Method for improving spectral resolution in hemoglobin and its application

Abstract

This invention provides a method for improving the spectral resolution of hemoglobin and its application. The method uses CWT-2T2D to analyze hemoglobin spectral samples. First, CWT processing is performed, selecting a Symlet wavelet with a vanishing moment of 4 and setting the scale parameter to 20 to process the hemoglobin spectral samples, eliminating background interference and noise in the spectrum. Then, a two-trace two-dimensional correlation spectral analysis method is used to analyze the CWT-processed hemoglobin spectral samples. This invention solves problems such as spectral mixing and insufficient resolution through the CWT method, significantly improving spectral resolution and thus identifying significant differences in highly similar NIRS raw spectra. In the second stage, two-trace two-dimensional correlation (2T2D-COS) technology is introduced to further reveal the spectral differences and their biological significance between lung cancer patients and healthy controls. Based on the results of the two-trace two-dimensional correlation spectral analysis, the HB fingerprint band of lung cancer patients is located.

Description

Technical Field

[0001] This invention relates to the field of spectroscopy, and in particular to a method for improving the spectral resolution of hemoglobin and its application. Background Technology

[0002] Lung cancer is one of the leading causes of cancer-related morbidity and mortality worldwide. Despite advancements in medical technology, the lack of specific early symptoms means that most lung cancer patients are diagnosed at an advanced stage, significantly impacting treatment outcomes and survival rates. Currently, routine detection methods for lung cancer include chest X-rays, CT scans, and tissue biopsies. However, these methods have several limitations, such as radiation exposure risks, high invasiveness, and high cost. Furthermore, the sensitivity and specificity of these methods in early lung cancer detection still need improvement; for suspected lung cancer patients lacking specific symptoms, diagnostic accuracy remains low, and false negatives or false positives are common. Therefore, exploring new, non-invasive methods for early lung cancer diagnosis is of significant clinical importance.

[0003] In recent years, optical methods have attracted widespread attention in the field of disease biomarker detection and diagnosis due to their advantages such as being non-invasive, rapid, and highly sensitive. This is because optical methods can provide the "spectral fingerprint" of all molecules in a biological sample. Near-infrared spectroscopy (NIR spectroscopy) has garnered significant attention due to its unique advantages. Within the near-infrared range of 650 to 950 nm, it possesses a known optical window, allowing light to penetrate deep layers or large amounts of tissue (up to several centimeters) with relatively low tissue absorption. Feng et al. collected 171 plasma samples (Zhang, P., Zhang, WY, Zhu, J., Chen, ZJ, Feng, JG, Plasma-based near-infrared spectroscopy for early diagnosis of lung cancer, J. Pharm. Biomed. Anal. 249(2024) 116376.), including 73 healthy controls (HC), 73 lung cancer patients (LC), and 25 benign lung tumors (B). Near-infrared spectroscopy was performed on the samples, and after preprocessing, diagnostic models based on four machine learning algorithms were established using the training set data. The results showed that the model could distinguish between HC and LC relatively well, but its ability to distinguish between B and LC was poor. NIRS can also be used to determine the stage of cancer and whether it has spread to other parts of the body. To date, related research has mainly focused on the analysis of lung cancer tumor tissue, serum, and plasma samples.

[0004] Compared to histopathological and imaging examinations, blood tests, which rely on the analysis of specific analytes and tumor biomarkers (such as tumor DNA / proteins and circulating tumor cells), are non-invasive and more efficient. However, due to the challenges of validating many potential biomarkers for different cancer types, only a few biomarkers have been validated and are used clinically. Therefore, there is an urgent need to conduct research on new blood biomarkers, such as the potential value of hemoglobin (HB) in lung cancer diagnosis. Hemoglobin, as one of the most important proteins in the human body, plays a crucial role in oxygen transport and metabolism. Recent studies have found that lung cancer may affect the structure and function of hemoglobin. Given the simplicity and widespread availability of hemoglobin testing, in-depth research on its application in lung cancer diagnosis has significant clinical translational value. When detecting blood components in lung cancer patients using NIRS, the characteristic absorption of HB is not obvious. To address this issue, it is necessary to improve the spectral resolution of hemoglobin. Summary of the Invention

[0005] The technical problem to be solved by this invention is how to improve the spectral resolution of hemoglobin.

