Determination device, determination system, determination method, and determination program

Abstract

This determination device comprises a processor that acquires a Raman spectrum which has been detected by irradiating a gastrointestinal mucosal tissue with excitation light having a prescribed wavelength, that calculates, from the Raman spectrum in a first wavenumber band, a first feature amount which represents the state or amount of a substance contained in the mucosal epithelium, that determines the risk of future deterioration of the gastrointestinal mucosal tissue on the basis of the first feature amount, and that outputs the result of the determination. Thus, provided is a determination device which is capable of accurately determining the risk of future deterioration of a gastrointestinal mucosal tissue.

Description

Judgment device, judgment system, judgment method, and judgment program

[0001] The present invention relates to a determination device, a determination system, a determination method, and a determination program.

[0002] Conventionally, a technique has been known in which the insertion part of an endoscope is inserted into a subject and the gastrointestinal mucosa of the subject is observed. Patent Document 1 discloses a technique for determining whether ulcerative colitis (UC) is in pathological remission or pathological non-remission using images acquired by a colonoscope.

[0003] Ulcerative colitis is a chronic or remittent / relapsing inflammatory disease of the intestines. Treatment for ulcerative colitis involves induction therapy with medications aimed at achieving clinical remission, followed by maintenance therapy to maintain remission after clinical remission is achieved. Maintenance therapy aims to achieve mucosal healing, defined as a Mayo Endoscopic Score (MES), which is an endoscopic finding based on the surface morphology and color of the mucosa, being 0 or 1.

[0004] Furthermore, in the gastric and esophageal mucosa, chronic inflammation can lead to changes in the properties of mucosal cells, such as intestinal metaplasia, which can degenerate the mucosa into one with a high risk of cancer. In such cases of mucosal abnormalities, the surface morphology and color of the mucosa can be evaluated using endoscopy to diagnose the future risk of cancer.

[0005] Furthermore, in gastroesophageal reflux disease, the degree of inflammation is evaluated using the revised Los Angeles classification through endoscopic examination, and this is used to determine treatment strategies and predict prognosis.

[0006] International Publication No. 2021 / 060159

[0007] However, the morphology and color tone of the mucosa used for evaluation in endoscopic examination are only one element representing the overall properties of the mucosa that appear as a result of physiological phenomena and functional expressions generated as a result of the expression of various molecules and biochemical reactions in the mucosa, and are not necessarily indicators that can sensitively reflect changes in the mucosa correlated with physiological and functional abnormalities of the mucosa that cause diseases. For example, in ulcerative colitis, in cases where mucosal healing defined by the Mayo endoscopic score has been achieved, recurrence of ulcerative colitis occurs at a certain rate and cannot be avoided. Therefore, there is a need for a method to accurately determine the future deterioration risk of gastrointestinal mucosal tissue by detecting biochemical changes in the mucosa that reflect physiological and functional abnormalities of the mucosa and to assist the diagnosis by physicians.

[0008] The present invention has been made in view of the above, and an object thereof is to provide a determination device, a determination system, a determination method, and a determination program that can accurately determine the future deterioration risk such as the occurrence of inflammation and carcinogenesis in gastrointestinal mucosal tissue by detecting changes in substances involved in the mucosal barrier function and inflammatory response of gastrointestinal mucosal tissue.

[0009] In order to solve the above-described problems and achieve the object, a determination device according to one aspect of the present invention acquires a Raman spectrum detected by irradiating a gastrointestinal mucosal tissue with excitation light having a predetermined wavelength, calculates a first feature amount representing the state or amount of substances contained in the mucosal epithelium from the Raman spectrum in a first wavenumber band, determines the future deterioration risk such as the occurrence of inflammation and carcinogenesis in the gastrointestinal mucosal tissue based on the first feature amount, and includes a processor that outputs a determination result.

[0010] Further, in the determination device according to one aspect of the present invention, the first feature amount represents the glycosylation state of mucin.

[0011] Further, in the determination device according to one aspect of the present invention, the first feature amount represents the amount of phospholipids.

[0012] Further, in the determination device according to one aspect of the present invention, the first feature amount represents the glycosylation state of mucin and the amount of phospholipids.

