needs to complete the distance calculation with the center values ​​of all types of spores in the database in turn.

[0061] The weighting coefficients were adaptively determined using historical data, which consisted of spore samples of known types of Aspergillus niger, Aspergillus flavus, and Aspergillus terreus. Each type had at least 30 samples to ensure statistical representativeness. The distribution dispersion of spores for each parameter was analyzed, using variance as a measure of dispersion. Parameters with large variance between different spore types and small variance within the same type indicated a more concentrated distribution, more significant differences between types, and stronger identification ability, thus receiving a higher weighting coefficient. Conversely, parameters with smaller variance were assigned lower weights.

[0062] In this embodiment, by applying the weighted Euclidean distance algorithm and setting scientific weight coefficients, the role of key feature parameters is highlighted and the interference of secondary parameters is weakened. This allows the distance calculation results to better reflect the true correlation between the spores to be tested and known types, thereby improving the accuracy of classification decisions, reducing misjudgments caused by equal parameter weights, and making the type identification results of Aspergillus conidia more reliable.

[0063] In another embodiment of the invention, the construction of the historical database requires the collection of Aspergillus conidia sample data of known types, including Aspergillus niger spores, Aspergillus flavus spores, and Aspergillus terreus spores. The number of samples of each type can be selected between 30 and 50 to ensure that the data has sufficient statistical representativeness. The feature set data of the samples is obtained through the aforementioned detection method, covering five parameters: height value, first to third harmonic signal intensity values, and amplitude peak value. After the data collection is completed, it is classified and stored according to spore type and organized using a conventional database format for easy subsequent retrieval and calculation.

[0064] The Fisher discriminant ratio is calculated for each parameter in the feature set individually. First, statistical analysis software is used to calculate the inter-class variance of each parameter among the three spores: *Aspergillus niger*, *Aspergillus flavus*, and *Aspergillus terreus*. Then, the intra-class variance of the same parameter within each spore type is calculated and summed. Finally, the Fisher discriminant ratio for that parameter is calculated as the ratio of the inter-class variance to the sum of the intra-class variances. During the calculation, it is crucial to ensure accurate data entry to avoid data errors affecting the discriminant ratio result. The stability of the values ​​can be verified by repeating the calculation 2-3 times.

[0065] When normalizing the Fisher discriminant ratios of the five parameters, the discriminant ratios of all parameters are first summed. Then, the discriminant ratio of each parameter is divided by the sum to obtain the corresponding weight coefficient. After normalization, the sum of all weight coefficients is 1. This process can be completed using conventional data processing tools to ensure rigorous calculation logic and that the weight allocation conforms to the actual discrimination ability of each parameter, avoiding weight imbalance.

[0066] The experimental method employed repeated detection, scanning, signal processing, and parameter extraction of each known spore type sample under identical conditions to minimize the impact of random factors on the data. Statistical analysis of the parameter data obtained from repeated detections verified the reliability of the calculated between-class and within-class variances, ensuring that the Fisher discriminant ratio accurately reflects the classification and discrimination capabilities of each parameter.

[0067] The weight coefficient determination method provided in this embodiment is based on objective historical data and statistical analysis, avoiding biases caused by subjective assignment and ensuring a high degree of match between weight allocation and the actual discriminative ability of the parameters. The normalized weight coefficients are reasonable and operable, providing a reliable basis for parameter priority in the weighted Euclidean distance algorithm, which helps improve the stability and accuracy of the entire classification and recognition process.

[0068] In another embodiment of the present invention, the detection targets are limited to Aspergillus niger spores, Aspergillus flavus spores, and Aspergillus terreus spores. Aspergillus niger spores can be obtained from natural environmental samples such as soil, air, and decaying plant matter, or through conventional laboratory fungal culture methods. During culture, suitable culture media and temperature and humidity conditions are used to ensure stable spore growth and purity meeting detection requirements. Aspergillus flavus spores are commonly found in agricultural products such as grains and oilseeds, and in related storage environments. They can be obtained through isolation and purification from environmental samples or through the culture of commercial standard strains. During acquisition, contamination by other microorganisms must be avoided to ensure the uniqueness of the spore sample, providing a reliable sample basis for subsequent feature extraction and type identification. Aspergillus terreus spores are mainly found in environmental media such as soil and plant debris. They can be purified from environmental samples using a gradient dilution plate separation method, or spores can be prepared using standardized cultured strains to ensure the typicality of the samples used for detection.

[0069] During the experiment, the three types of spores were processed according to the same detection procedure, including preparing spore suspensions, adding and fixing them onto a low-resistivity silicon substrate, scanning and acquiring signals, and extracting characteristic parameters. By collecting multiple sets of sample data, the adaptability of the method to different types of spores was verified, ensuring that the characteristic parameters can effectively distinguish between various types of spores.

[0070] This implementation clearly defines the specific application targets of the detection method, making the optimization and validation of the method more targeted and avoiding the problem of insufficient technical adaptability caused by a broad range of detection targets. Furthermore, these three types of spores are all types that require key detection in the medical, food industry, and agricultural fields. This limitation makes the detection method more aligned with actual application scenarios, enhancing its practical value and operability.