[0006] The present invention solves the above-mentioned technical problems through the following technical means: a method for improving the spectral resolution of hemoglobin, which uses CWT-2T2D to analyze hemoglobin spectral samples. First, CWT processing is performed, a Symlet wavelet with a vanishing moment of 4 is selected, and the scale parameter is set to 20 to process the hemoglobin spectral samples to eliminate background interference and noise in the spectrum. Then, the two-trace two-dimensional correlation spectral analysis method is used to analyze the hemoglobin spectral samples processed by CWT.

[0007] As a further optimized technical solution, in the dual-trace two-dimensional correlation spectral analysis method, the synchronous spectrum Φ(v1,v2) and the asynchronous spectrum Ψ(v1,v2) are respectively:

[0008]

[0009] Where r(v) is the reference spectrum, s(v) is the sample spectrum, r(v) is the average spectrum of healthy individuals, and s(v) is the average spectrum of lung cancer patients.

[0010] As a further optimized technical solution, this method for improving the spectral resolution of hemoglobin also includes a model building step. Before modeling, all samples are labeled as healthy individuals and lung cancer patients. First, the KS algorithm is used to divide the spectral dataset into a calibration set and a prediction set. After analyzing the samples using CWT-2T2D, a spectrum composed of selected bands is obtained. Multiple machine learning algorithms are used for the differential analysis between the healthy control group and lung cancer patients. The performance evaluation of the classification model is based on measurements after several iterations. The accuracy, precision, specificity, and sensitivity of the model are calculated using the following formulas:

[0011]

[0012]

[0013] TP represents true positive; TN represents true negative; FP represents false positive; and FN represents false negative. For each indicator, a higher value indicates better performance of the corresponding model.

[0014] As a further optimized technical solution, all data processing was performed using MATLAB software 2020a, and the plotting of the 2T2D correlation spectrum was performed using the two-dimensional correlation spectroscopy software 2DCS Professional Edition.

[0015] As a further optimized technical solution, the hemoglobin extraction process is as follows:

[0016] First, a 1.5ml blood sample was centrifuged at 2000r / min for 10 minutes at low temperature to achieve preliminary separation of red blood cells and plasma;

[0017] Subsequently, the upper plasma layer and the middle white blood cell layer were carefully removed using a high-precision pipette, leaving only the bottom red blood cell layer. 0.5 ml of pre-cooled isotonic saline was precisely added to the remaining red blood cells, and after gentle mixing, the mixture was centrifuged again under the same conditions. This washing step was repeated 1-2 times.

[0018] After washing, carefully remove the supernatant, precisely add 1 ml of red blood cell lysis buffer to the purified red blood cells, mix gently, and let stand at room temperature for 5-10 minutes.

[0019] Subsequently, the lysed sample was centrifuged at 4°C and 2000 r / min for 10 minutes to effectively separate unlysed cells and cell debris.

[0020] After centrifugation, the sample exhibited a typical three-layer distribution: a small amount of dark red precipitate at the bottom, white flocculent matter in the middle, and a red transparent solution at the top. Using a precision pipette, the upper layer of hemoglobin solution was collected and immediately stored at 4°C.

[0021] As a further optimized technical solution, the low temperature environment refers to 4℃, and the same conditions refer to 4℃, 2000r / min, and 10 minutes.

[0022] As a further optimized technical solution, the main component of the red blood cell lysis fluid is ammonium chloride.

[0023] As a further optimized technical solution, the spectral acquisition process is as follows: all blood near-infrared spectra were obtained using an MB3600 spectrometer with an InGaAs detector.

[0024] As a further optimized technical solution, both the air reference and sample spectra were measured with a scan number of 64 during the spectral acquisition process. A quartz cuvette with a 1mm optical path length and a liquid cell were used in the experiment. The spectral acquisition mode was transmission mode, and the instrument resolution was set to 4cm². -1 The spectral acquisition range is controlled within 4000-11000 cm⁻¹ -1 .

[0025] This invention also provides the application of the method for improving the spectral resolution of hemoglobin described in any of the above-mentioned schemes in HB fingerprint band localization. Based on the results of the dual-trace two-dimensional correlation spectral analysis, the HB fingerprint band of lung cancer patients is located in the range of 4870-4850, 4625-4580, 4440-4425, 4380-4350 and 4270-4250 cm-1.

[0026] The advantages of this invention are: it proposes a new framework for improving the spectral resolution of hemoglobin, comprising two stages:

[0027] In the first stage, the CWT method was used to address issues such as spectral mixing and insufficient resolution, significantly improving spectral resolution and enabling the identification of significant differences in highly similar raw NIRS spectra.