[0013] Further, in the determination device according to one aspect of the present invention, the first frequency band is set according to the types of monosaccharide molecules contained in the O-linked sugar chains of the mucin and the functional groups added to the sugar chain terminals.

[0014] Further, in the determination device according to one aspect of the present invention, the first frequency band is set according to the functional groups contained in the phospholipids.

[0015] Further, in the determination device according to one aspect of the present invention, the processor uses the first feature amount for a plurality of patients to whom a label indicating whether or not the gastrointestinal mucosal tissue has deteriorated in the future is assigned with respect to the time point of measurement of the Raman spectrum of the gastrointestinal mucosal tissue, and determines the future deterioration risk of the gastrointestinal mucosal tissue using a discriminator generated thereby.

[0016] Further, in the determination device according to one aspect of the present invention, the processor calculates a second feature amount representing the amount of neutrophil mucosal infiltration from the Raman spectrum in the second frequency band, and determines the future deterioration risk of the gastrointestinal mucosal tissue based on the first feature amount and the second feature amount.

[0017] Further, in the determination device according to one aspect of the present invention, the processor uses the first feature amount for a plurality of patients to whom a label indicating whether or not inflammatory bowel disease has relapsed in the gastrointestinal mucosal tissue is assigned, and determines the future relapse risk of inflammatory bowel disease in the gastrointestinal mucosal tissue using a discriminator generated thereby.

[0018] Further, in the determination device according to one aspect of the present invention, the processor uses the first feature amount for a plurality of patients to whom a label indicating whether or not the gastrointestinal mucosal tissue of the stomach or esophagus has become cancerous is assigned, and determines the future canceration risk of the gastrointestinal mucosal tissue using a discriminator generated thereby.

[0019] Further, in the determination device according to one aspect of the present invention, the processor uses a discriminator generated using the first feature amount for a plurality of patients to whom a label indicating whether or not gastroesophageal reflux disease has relapsed in the gastrointestinal mucosal tissue is assigned, and determines the future recurrence risk of gastroesophageal reflux disease in the gastrointestinal mucosal tissue.

[0020] Also, the determination device according to one aspect of the present invention, the processor uses the first feature amount and the second feature amount for a plurality of patients to whom a label indicating whether inflammatory bowel disease has relapsed in the gastrointestinal mucosal tissue is assigned, and determines the future relapse risk of inflammatory bowel disease in the gastrointestinal mucosal tissue by a discriminator generated by using the first feature amount and the second feature amount.

[0021] Also, the determination device according to one aspect of the present invention, the first frequency band is a frequency band including Raman peaks of N-acetylglucosamine (GlcNAc), galactose, fucose, mannose, glucose, sulfate groups bound to the sugar chain terminus, N-acetylgalactosamine (GalNAc), or sialic acid.

[0022] Also, the determination device according to one aspect of the present invention, the first frequency band is 780 cm －１ or more and 900 cm －１ or less, 1020 cm －１ or more and 1160 cm －１ or less, 1320 cm －１ or more and 1400 cm －１ or less, or 1400 cm －１ or more and 1500 cm －１ or less.

[0023] Also, the determination device according to one aspect of the present invention, the second frequency band is a frequency band including Raman peaks of myeloperoxidase contained in neutrophils.

[0024] Also, the determination device according to one aspect of the present invention, the second frequency band is 1550 cm －１ or more and 1600 cm －１ or less.

[0025] Also, the determination device according to one aspect of the present invention, the first feature amount is an amount obtained by dividing the area intensity of the Raman spectrum of the first frequency band by the area intensity of the Raman spectrum of a frequency band including Raman peaks derived from amino acids constituting proteins.

[0026] Furthermore, in a determination device according to one aspect of the present invention, the second characteristic quantity is the amount obtained by dividing the area intensity of the Raman spectrum in the second wavenumber band by the area intensity of the Raman spectrum in the wavenumber band that includes the Raman peak derived from the amino acids constituting the protein.