[0071] In another embodiment of the present invention, the parameters of Aspergillus niger spores are determined based on the detection data of standard Aspergillus niger strains, referring to the typical range of various parameters of Aspergillus niger spores in the database. The experimental object is Aspergillus niger spores with qualified purity. Using the aforementioned detection method, multiple samples are sequentially subjected to suspension preparation, substrate fixation, stepwise scanning, signal processing, and feature extraction. After statistical analysis to remove abnormal data, the typical range of its height value is determined to be 380-420nm, and the first to third harmonic signal intensity values ​​are 0.87-1.98au, 0-0.679au, and 0-0.43au, respectively.

[0072] The parameter range of Aspergillus flavus spores was derived from multiple repeated tests of standard Aspergillus flavus spore samples. During the experiment, the temperature, humidity, and scanning parameters of the detection environment were strictly controlled to ensure consistency with the detection environment of Aspergillus niger spores, avoiding parameter deviations caused by environmental factors. Statistical analysis revealed that its height value was 610-650 nm, and the first to third harmonic signal intensity values ​​were 0.73-1.89 au, 0-0.671 au, and 0-0.37 au, respectively.

[0073] The parameter range of Aspergillus terreus spores was obtained by detecting multiple samples of standard Aspergillus terreus strains. The detection procedure was consistent with that for Aspergillus niger and Aspergillus flavus spores to ensure parameter comparability. After statistical analysis of the parameter dispersion of similar samples, the height value was determined to be 1080-1120 nm, and the first to third harmonic signal intensity values ​​were 0.32-1.82 au, 0-0.5 au, and 0-0.33 au, respectively.

[0074] The extraction of amplitude peak values ​​was limited to a frequency range of 1.5-2.0 THz. This frequency band was determined by testing the spectral responses of the three spores across the entire frequency range of 0.2-2.0 THz. Within this frequency band, the differences in amplitude peak values ​​among the three spores are more significant, facilitating differentiation. During extraction, frequency domain signal analysis was used to precisely select peak data within this frequency band as characteristic parameters, ensuring the specificity and effectiveness of the parameters.

[0075] In this embodiment, these clearly defined parameter ranges provide specific and operable comparison standards for the reference database, enabling the various characteristic parameters of the spores to be tested to be directly quantitatively matched with them. This reduces the ambiguity and subjectivity of parameter comparison, improves the efficiency and consistency of type identification, and provides a reliable numerical basis for the accurate classification of Aspergillus conidia.

[0076] The present invention will be further explained and illustrated below with reference to specific embodiments and comparative examples.

[0077] Experimental equipment: such as Figure 1 As shown, a scattering-type terahertz time domain spectroscopy near-field scanning system (THz-TDSs-SNOM) is used, integrating a commercial SNOM platform (Neaspec, attocube systems AG) and a commercial THz time domain spectroscopy THz source (TeraSmart, menlosystems TERA15-TX-FC). This system consists of a Menlosystems terahertz time domain spectrometer and a Nearspec near-field atomic force microscope.

[0078] Example 1: Constructing a reference database.

[0079] First, samples were prepared. The mold strains used in this invention were standard strains of *Aspergillus niger* (ATCC16404), *Aspergillus flavus Link* (AS3.3950), and *Aspergillus terreus* (ATCC10690). Glycerol lysates of *Aspergillus niger* and *Aspergillus terreus* were dissolved and inoculated onto potato dextrose agar (PDA) slants. Glycerol lysates of *Aspergillus flavus* were dissolved and inoculated onto malt extract agar slants. The samples were incubated at 28°C for approximately 20 days. The slants were washed repeatedly with sterile water to collect bacterial cells and spores, forming a bacterial suspension. The suspension was transferred to a small conical flask containing glass beads and shaken at 200 rpm for 45 minutes to separate the bacterial cells and spores. The suspension was filtered through sterile defatted cotton, centrifuged at 5000 rpm for 10 minutes, the supernatant was discarded, and the spores were resuspended in sterile water. A hemocytometer was used to count the spores. The concentration was 1 × 10⁻⁶. 4 CFU / mL. The prepared mold spore suspension was dropped onto a low-resistivity silicon substrate using a 10 μL pipette, with each drop containing approximately 10 μL of liquid. The substrate was then rapidly dried using a hot plate (set to 37 °C). After complete drying, the substrate was fixed on the sample stage within the THz-TDS s-SNOM system.