[0028] In the second phase, the two-trace two-dimensional correlation (2T2D-COS) technique was introduced to further reveal the spectral differences and their biological significance between lung cancer patients and healthy controls, and to assign unknown characteristic peaks. Among these, 4432 cm⁻¹ was identified as the α-helical structure fingerprint spectral band associated with hemoglobin (HB). Furthermore, by combining synchronous and asynchronous spectral information, additional information on HB secondary structure changes was obtained: 4432, 4615, and 4360 cm⁻¹ belong to the α-helical structure, with changes lagging behind the β-sheet structure corresponding to 4862 cm⁻¹ and leading the CH side chain structure corresponding to 4260 cm⁻¹. This finding provides important evidence for understanding the mechanism of hemoglobin state changes in the blood of lung cancer patients.

[0029] Based on the analysis results of the CWT-2T2D framework, the HB fingerprint band of lung cancer patients can be located in the range of 4870-4850, 4625-4580, 4440-4425, 4380-4350 and 4270-4250 cm-1. Attached Figure Description

[0030] Figure 1 These are the raw NIRS spectra of healthy individuals and lung cancer patients in the range of 11000-4000 cm⁻¹;

[0031] Figure 2 For both healthy individuals and lung cancer patients, the ideal height is between 6250-5600cm. -1 (a) 4900-4200cm -1 (b) CWT-NIRS;

[0032] Figure 3 The 2T2D correlation spectrum analysis is performed in the 4900-4500cm-1 band (red indicates the positive correlation region, blue indicates the negative correlation region, and '+' in the figure indicates the position of the cross peak mentioned in the text).

[0033] Figure 4 The 2T2D correlation spectrum analysis is performed in the 4500-4200cm-1 band (red indicates the positive correlation region, blue indicates the negative correlation region, and '+' in the figure indicates the position of the cross peak mentioned in the text).

[0034] Figure 5 This is a flowchart of data processing in an embodiment of the present invention. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] 1. Sample preparation and spectral acquisition

[0037] Hemoglobin Extraction: This invention employs a rigorous blood processing procedure to obtain high-purity hemoglobin samples for subsequent analysis. A total of 134 samples were collected, all from whole blood samples of the Department of Hematology, Hefei Cancer Hospital, Chinese Academy of Sciences, including 78 healthy volunteers and 56 clinically diagnosed lung cancer patients. Lung cancer diagnoses were confirmed by pathologists through histopathological examination or imaging analysis to ensure the reliability and representativeness of the samples. All subjects had similar socioeconomic backgrounds. This invention has been approved by the hospital's ethics committee, and all subjects were informed and consented to their participation. To focus on the hemoglobin spectral characteristics of lung cancer patients and healthy controls, patients with anemia due to non-tumor-related bleeding were excluded.

[0038] The hemoglobin extraction process is as follows: First, a 1.5ml blood sample is centrifuged at 2000 rpm for 10 minutes at a low temperature (4℃) to achieve initial separation of red blood cells and plasma. Then, using a high-precision pipette, the upper plasma layer and the middle leukocyte layer are carefully removed, leaving only the bottom red blood cell layer. To further purify the red blood cells, 0.5ml of pre-cooled isotonic saline is precisely added to the remaining red blood cells, gently mixed, and centrifuged again under the same conditions (4℃, 2000 rpm, 10 minutes). This washing step is repeated 1-2 times to remove residual plasma and leukocytes to the greatest extent possible, ensuring sample purity. After washing, the supernatant is carefully removed, and 1ml of red blood cell lysis buffer (mainly composed of ammonium chloride) is precisely added to the purified red blood cells. After gentle mixing, the sample is allowed to stand at room temperature for 5-10 minutes to ensure complete lysis of the red blood cells. Subsequently, the lysed sample is centrifuged at 4℃ and 2000 rpm for 10 minutes to effectively separate unlysed cells and cell debris. After centrifugation, the sample exhibited a typical three-layer distribution: a bottom layer consisting of a small amount of dark red precipitate (mainly platelets and other substances), a middle layer of white flocculent matter (broken red blood cell membranes), and an upper layer of clear red solution (hemoglobin solution). Using a precision pipette, the upper hemoglobin solution was carefully collected, taking special care to avoid disturbing the other layers to ensure sample purity. The collected hemoglobin solution was immediately stored at 4°C for subsequent analysis.