[0027] Furthermore, a determination system according to one aspect of the present invention includes a light source device that irradiates gastrointestinal mucosal tissue with excitation light of a predetermined wavelength, a photodetector that spectrally measures Raman scattered light from the gastrointestinal mucosal tissue to generate a Raman spectrum, and a determination device that includes a processor that calculates a first feature quantity representing the state or amount of substances contained in the mucosal epithelium from the Raman spectrum in the first wavenumber band, determines the future deterioration risk of the gastrointestinal mucosal tissue based on the first feature quantity, and outputs a determination result.

[0028] Furthermore, a determination system according to one aspect of the present invention comprises an endoscope inserted into a subject, and an optical fiber scope inserted into the subject via a treatment instrument conduit of the endoscope, which irradiates the gastrointestinal mucosal tissue in the subject with the excitation light output by the light source device, and collects the Raman scattered light from the gastrointestinal mucosal tissue and guides it to the photodetector.

[0029] Furthermore, a determination method according to one aspect of the present invention involves irradiating gastrointestinal mucosal tissue with excitation light of a predetermined wavelength to obtain a detected Raman spectrum, calculating a first characteristic quantity representing the state or amount of substances contained in the mucosal epithelium from the Raman spectrum in the first wavenumber band, determining the future deterioration risk of the gastrointestinal mucosal tissue based on the first characteristic quantity, and outputting the determination result.

[0030] Furthermore, a determination program according to one aspect of the present invention causes a processor to perform the following actions: irradiate gastrointestinal mucosal tissue with excitation light of a predetermined wavelength to obtain a detected Raman spectrum; calculate a first feature quantity representing the state or amount of substances contained in the mucosal epithelium from the Raman spectrum in the first wavenumber band; determine the future deterioration risk of the gastrointestinal mucosal tissue based on the first feature quantity; and output the determination result.

[0031] According to the present invention, it is possible to realize a determination device, determination system, determination method, and determination program that can accurately determine the future risk of deterioration of gastrointestinal mucosal tissue.

[0032] Figure 1 is a schematic diagram showing the configuration of a determination system including a determination device according to Embodiment 1. Figure 2 is a diagram showing the irradiation of the gastrointestinal mucosal tissue with excitation light from the optical fiber scope shown in Figure 1. Figure 3 is a diagram showing the optical fiber scope shown in Figure 1 as viewed from the tip. Figure 4 is a block diagram showing the configuration of the determination device shown in Figure 1. Figure 5 is a flowchart showing an overview of the processing performed by the determination device according to Embodiment 1. Figure 6 is a diagram showing an example of a Raman spectrum in the mucosal tissue of the large intestine. Figure 7 is a flowchart showing a method for generating a determination device. Figure 8 is a diagram showing feature data calculated from the accumulated Raman spectra. Figure 9 is a diagram showing the wavenumber band used to calculate feature quantities in the Raman spectrum of the mucosal tissue of the large intestine. Figure 10 is a diagram showing a microscopic image of the mucosa captured with a microscope. Figure 11 is a diagram showing the distribution of proteins in observation region A1. Figure 12 is a diagram showing the distribution of mucin in observation region A1. Figure 13 is a diagram showing the Raman spectrum of the mucosal tissue of the large intestine (solid line) and the Raman spectrum of mucin included in the Raman spectrum of the mucosal tissue of the large intestine (dashed line). Figure 14 shows the correlation between feature XA and the glycosylation status of mucin. Figure 15 shows the correlation between feature XB and the glycosylation status of mucin. Figure 16 shows the Raman spectrum of the colonic mucosal tissue (solid line) and the Raman spectrum of phospholipids included in the Raman spectrum of the colonic mucosal tissue (dashed line). Figure 17 shows the correlation between feature XA and the amount of phospholipids. Figure 18 shows the correlation between feature XB and the amount of phospholipids. Figure 19 shows a two-dimensional scatter plot of two features calculated for the relapsed and non-relapsed groups of ulcerative colitis, and the discrimination boundary obtained by training the labeled features using Fisher's linear discriminant method. Figure 20 shows the ROC curve obtained by changing the threshold using Fisher's linear discriminant method. Figure 21 shows the table for judgment. Figure 22 shows the scores of the two features. Figure 23 shows the table for judgment. Figure 24 shows the wavenumber bands used to calculate features in the Raman spectrum of colonic mucosal tissue.Figure 25 shows the Raman spectrum of the colonic mucosal tissue (solid line) and the Raman spectrum of mucin included in the colonic mucosal tissue (dashed line). Figure 26 shows the correlation between feature XA and the glycosylation status of mucin. Figure 27 shows the Raman spectrum of the colonic mucosal tissue (solid line) and the Raman spectrum of phospholipids included in the colonic mucosal tissue (dashed line). Figure 28 shows the correlation between feature XA and the amount of phospholipids. Figure 29 shows a two-dimensional scatter plot of two feature quantities calculated for the relapsed and non-relapsed groups of ulcerative colitis, and the discrimination boundary obtained by training the labeled feature quantities using Fisher's linear discriminant method. Figure 30 shows the ROC curve obtained by changing the threshold using Fisher's linear discriminant method. Figure 31 shows the wavenumber band used to calculate the feature quantities in the Raman spectrum of the colonic mucosal tissue. Figure 32 shows the Raman spectrum of the colonic mucosal tissue (solid line) and the Raman spectrum of mucin included in the colonic mucosal tissue (dashed line). Figure 33 shows the correlation between feature XA and the glycosylation status of mucin. Figure 34 shows the Raman spectrum of the colonic mucosal tissue (solid line) and the Raman spectrum of phospholipids included in the colonic mucosal tissue (dashed line). Figure 35 shows the correlation between feature XA and the amount of phospholipids. Figure 36 shows a two-dimensional scatter plot of two features calculated for the relapsed and non-relapsed groups of ulcerative colitis, and the discrimination boundary obtained by training the labeled features using Fisher's linear discriminant method. Figure 37 shows the ROC curve obtained by changing the threshold using Fisher's linear discriminant method. Figure 38 shows the Raman spectrum of the colonic mucosal tissue and the wavenumber band of a modified example. Figure 39 shows the process of setting the threshold. Figure 40 shows ROC curves obtained by varying the threshold using Fisher's linear discriminant method. Figure 41 shows an example of a Raman spectrum in the colonic mucosa.