[0080] Then, the system parameters are set. In this invention, the core light source of the THz-TDS s-SNOM system is a high-performance 1560nm erbium-doped fiber femtosecond laser with a repetition rate of 100MHz, an average output power of 100mW, and a pulse width of 45fs. The laser output is split into two beams by a beam splitter, each with an average power of 50mW. One beam, after power attenuation, is incident on a terahertz transmitting antenna based on an InGaAs / InAlAs superlattice heterojunction structure with an average power of less than 30mW, exciting terahertz radiation in the frequency range of 0.2-2.0THz. The other beam passes through an optical delay line and then reaches a terahertz receiving antenna for detection. Both excitation and detection beams are transmitted through a 6m dispersion-compensated fiber, resulting in laser pulse broadening to 100fs. This configuration ensures the effective generation and coherent detection of the terahertz pulse. The atomic force microscope (AFM) operated in impact mode, with the microcantilever beam oscillating at a frequency of 50 kHz. The probe used was an RMN 25PtIr200B-H with a tip radius of 20 nm. The impact amplitude in the probe experiments was approximately 150 nm, and the cantilever beam length was 285 μm. The extraction of the near-field terahertz signal relied on phase-locked demodulation of the microcantilever beam oscillation, where the oscillation frequency φ was specifically used to demodulate the nth-order near-field signal component. Sn(τ) Where n is the harmonic order, including first, second, and third order. τ This is the optical delay time.

[0081] The oscillation frequency itself does not affect the near-field signal. The detected terahertz time-domain signal is demodulated at the probe oscillation frequency using a lock-in amplifier to obtain the nth-order near-field signal. A Fast Fourier Transform (FFT) is then performed on the demodulated time-domain signal to finally obtain the near-field spectrum. Sn(w) Subsequently, the normalized first harmonic near-field spectrum is calculated using the following formula: E 1,norm ( w )= E 1,sam ( w ) / E 1,ref ( w ),in, E 1,sam ( w () is the first harmonic spectrum of the spore sample. E 1,ref ( w ) is the first harmonic spectrum of the reference substrate, and then from E 1,norm ( w The amplitude s1 is extracted, while for higher-order harmonics, it is directly extracted from... Sn(w) Extracting amplitude s nThis allows us to obtain quantized signal strength data.

[0082] For THz nanoimaging of Aspergillus spore samples, a 150×150 pixel scan was first performed within a 50×50 μm area. After determining the approximate location of the spores, a more detailed 10×10 μm area was then imaged. The temporal signal of each pixel was recorded using a 20×20 grid and processed using time-domain Fourier transform. Subsequently, spectral information was acquired by selecting regions of interest on the low-resistivity silicon substrate and the spores. Depending on the number of regions selected, acquiring a set of data took approximately one hour.

[0083] THz-TDS s-SNOM combines AFM and THz imaging techniques. AFM images can reflect the texture features and roughness information of Aspergillus conidia, and AFM analysis of Aspergillus conidia is a key means of providing information for fungal morphological identification. THz near-field images can easily identify clear regions between biological samples and the substrate. Figure 2 The image shows complete AFM morphology images and THz near-field images of individual Aspergillus niger spores, Aspergillus flavus spores, and Aspergillus terreus spores. Figure 2 (a), (e), and (i) are AFM morphology images of Aspergillus niger, Aspergillus flavus, and Aspergillus terreus spores, respectively, clearly showing the morphological information of individual Aspergillus spores. It can be seen that the three types of spores are approximately round or elliptical in shape, with relatively smooth surface contours. The shape gradually increases from the edge to the center, with the highest point being approximately 400 nm, 630 nm, and 1.1 μm, respectively. Some spores show fine protrusions or textures on their surface, which are related to the glycoproteins on the spore surface and the spore wall structure. The size varies depending on the culture conditions and spore maturity, but the size of a single spore is generally between 2 and 5 μm.

[0084] To reduce background noise interference with THz images, harmonic demodulation is used to process the THz images. For example... Figure 2(bd), (fh), and (jl) correspond to the first to third order THz near-field images of Aspergillus niger, Aspergillus flavus, and Aspergillus terreus spores after harmonic demodulation, respectively. In the THz near-field images, the first to third order harmonic signal intensity ranges for Aspergillus niger spores are 0.87–1.98 au, 0–0.679 au, and 0–0.43 au, respectively; for Aspergillus flavus, they are 0.73–1.89 au, 0–0.671 au, and 0–0.37 au; and for Aspergillus terreus, they are 0.32–1.82 au, 0–0.5 au, and 0–0.33 au. The distribution patterns of bright and dark areas are distinct, reflecting significant differences in THz signal intensity and spatial distribution characteristics among different spores. The first-order signal reflects the overall morphology of the sample and its local enhancement effect on near-field light, showing good consistency with the AFM morphology. Second and third harmonic signals exhibit greater sensitivity to subtle variations in near-surface dielectric properties and nanoscale compositional inhomogeneities within the spores. Higher-order demodulated images reveal additional features, including enhanced local contrast at spore edges and fine internal structures not observable in AFM morphology alone. Notably, compared to *Aspergillus niger* and *Aspergillus flavus*, *Aspergillus terreus* conidia show relatively weaker near-field amplitude signals at higher harmonic orders. This suggests that thicker spore walls and larger volumes may induce weaker local confinement and scattering effects, thereby modulating the detected terahertz near-field amplitude. While the internal materials of *Aspergillus niger*, *Aspergillus flavus*, and *Aspergillus terreus* spores are similar in basic organelle composition, they differ significantly in the types of toxins and the distribution of secondary metabolites such as spore wall pigments. The uniqueness of these internal material components and their spatial distribution is the fundamental physical basis for their distinguishable features in AFM morphology and THz near-field images.