[0039] Spectral Acquisition: All near-infrared spectra of blood samples were acquired using an MB3600 spectrometer (ABB Ltd., Canada) with an InGaAs detector. To better acquire spectral information of hemoglobin and improve the signal-to-noise ratio, both the air reference and sample spectra were measured at 64 scans. During the experiment, a quartz cuvette with a 1 mm optical path length and a liquid cell was used. The spectral acquisition mode was transmission mode, and the instrument resolution was set to 4 cm⁻¹. -1 The spectral acquisition range is controlled within 4000-11000 cm⁻¹ -1

[0040] 2. Data processing, the process is as follows: Figure 5 As shown, it includes:

[0041] 2.1 Introduction to CWT

[0042] To delve deeper into the variation patterns of hemoglobin (Hb), continuous wavelet transform (CWT) technology was employed to improve spectral resolution. CWT, as a powerful signal processing tool, effectively eliminates background interference and noise in the spectrum, and its effectiveness in analyzing complex signals has been widely validated. Unlike discrete wavelet transform (DWT), which has binary scale and shift parameters, CWT has continuous scale and shift parameters throughout the signal domain, resulting in highly redundant basis sets. CWT coefficients can be viewed as correlation coefficients, measuring the similarity between the wavelet function and the original signal. Larger scale parameters correspond to wider basis functions, thus providing wavelet coefficients correlated with low-frequency bands, and vice versa. CWT can reveal the characteristic information of a signal at different scales, each scale reflecting a certain frequency range of the signal. The scale parameter in CWT can vary continuously; therefore, the wavelet function can be smoothly shifted throughout the signal domain at any scale. In this invention, a Symlet wavelet (Sym4) with a vanishing moment of 4 is selected, which makes the wavelet transform of the smooth portion of the spectral signal very small, which is beneficial for highlighting the location of feature points. The scale parameter is set to 20. The CWT using Sym4 is approximately equivalent to the fourth derivative and has good smoothing properties. This property is particularly important for preserving subtle features in the spectrum.

[0043] 2.2 Introduction to Two-Trace Two-Dimensional Correlation Analysis (2T2D-COS)

[0044] Two-trace two-dimensional (2T2D-COS) correlation analysis, recently proposed by Noda et al., is an innovative two-dimensional correlation analysis method. This method generates highly informative two-dimensional correlation spectra using only two spectra, and can more sensitively highlight potential asynchronous spectral behavior caused by different sources in a pair of comparative spectra. Furthermore, it better identifies the characteristics of widely overlapping and almost imperceptible neighboring spectra, provided their sources are different. Due to this advantage, 2T2D is currently being used as a tool in disease-related discriminant analysis. In 2T2D correlation analysis, the synchronous spectrum Φ(v1,v2) and the asynchronous spectrum Ψ(v1,v2) are respectively:

[0045]

[0046] Where r(v) is the reference spectrum and s(v) is the sample spectrum. In this invention, unlike traditional generalized two-dimensional correlation methods, 2T2D correlation analysis is used, which can better amplify the differences between lung cancer patients and healthy controls. This is beneficial for identifying HB fingerprint information in the blood, thereby understanding the changes in HB under different disease states. To better represent the overall characteristics of each group and reduce the influence of individual differences, the average spectrum of healthy individuals is selected as r(v), and the average spectrum of lung cancer patients is selected as s(v), which helps to identify consistent HB spectral changes associated with lung cancer.

[0047] Therefore, synchronous spectroscopy reflects the correlation (positive or negative) between spectral intensity at v1 and v2 and lung cancer status, while asynchronous spectroscopy reveals the sequence of spectral changes as lung cancer progresses. In synchronous spectra, the contours along the diagonal reflect the overall changes in hemoglobin (HB) between healthy individuals and lung cancer patients; these are called autocorrelation peaks (always positive). Peaks appearing off-diagonally (v1 ≠ v2) are called cross-peaks, representing the correlation between different wavenumbers. Cross-peaks are positive when the spectral intensity changes at two wavenumbers are in the same direction (i.e., increasing or decreasing simultaneously); they are negative when the changes are in opposite directions. Asynchronous spectra do not contain autocorrelation peaks, only cross-peaks. Cross-peaks (which can be positive or negative) occur when the order of spectral intensity changes at two wavenumbers is different. Notably, according to the precise description by Noda et al., if the synchronous cross-peak is positive, a positive or zero asynchronous cross-peak indicates that the corresponding characteristic peaks originate from the same functional group structure; if the asynchronous cross-peak is negative, it indicates that they originate from different functional group structures. In this invention, the obtained hemoglobin characteristic band (CWT-2T2D-SS) spectral size is a set of one-dimensional spectra (n*m), where m and n represent the number of samples and the number of data points (wavenumber), respectively.

[0048] 2.3 Model Establishment and Evaluation

[0049] Before modeling, all samples were labeled as two groups: healthy individuals and lung cancer patients. First, the Kennard-Stone (KS) algorithm was used to divide the spectral dataset into a calibration set (94 samples) and a prediction set (40 samples), with a ratio of 7:3.

[0050] CWT was used to process blood spectra from lung cancer patients and healthy individuals. Due to the derivative-like properties of CWT, more spectral details were revealed, showing changes in the hepatic oxalate (HB) structure, but the relationships between characteristic peaks remained unclear. Therefore, the introduction of 2T2D can further enhance resolution and provide a deeper understanding of the interactions between characteristic peaks. Therefore, the focus was placed on the protein-rich region of 4900-4200 cm⁻¹. -1By utilizing the differences in hepatitis B (HB) status among control groups as an external perturbation factor, changes in HB status can be sensitively captured, and more HB fingerprint spectra can be obtained. Multiple machine learning algorithms were used for the differential analysis between healthy control groups and lung cancer patients. The performance evaluation of the classification model was based on measurements after 50 iterations, calculating the model's accuracy, precision, specificity, and sensitivity using the following formulas:

[0051] Accuracy

[0052] Accuracy

[0053] Specificity

[0054] Sensitivity

[0055] TP represents a true positive; TN represents a true negative; FP represents a false positive; and FN represents a false negative. For each indicator, a higher value indicates better model performance. Sensitivity and specificity are particularly important in medical diagnosis. High sensitivity (reducing the false negative rate) means ensuring that most patients with lung cancer are detected in a timely manner, avoiding delays in treatment. High specificity (reducing the false positive rate) reduces the probability of healthy individuals being misdiagnosed as lung cancer patients, avoiding unnecessary psychological stress and medical intervention. All data processing in this invention was performed using MATLAB software 2020a (The MathWorks, Natick, MA, USA), with the 2T2D correlation spectra plotted using the two-dimensional correlation spectroscopy software 2DCS Professional Edition.

[0056] 3. Results

[0057] 3.1 Preliminary NIRS Spectroscopic Analysis of Hemoglobin

[0058] Figure 1 The hemoglobin NIRS of healthy controls and lung cancer patients is shown. Three major broad absorption bands are clearly visible in the spectra, located at approximately 6900 cm⁻¹. -1 5200cm -1 and 4950cm -1 Location. Specifically, approximately 6900cm. -1 The broad peak at approximately 5200 cm⁻¹ is attributed to the first overtone absorption of the stretching vibration of the OH molecules in water; -1 The peak at [location] corresponds to the combination frequency absorption of the bending and stretching vibrations (v2+v3) of water molecules; while the peak at approximately 4950 cm⁻¹ corresponds to the absorption of the combination frequency of these vibrations. -1The strong absorption band at this point originates from the combined frequency absorption of the antisymmetric stretching and bending vibrations of the OH group in water molecules. These significant water molecule absorption characteristics may mask the absorption signal of the target component HB. Furthermore, the HB spectra of healthy individuals and lung cancer patients did not show significant differences in peak shape and position. Therefore, to further explore the spectral characteristics of HB in disease states and enhance its characteristic information, this invention performed continuous wavelet transform (CWT) processing on the original spectral data and analyzed the corresponding CWT-NIR spectra.