[0085] To accurately compare the dielectric properties of Aspergillus conidia, the influence of spore height on near-field intensity is significant. To elucidate the correlation between spore morphology and THz near-field optical response, the same profile was analyzed for AFM height data and near-field amplitude. Figure 3 (a) shows the extracted AFM cross-section. Figure 3 (b) shows the corresponding surface height distribution along the same line. The gradually increasing and symmetrical peaks in the surface height distribution map confirm the nearly circular nature of the spores. The width and height parameters are consistent with observations in the 2D topographic map, and significant differences exist in the surface height distribution among different Aspergillus spores. However, Figure 4The near-field amplitude distributions revealed unique behaviors. For all three spore species, when the tip scanned the sporophyte, the amplitude signal decreased compared to the surrounding low-resistivity silicon substrate, resulting in a distinct groove in the profile. This signal suppression suggests that the spores induce localized changes in the tip-sample interaction due to their unique dielectric properties, detectable even at the subwavelength scale. *Aspergillus terreus* exhibited the most pronounced signal drop and the widest valley in the amplitude profile, consistent with its greater height and volume. *Aspergillus niger*, on the other hand, showed the smallest near-field signal variation, indicating a relatively weak interaction with the terahertz field under the measurement conditions. This observation suggests that internal composition, such as water content, cell wall thickness, and possible microstructural heterogeneity, plays a crucial role in modulating the amplitude of terahertz near-field scattering.

[0086] To investigate the attenuation characteristics of near-field signals from Aspergillus niger spores at different harmonic orders and their ability to distinguish them from the substrate, harmonic demodulation techniques were used to obtain harmonic images of Aspergillus niger spores from the 1st to the 4th harmonic orders, as shown below. Figure 5 (ad) and a near-field intensity profile along the black dashed line was drawn as follows. Figure 5 (e) It can be seen that as the demodulation frequency increases from 1Ω to 2Ω, the intensity of the near-field signal decreases sharply. This can be attributed to the high-frequency attenuation experienced by the demodulated signal during the locking process. Due to the influence of the far-field background, the relative contrast of the 1Ω THz nanometer image is relatively weak. High-order harmonic signal demodulation is usually used to extract the near-field signal of pure bacteria. Increasing the order of harmonic demodulation can reduce background noise, but it will also reduce the signal intensity, thereby reducing the signal-to-noise ratio, which is not conducive to the analysis of the internal substances of spores.

[0087] To investigate the THz near-field signal frequency response characteristics of different Aspergillus spores, spectral analysis was performed on the original terahertz near-field time-domain signals at specific locations (spore center and base) in each scan. The amplitude response within the characteristic frequency band (1.5–2.0 THz) was extracted, and amplitude curves for different spores were plotted, as shown below. Figure 6 As shown, a significant difference was observed between the near-field amplitude signals recorded at the spore center and the reference substrate region. Furthermore, the localization spectra were consistent across multiple measurement positions and different individual spores, confirming the reproducibility and reliability of the observed spectral features. The three types of spores exhibited significantly different signal intensities in the THz band, and these three types of Aspergillus spores can also be easily identified by the differences in their terahertz spectral information.

[0088] In summary, based on the reference database, the height of *Aspergillus niger* spores is 380-420 nm, with first- to third-order harmonic signal intensities of 0.87-1.98 au, 0-0.679 au, and 0-0.43 au, respectively, and a peak amplitude of 1.40-1.50 au in the 1.5-2.0 THz frequency range; the height of *Aspergillus flavus* spores is 610-650 nm, with first- to third-order harmonic signal intensities of 0.73-1.89 au, 0-0.671 au, and 0-0.37 au, respectively, and a peak amplitude of 2.32-2.42 au in the 1.5-2.0 THz frequency range; and the height of *Aspergillus terreus* spores is 1080-1120 nm, with first- to third-order harmonic signal intensities of 0.32-1.82 au, 0-0.5 au, and 0-0.33 au, respectively, and a peak amplitude of 1.40-1.50 au in the 1.5-2.0 THz frequency range. The peak amplitude in the THz frequency range is 1.85-1.95au.

[0089] Example 2: Rapid label-free detection of Aspergillus conidia using terahertz near-field imaging.

[0090] The experimental materials used were standard strains of *Aspergillus niger* (ATCC16404), *Aspergillus flavus* (AS3.3950), and *Aspergillus terreus* (ATCC10690). Thirty independent spore samples were prepared for each strain, for a total of 90 samples. The experiment was repeated three times to verify reproducibility. Sample preparation, THz-TDS s-SNOM system parameter settings, and imaging scanning were strictly performed according to the protocol in Example 1. AFM morphology images of each spore and terahertz near-field time-domain signals of each pixel were acquired simultaneously.