[0059] Changes in peak intensity in CWT-NIRS accurately reflect changes in sample component content. Since CWT using the Sym4 wavelet function is approximately equivalent to the fourth derivative, CWT technology can enhance subtle differences in NIRS, significantly improving the sensitivity of spectral analysis. The positive peak in the spectrum corresponds to the positive absorption peak in the original spectrum. Given 5200 cm⁻¹... -1 The wavelength band exhibits supersaturated absorption by water, making this region unsuitable for effective study. Therefore, Figure 2 (a) and Figure 2 (b) Focus on 4200-4900cm -1 and 5600-6250cm -1 Spectral variations within the wavenumber range. Figure 2 (a) shows that at 4200-4900cm -1 Within the range, lung cancer patients, compared to the healthy control group, had values ​​of 4862, 4615, 4432, 4366, and 4262 cm⁻¹. -1 The peak intensity at 4862 cm⁻¹ shows a significant decreasing trend. -1 The peak can be attributed to the β-sheet structure of the protein; 4615 cm⁻¹ -1 and 4366cm -1 The peak at 4262 cm⁻¹ belongs to the α-helix structure of the protein; while the peak at 4262 cm⁻¹ belongs to the α-helix structure of the protein. -1 The peak corresponds to the CH side chain vibration of the protein. Figure 2 (b) shows that the distance is 5600-6250cm. -1 Within the range, 5913cm -1 and 5744cm -1 The peak intensity at 5913 cm⁻¹ shows a decreasing trend, with the peak intensity at 5913 cm⁻¹ decreasing. -1 The peaks can be attributed to the β-sheet structure of the protein. These known changes in the intensity of characteristic HB peaks strongly suggest that the secondary structure of HB in lung cancer patients is significantly altered compared to healthy controls. Previous studies have shown that most cancer patients suffer from tumor-related anemia, leading to a reduction in HB levels. The mechanisms underlying this are complex, primarily involving cancer-associated chronic inflammation and the synthesis of pro-inflammatory cytokines by immune cells and cancer cells. Furthermore, peaks at 4820, 4648, 4464, 4432, and 4400 cm⁻¹...-1 The positions and intensities of characteristic peaks also changed. However, because the structure of hemoglobin is affected by various environmental factors in lung cancer patients, and the synergistic relationship between these characteristic peaks is not yet clear, it is currently impossible to accurately attribute the source of these characteristic peaks. Therefore, there is an urgent need to develop new analytical techniques to further explore the spectral differences between healthy controls and lung cancer patients and their biological significance.

[0060] 3.2 2T2D reveals the spectral fingerprint of lung cancer

[0061] To more objectively and convincingly identify the differences between lung cancer patients and healthy controls, this invention employs two-dimensional correlation analysis (2T2D) technology, focusing on the 4900–4500 cm⁻¹ region where protein structures are more abundant. -1 and 4650–4200cm -1 The wavenumber range. 2T2D-NIR analysis provides in-depth information on molecular structure and intramolecular and intermolecular functional group interactions, requiring only two contrast spectra to sensitively capture the influence of external perturbations on the components of the studied system. The presence of lung cancer is considered an external perturbation factor, and 2T2D can more sensitively capture the synchronous and asynchronous spectral behavior between two contrast spectra (average spectra of healthy individuals and lung cancer patients). Considering the broad peak characteristics of raw blood NIRS, this invention uses CWT-processed hemoglobin NIRS for analysis.

[0062] Figure 3 (a) Shows the 2T2D synchronous correlation spectrum in the range of 4900–4500 cm⁻¹. Autocorrelation peaks can be observed at 4862, 4821, 4650, 4615, and 4553 cm⁻¹. -1 place, by Figure 3 (a) The scale variation shown in the two-dimensional graph is obvious, with 4862 cm⁻¹ being particularly prominent. -1 The nearby spectral range showed the highest sensitivity, consistent with previous CWT-NIR analysis results. The sign of the cross peaks in the synchronous correlation spectrum indicates the direction of spectral intensity changes (positive for same-direction changes, negative for opposite-direction changes). (4648, 4550 cm⁻¹), (4820, 4550 cm⁻¹), (4862, 4615 cm⁻¹) -1 ) and (4820, 4648cm -1 The positive correlation at () indicates that the functional groups corresponding to these cross peaks change in the same direction, possibly originating from the same structural change. Other synchronous cross peaks (see Table 1) include (4615, 4550 cm⁻¹). -1 () and (4862, 4648cm) -1The negative correlation at the () indicates that the functional groups corresponding to these cross peaks change in opposite directions, necessarily originating from different structural changes. Combined with the known 4862 cm⁻¹ -1 and 4615cm -1 These are characteristic absorption bands of hemoglobin, from which 4820, 4648, and 4550 cm⁻¹ can be inferred. -1 It is not a characteristic absorption band of hemoglobin.