[0091] The terahertz near-field time-domain signal is amplified and demodulated using phase-locked loop amplification. The first, second, and third harmonic near-field signal components are extracted using the micro-cantilever beam oscillation frequency and its integer multiples as reference signals, and then converted to frequency-domain signals via fast Fourier transform. The first harmonic spectrum is calculated using the formula... E 1,norm ( w )= E 1,sam ( w ) / E 1,ref ( w Normalization E 1,sam ( w () represents the first harmonic spectrum of the spore. E 1,ref ( w (Referring to the first harmonic spectrum of the reference substrate), the amplitude s1 is extracted, and the amplitude s of higher harmonics is directly extracted. nA third-order harmonic near-field image was generated; height values ​​(range 200-1500nm) were extracted from the AFM morphology image; first-order (0.20-2.00au), second-order (0.00-0.80au), and third-order (0.00-0.50au) harmonic signal intensity values ​​were extracted from the third-order harmonic image; and amplitude peak values ​​(1.25-2.50au) were extracted from the local near-field amplitude spectrum (1.5-2.0THz band) at the spore center.

[0092] The classification and identification employs a weighted Euclidean distance algorithm: Based on the reference database constructed in Example 1, the Fisher discriminant ratio (the ratio of the sum of inter-class variance and intra-class variance) for each parameter is calculated, and after normalization, weight coefficients are obtained (height value 0.22, first harmonic intensity 0.18, second harmonic intensity 0.25, third harmonic intensity 0.15, peak amplitude 0.20); a five-dimensional feature space is constructed, and the set of spore features to be tested is compared with the center values ​​of the typical ranges of various spore parameters in the reference database (Aspergillus niger: height 4). (00nm, first order 1.42au, second order 0.34au, third order 0.21au, peak amplitude 1.45au; Aspergillus flavus: height 630nm, first order 1.31au, second order 0.33au, third order 0.18au, peak amplitude 2.37au; Aspergillus terreus: height 1100nm, first order 1.07au, second order 0.25au, third order 0.16au, peak amplitude 1.90au) Calculate the weighted Euclidean distance and classify the spores to be tested into the category with the smallest distance.

[0093] The test metrics include recognition accuracy, single-sample detection time, and repeatability. Accuracy is calculated as the percentage of correctly identified samples out of the total number of samples; detection time is the time from sample fixation to obtaining the classification result; repeatability is evaluated by scanning the same spore five times and calculating the coefficient of variation (CV = standard deviation / mean × 100%) for each feature parameter.

[0094] Experimental results showed that in three repeated experiments, 89 out of 90 samples were correctly identified, with an average recognition accuracy of 98.9%. The average detection time per sample was 15 minutes, including 2 minutes for preliminary scanning, 8 minutes for fine scanning, and 5 minutes for signal processing and classification. The coefficients of variation for each feature parameter were all less than 3%, with height CV=1.2%, first harmonic intensity CV=2.1%, and peak amplitude CV=1.8%, indicating that the method has good repeatability.

[0095] Comparative Example 1: Morphological Detection Method Using Optical Microscopy.

[0096] The experimental samples were identical to those in Example 2, consisting of 30 spore samples each of *Aspergillus niger*, *Aspergillus flavus*, and *Aspergillus terreus*, for a total of 90 samples, repeated three times. Sample preparation: 10 μL of the spore suspension prepared in Example 2 was added to a regular glass slide, allowed to dry naturally, and then one drop of physiological saline was added. A coverslip was then placed on top to prepare a temporary slide. The detection equipment used was an Olympus BX53 optical microscope equipped with a 1000x oil immersion objective. Detection procedure: The slide was placed on the microscope stage, and the focus was adjusted until individual spores were clearly visible. Three lab technicians with more than 5 years of experience in microbial morphology identification independently observed the spores and classified them according to their diameter, shape, surface texture, and other morphological characteristics. The results of the three lab technicians who agreed on the classification were recorded as the final classification result; if their opinions differed, the identification was considered a failure.

[0097] The test metrics were the same as in Example 2, including recognition accuracy, single-sample detection time, and repeatability. Detection time was measured from slide preparation to obtaining the classification result; repeatability was achieved by having three experimenters test the same batch of samples separately and calculating the percentage of consistency in the recognition results.

[0098] The experimental results showed that in three repeated experiments, 65 out of 90 samples were correctly identified, with an average identification accuracy of 72.2%. The three researchers disagreed on 20 samples, mainly concerning the distinction between Aspergillus niger and Aspergillus flavus, due to the overlap in their diameters and shapes under the microscope. The average testing time for a single sample was 30 minutes, including 10 minutes for slide preparation and 20 minutes for microscopic observation. Regarding repeatability, the consistency of the identification results among the three researchers was 78.9%, indicating that traditional morphological methods are highly subjective and have a high misjudgment rate.

[0099] Comparative Example 2: Raman spectroscopy detection method.

[0100] The experimental samples were the same as in Example 2, with 30 spore samples each of *Aspergillus niger*, *Aspergillus flavus*, and *Aspergillus terreus*, for a total of 90 samples, repeated three times. Sample preparation: 10 μL of spore suspension was dropped onto an aluminum foil substrate, dried at 37°C, and then fixed on the sample stage of the Raman spectrometer. The detection equipment used was a Thermo Scientific DXR3xi Raman spectrometer, with an excitation wavelength of 785 nm, a laser power of 50 mW, and a scanning range of 400-2000 cm⁻¹. -1 The integration time was 10 seconds, and each sample was scanned three times to obtain the average spectrum. Signal processing: Baseline noise was subtracted using the instrument's built-in software, and characteristic peaks (protein amide I band at 1650 cm⁻¹) were extracted. -1 Polysaccharide COC stretching vibration 1100cm -1 Lipid CH bending vibration 1450cm -1 The peak height and peak area of ​​the characteristic peaks are calculated as classification parameters.

[0101] The classification method employs Principal Component Analysis (PCA) combined with Support Vector Machine (SVM) algorithms to construct the classification model. Test metrics include recognition accuracy, single-sample detection time, and signal stability. Detection time is measured from sample fixation to obtaining the classification result; signal stability is evaluated using the coefficient of variation of the peak heights of the characteristic peaks from three scans of the same spore.

[0102] The experimental results showed that in three repeated experiments, 73 out of 90 samples were correctly identified, with an average identification accuracy of 81.1%. The Raman characteristic peaks of Aspergillus niger and Aspergillus terreus partially overlapped, leading to 12 misidentifications. The average detection time for a single sample was 30 minutes, including 10 minutes for scanning and 20 minutes for signal processing. The average coefficient of variation of the characteristic peak height was 8.7%. Due to the small spore size, the far-field Raman signal was weak, the spatial resolution was insufficient, and the signal stability was poor.

[0103] Comparison with Example 1: Terahertz detection method with few parameters.

[0104] The experimental samples, equipment, and scanning procedures were completely identical to those in Example 2, with simplification only in the feature extraction stage: only the AFM topography height value and the first harmonic signal intensity were retained, while the second harmonic intensity, third harmonic intensity, and local near-field amplitude peak parameters were removed. The classification method still used the weighted Euclidean distance algorithm from Example 2, with the weight coefficients obtained by normalizing the Fisher discriminant ratio of the remaining two parameters (height value 0.55, first harmonic intensity 0.45). The test indicators were the same as in Example 2, including recognition accuracy, single-sample detection time, and repeatability.

[0105] Experimental results showed that in three repeated experiments, 77 out of 90 samples were correctly identified, with an average identification accuracy of 85.6%. The average detection time per sample was 12 minutes, which was reduced by 3 minutes due to the reduction in feature parameters. In terms of repeatability, the coefficients of variation of the feature parameters were similar to those in the previous example (height value CV = 1.3%, first harmonic intensity CV = 2.3%), but the identification accuracy decreased by 13.3 percentage points compared to the previous example. The reason for this may be that second- and third-order harmonic signals are more sensitive to the near-surface dielectric properties and nanoscale compositional inhomogeneities of spores. The peak amplitude in the 1.5-2.0 THz frequency band can reflect the differences in internal spore composition. Without these parameters, it is impossible to fully capture the characteristic differences of different types of spores, making it difficult to distinguish some morphologically similar spores.

[0106] Comparison with Example 2: Ordinary Euclidean distance terahertz detection method.

[0107] The experimental samples, equipment, scanning procedures, and feature extraction were completely consistent with Example 2, with adjustments made only to the classification algorithm: the ordinary Euclidean distance algorithm was used instead of the weighted Euclidean distance algorithm. Specifically, equal weights (0.2 each) were assigned to the five parameters: height, first to third harmonic intensity, and peak amplitude. The ordinary Euclidean distance between the feature set of the spores to be tested and the center values ​​of various spore parameters in the reference database was calculated, and the spores to be tested were classified into the category with the smallest distance. The test indicators were the same as in Example 2, with a focus on comparing the recognition accuracy.

[0108] Experimental results showed that in three repeated experiments, 83 out of 90 samples were correctly identified, with an average accuracy of 92.2%, a decrease of 6.7 percentage points compared to 98.9% in Example 2. Of the seven misclassified samples, five were confusions between *Aspergillus niger* and *Aspergillus flavus*, and two were confusions between *Aspergillus flavus* and *Aspergillus terreus*. The reason for this misclassification may be that the ordinary Euclidean distance does not consider the differences in classification and discrimination capabilities of various parameters. In Example 2, the weighting coefficients determined by the Fisher discriminant ratio highlighted the role of the second harmonic intensity (weight 0.25) and peak amplitude (weight 0.20), two parameters with strong discriminative power, while weakening the influence of the weaker third harmonic intensity (weight 0.15). This made the classification decision more closely reflect the actual contribution of different parameters, thereby improving the recognition accuracy.