[0063] Figure 3 (b) Showcasing 4900-4500cm -1 The asynchronous two-dimensional correlation spectrum within the range. The asynchronous correlation spectrum has no autocorrelation peaks, and the sign of the cross peaks indicates the order of spectral intensity changes. Based on the signs of synchronous and asynchronous cross peaks (see Table 1), we can assign the origin and order of change of the cross peaks. Among them, (4648, 4550 cm⁻¹) -1 ) and (4820, 4550cm -1 A negative peak appears at (4820, 4648cm), while (4820, 4648cm) -1 A positive peak appears at 4648cm, indicating that... -1 and 4820cm -1 The corresponding functional groups have the same origin and are related to 4550cm. -1 The corresponding functional groups originate from different sources. 4550cm -1 This may originate from the combination frequency absorption of water molecules in the hemoglobin solution, with 4820 and 4648 cm⁻¹ being the most significant. -1 This may originate from the overtone absorption of the bending vibrations of the OH groups in water molecules. Of particular note is the (4862, 4615 cm⁻¹)... -1 A negative cross peak appeared at 4615 cm⁻¹. According to Noda's rule for determining the sequence of successive events, this indicates that the peak occurred at 4615 cm⁻¹. -1 The corresponding hemoglobin α-helix structure lags behind 4862 cm in rate of change. -1 The corresponding hemoglobin β-sheet structure. This finding is consistent with previous studies on the sequence of protein structural changes in gastric cancer tissue, further validating the effectiveness of the 2T2D method.

[0064] Table 1: Cross-peaks of synchronous / asynchronous correlation spectra from 4900 to 4200

[0065]

[0066] Note:where'+'denotes a positive peak and'-'denotes a negative peak.

[0067] Figure 4(a) Shows the 2T2D synchronous correlation spectrum in the range of 4500–4200 cm⁻¹. Autocorrelation peaks appear at 4461, 4432, 4400, 4366, 4303, and 4262 cm⁻¹. -1 place, by Figure 4 (a) The two-dimensional spectral amplitude changes show that 4366 cm⁻¹ is a significant value. -1 The area is most sensitive. (4461, 4400cm) -1 (4432, 4366cm) -1 The presence of positive cross peaks indicates that the functional groups corresponding to these cross peaks exhibit the same direction of change and may originate from the same source. (4400, 4366 cm⁻¹) -1 (4303, 4262cm) -1 The presence of negative cross peaks indicates that the functional groups corresponding to these cross peaks exhibit opposite directions of change, suggesting they originate from different sources. (4366cm) -1 4262cm -1 It can be confirmed as an HB-related structure, therefore 4400cm -1 and 4303cm -1 It is definitely not an HB characteristic peak (the reasoning also applies to other cross peaks in Table 1).

[0068] Figure 4 (b) Showcasing 4500-4200cm -1 Asynchronous two-dimensional correlation spectrum within the range, (4461, 4303 cm⁻¹) -1 ) and (4400, 4303cm -1 The asynchronous spectrum is shown as negative, while (4461, 4400cm) -1 No cross peaks were observed, indicating that 4303 cm⁻¹ -1 With 4400, 4461cm -1 Different origins, speculated to be 4303cm -1 The absorption originates from the bending vibration of the OH group of water molecules, and 4400cm -1 4461cm -1 The presence of glycated hemoglobin due to changes in the lung cancer environment leads to the appearance of glucose absorption bands, and according to Noda's rule, changes in glucose molecules precede changes in water molecule absorption. Furthermore, and more importantly, this occurs at (4432, 4262 cm⁻¹). -1 (4366, 4262cm) -1 A negative peak appeared at (4432, 4366 cm⁻¹), while (4432, 4366 cm⁻¹) showed a negative peak. -1 However, no cross peaks were observed, so it can be inferred that 4262cm... -1 With 4432, 4366cm -1 The functional groups are different, among which 4262cm -1It is the CH side chain absorption of HB, and 4432cm -1 With 4366cm -1 They share the same origin and are both α-helical structures of HB. Due to (4262, 4432 cm⁻¹) -1 The synchronous spectrum shows a positive peak, while the asynchronous spectrum shows a negative peak at the same position, indicating that the α-helix structure must change before the CH side chain.

[0069] In summary, 2T2D-NIR can more reliably reveal the dynamic spectral characteristics of hemoglobin (HB) in the blood of lung cancer patients and healthy individuals, improving spectral resolution and uncovering more hidden information. Therefore, the spectral ranges of 4870-4850, 4625-4580, 4440-4425, 4380-4350 cm⁻¹, and 4270-4250 cm⁻¹ can be analyzed. -1 The HB fingerprint spectral region within the range is used for diagnostic analysis of healthy individuals and lung cancer patients. These bands also correspond to the locations of HB secondary structure changes shown by CWT-NIRS. The specific band allocation is shown in Table 2.