[0109] In summary, the terahertz near-field imaging rapid label-free detection method used in Example 2 significantly outperformed Comparative Example 1 (optical microscopy 72.2%) and Comparative Example 2 (Raman spectroscopy 81.1%) in terms of recognition accuracy (98.9%). Compared with traditional optical microscopy, Example 2 overcomes the diffraction limit, combining nanoscale morphology with terahertz spectral information to completely solve the problem of high misclassification rate caused by morphological overlap. Compared with Raman spectroscopy, terahertz near-field technology overcomes the spatial resolution limitation of far-field spectroscopy, and multi-harmonic signals and local spectral peaks can more sensitively capture the dielectric properties and compositional differences inside spores, resulting in better signal stability. The results of Comparative Example 1 and Comparative Example 2 further verify that multi-parameter extraction (height, multi-harmonics, and amplitude peaks) is the key to achieving high accuracy, and the lack of any category parameter will lead to a decrease in distinguishing ability. The weighted Euclidean distance algorithm, through adaptive weight allocation, fully utilizes the role of high-discrimination parameters and is more suitable for the accurate classification of Aspergillus spores than ordinary Euclidean distance. Furthermore, the single-sample detection time in Example 2 is only 15 minutes, which is much shorter than that of optical microscopy and Raman spectroscopy. It also requires no labeling reagents, fully meeting the practical application requirements of rapid and label-free testing. The high accuracy and low coefficient of variation in the three repeated experiments demonstrate that the method has excellent reproducibility, providing a reliable guarantee for batch sample screening in actual testing.

[0110] The number of devices and processing scale described herein are for simplification of the invention. Applications, modifications, and variations of the present invention's method for rapid, label-free detection of Aspergillus conidia using terahertz near-field imaging will be readily apparent to those skilled in the art.

[0111] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A method for rapid label-free detection of Aspergillus conidia using terahertz near-field imaging, characterized in that, include: Step 1: Prepare Aspergillus conidia into a spore suspension and drop it onto a low-resistivity silicon substrate. After drying, fix it on the sample stage of a scattering terahertz time-domain spectroscopy near-field scanning system. Step 2: Use a scattering-type terahertz time-domain spectroscopy near-field scanning system to scan a single spore fixed on a low-resistivity silicon substrate, and simultaneously acquire atomic force microscopy morphology images of the spore and terahertz near-field time-domain signals of each pixel. Step 3: Perform phase-locked amplification and demodulation on the terahertz near-field time-domain signal to extract the first-order harmonic near-field image, the second-order harmonic near-field image, and the third-order harmonic near-field image; extract the local near-field amplitude spectrum from the terahertz near-field time-domain signal for the spore center position; Step 4: Extract the morphological features of each spore from the atomic force microscope images. The morphological features include height values ​​ranging from 200 to 1500 nm; extract the intensity values ​​of the first harmonic signal ranging from 0.20 to 2.00 au from the first harmonic near-field images; extract the intensity values ​​of the second harmonic signal ranging from 0.00 to 0.80 au from the second harmonic near-field images; and extract the intensity values ​​of the third harmonic signal ranging from 0.00 to 0.50 au from the third harmonic near-field images. Peak values ​​with amplitudes ranging from 1.25 to 2.50 au were extracted from the local near-field amplitude spectrum. Step 5: Combine the height value, first harmonic signal intensity value, second harmonic signal intensity value, third harmonic signal intensity value and amplitude peak value of each spore into a feature set. By comparing the differences of various parameters of Aspergillus conidia in the feature set, the type of Aspergillus conidia can be identified.

2. The method for rapid label-free detection of Aspergillus conidia using terahertz near-field imaging according to claim 1, characterized in that, The scattering-type terahertz time-domain spectroscopy near-field scanning system integrates an atomic force microscope module and a terahertz time-domain spectroscopy module. The terahertz time-domain spectroscopy module uses a femtosecond laser as the light source, with a working repetition frequency of 100 MHz, an average output power of 100 mW, a pulse width of 45 fs, and generates terahertz radiation with a frequency range of 0.2–2.0 THz. The atomic force microscope module operates in knocking mode, with a microcantilever beam oscillation frequency of 50 kHz, a probe tip radius of 20 nm, a knocking amplitude of approximately 150 nm, and a cantilever beam length of 285 μm.

3. The method for rapid label-free detection of Aspergillus conidia using terahertz near-field imaging according to claim 2, characterized in that, A scattering-type terahertz time-domain spectroscopy near-field scanning system was used to scan individual spores fixed on a low-resistivity silicon substrate. Specifically, the scan was first performed in a 50×50 μm area to locate the spores, and then a fine scan was performed in a 10×10 μm or smaller area.