[0070] Table 2. 2T2D screening for 4900-4200, used for HB characteristic peak band allocation.

[0071]

[0072]

[0073] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for improving the spectral resolution of hemoglobin, characterized in that: CWT-2T2D was used to analyze the hemoglobin spectral samples. First, CWT processing was performed, and a Symlet wavelet with a vanishing moment of 4 was selected. The scale parameter was set to 20 to process the hemoglobin spectral samples to eliminate background interference and noise in the spectrum. Then, the two-trace two-dimensional correlation spectral analysis method was used to analyze the hemoglobin spectral samples processed by CWT. In the dual-trace two-dimensional correlation spectral analysis method, synchronous spectroscopy and asynchronous spectrum They are respectively: （1） （2） in For reference spectrum, The sample spectrum is shown below; the average spectrum of healthy individuals is shown below. The average spectrum of lung cancer patients is .

2. The method for improving the spectral resolution of hemoglobin as described in claim 1, characterized in that: it further includes a model building step. Before modeling, all samples are labeled as two groups: healthy individuals and lung cancer patients. First, the KS algorithm is used to divide the spectral dataset into a calibration set and a prediction set. After analyzing the samples using CWT-2T2D, a spectrum composed of selected bands is obtained. Multiple machine learning algorithms are used for the differential analysis between the healthy control group and lung cancer patients. The performance evaluation of the classification model is based on measurements after several iterations. The accuracy, precision, specificity, and sensitivity of the model are calculated using the following formulas: (3) (4) (5) (6) TP represents true positive; TN represents true negative; FP represents false positive; and FN represents false negative. For each indicator, a higher value indicates better performance of the corresponding model.

3. The method for improving the spectral resolution of hemoglobin as described in claim 1, characterized in that: the hemoglobin extraction process is as follows: First, a 1.5 ml blood sample was centrifuged at 2000 r / min for 10 minutes at low temperature to achieve preliminary separation of red blood cells and plasma; Subsequently, the upper plasma layer and the middle white blood cell layer were carefully removed using a high-precision pipette, leaving only the bottom red blood cell layer. 0.5 ml of pre-cooled isotonic saline was precisely added to the remaining red blood cells, and after gentle mixing, the mixture was centrifuged again under the same conditions. This washing step was repeated 1-2 times. After washing, carefully remove the supernatant, precisely add 1 ml of red blood cell lysis buffer to the purified red blood cells, mix gently, and let stand at room temperature for 5-10 minutes. Subsequently, the lysed sample was centrifuged at 4°C and 2000 r / min for 10 minutes to effectively separate unlysed cells and cell debris. After centrifugation, the sample exhibited a typical three-layer distribution: a small amount of dark red precipitate at the bottom, white flocculent matter in the middle, and a red transparent solution at the top. Using a precision pipette, the upper layer of hemoglobin solution was collected and immediately stored at 4°C.

4. The method for improving the spectral resolution of hemoglobin as described in claim 3, characterized in that: the low temperature environment refers to 4°C, and the same conditions refer to 4°C, 2000 r / min, and 10 minutes.

5. The method for improving the spectral resolution of hemoglobin as described in claim 3, characterized in that: the main component of the red blood cell lysis fluid is ammonium chloride.

6. The method for improving the spectral resolution of hemoglobin as described in claim 3, characterized in that: the spectral acquisition process is as follows: all blood near-infrared spectra are obtained by an MB3600 spectrometer using an InGaAs detector.

7. The method for improving the spectral resolution of hemoglobin as described in claim 6, characterized in that: during the spectral acquisition process, both the air reference and the sample spectrum are measured with a scan number of 64; during the experiment, a quartz cuvette with an optical path length of 1 mm and a liquid cell are used; the spectral acquisition mode is transmission mode; and the instrument resolution is set to 4. The spectral acquisition range is controlled between 4000-11000. .

8. The method for improving spectral resolution in hemoglobin according to any one of claims 1-7 is applied to hemoglobin fingerprint band localization, characterized in that: based on the results of the dual-trace two-dimensional correlation spectral analysis, the hemoglobin fingerprint band of lung cancer patients is located at 4870-4850, 4625-4580, 4440-4425, 4380-4350, and 4270-4250 cm⁻¹. - Within the range.

Images