4. The method for rapid label-free detection of Aspergillus conidia using terahertz near-field imaging according to claim 2, characterized in that, The terahertz near-field time-domain signal is subjected to phase-locked amplification and demodulation, specifically including: performing phase-locked demodulation on the terahertz near-field time-domain signal of each pixel, using the oscillation frequency of the micro-cantilever beam and its integer multiples of harmonics as reference signals, and extracting the nth-order near-field signal component. Sn(τ) Where n is the harmonic order, including first, second, and third order. τ The optical delay time is used to separate the modulation signal related to spore morphology and internal structure; the nth order near-field signal component of each extracted pixel is... Sn(τ) Perform a Fast Fourier Transform to convert the time-domain signal into a frequency-domain signal, and obtain the near-field spectrum. Sn(w) ,in, w This is a terahertz frequency, and this conversion allows the signal to be used to analyze the spectral characteristics of the terahertz band; based on the near-field spectrum. Sn(w) Amplitude information is extracted for each harmonic order n. For the first harmonic, the near-field spectrum is first normalized to eliminate the influence of system response and floor. The normalization formula is: E 1,norm ( w )= E 1,sam ( w ) / E 1,ref ( w ),in, E 1,sam ( w () is the first harmonic spectrum of the spore. E 1,ref ( w ) is the first harmonic spectrum of the reference substrate, and then from E 1,norm ( w The amplitude s1 is extracted, while for higher-order harmonics, it is directly extracted from... Sn(w) Extracting amplitude s n This allows us to obtain quantized signal strength data; the amplitude s of each harmonic order n is then extracted. n The pixels are combined into a two-dimensional matrix to form a terahertz near-field image, including first-order, second-order, and third-order harmonic near-field images. The first-order harmonic near-field image reflects the overall morphology of the spore and the near-field enhancement effect, while the second-order and third-order harmonic near-field images show stronger sensitivity to subtle changes in the near-surface dielectric properties and nanoscale compositional inhomogeneities within the spore.

5. The method for rapid label-free detection of Aspergillus conidia using terahertz near-field imaging according to claim 1, characterized in that, Step five specifically includes: constructing a multidimensional feature space, where the feature set of each spore corresponds to a data point in this space; based on a predefined Aspergillus conidia type reference database, which stores the typical range values ​​of various parameters in the feature set for different Aspergillus conidia types; and by calculating the Euclidean distance between the feature set of the spore to be tested and the typical range values ​​of various parameters in the reference database, performing parameter matching and classification decisions to identify the specific type of the spore to be tested.

6. The method for rapid label-free detection of Aspergillus conidia using terahertz near-field imaging according to claim 5, characterized in that, Parameter matching and classification decisions are made by calculating the Euclidean distance between the feature set of the spores to be tested and the typical range values ​​of each parameter in the reference database. Specifically, a weighted Euclidean distance algorithm is used to assign a weight coefficient based on its classification ability to each parameter in the feature set. The weighted Euclidean distance between the spores to be tested and the center value of the typical range value of each type of spore in the reference database is calculated, and the spores to be tested are classified into the category with the smallest weighted Euclidean distance. The weight coefficients are adaptively determined by analyzing the distribution dispersion of various types of spores on different parameters in historical data. For parameters that show greater differences and more concentrated distribution among different types of spores, higher weight coefficients are assigned.

7. The method for rapid label-free detection of Aspergillus conidia using terahertz near-field imaging according to claim 6, characterized in that, The method for determining the weighting coefficients includes: collecting and storing feature set data of multiple samples of known types of Aspergillus conidia to form a historical database. Each sample's feature set includes five parameters: height value, first harmonic signal intensity value, second harmonic signal intensity value, third harmonic signal intensity value, and peak amplitude. For each parameter in the feature set, the Fisher discriminant ratio is calculated to distinguish different spore types. The Fisher discriminant ratio is calculated as the ratio of the variance of the parameter among different spore types to the sum of the variances of the parameter within the same spore type. A higher ratio indicates a stronger ability of the parameter to distinguish different spore types. The calculated Fisher discriminant ratios for all five parameters are normalized to obtain the weighting coefficients for the five parameters.

8. The method for rapid label-free detection of Aspergillus conidia using terahertz near-field imaging according to claim 5, characterized in that, Aspergillus conidia include Aspergillus niger spores, Aspergillus flavus spores, and Aspergillus terreus spores.

9. The method for rapid label-free detection of Aspergillus conidia using terahertz near-field imaging according to claim 8, characterized in that, According to the reference database, the height of *Aspergillus niger* spores is 380-420 nm, with first- to third-order harmonic signal intensities of 0.87-1.98 au, 0-0.679 au, and 0-0.43 au, respectively, and a peak amplitude of 1.40-1.50 au in the 1.5-2.0 THz frequency range; the height of *Aspergillus flavus* spores is 610-650 nm, with first- to third-order harmonic signal intensities of 0.73-1.89 au, 0-0.671 au, and 0-0.37 au, respectively, and a peak amplitude of 2.32-2.42 au in the 1.5-2.0 THz frequency range; and the height of *Aspergillus terreus* spores is 1080-1120 nm, with first- to third-order harmonic signal intensities of 0.32-1.82 au, 0-0.5 au, and 0-0.33 au, respectively, and a peak amplitude of 1.40-1.50 au in the 1.5-2.0 THz frequency range. The peak amplitude in the THz frequency range is 1.85-1.95au.

Images