[0033] Embodiments of the determination device, determination system, determination method, and determination program according to the present invention will be described below with reference to the drawings. However, the present invention is not limited to these embodiments. In the following embodiments, a determination device, determination system, determination method, and determination program for determining the risk of relapse of inflammatory bowel disease such as ulcerative colitis, the risk of future cancerous transformation of gastrointestinal mucosal tissue, and the risk of recurrence of gastroesophageal reflux disease will be described as examples, but the present invention can be applied in general to a determination device, determination system, determination method, and determination program for determining the risk of future deterioration of gastrointestinal mucosal tissue.

[0034] Furthermore, in the drawings, identical or corresponding elements are appropriately denoted by the same reference numeral. It should also be noted that the drawings are schematic, and the dimensional relationships and proportions of each element may differ from reality. Even between drawings, there may be differences in dimensional relationships and proportions.

[0035] (Embodiment 1) [Outline Configuration of the Judgment System] Figure 1 is a schematic diagram showing the configuration of a judgment system including a judgment device according to Embodiment 1. The judgment system 1 uses the Raman spectrum detected by irradiating the gastrointestinal mucosal tissue with excitation light to determine the future risk of deterioration of the gastrointestinal mucosal tissue. The judgment system 1 of Embodiment 1 can determine the future risk of inflammation recurrence in gastrointestinal mucosal tissue that has achieved endoscopic remission, i.e., a mucosal healing state, through remission maintenance treatment for inflammatory bowel disease, and can assist in diagnosis by physicians.

[0036] As shown in Figure 1, this determination system 1 comprises an optical fiber scope 2, a light source device 3, a photodetector 4, a determination device 5, a display device 6, and an endoscope 7 (see Figure 2).

[0037] Figure 2 shows how excitation light is irradiated onto the gastrointestinal mucosal tissue from the optical fiber scope shown in Figure 1. As shown in Figure 2, the endoscope 7 has a long insertion section that is inserted into the patient, and the optical fiber scope 2 is extended from the treatment instrument protrusion port 71 located at the tip of the insertion section. The endoscope 7 also irradiates illumination light from the tip of the insertion section, captures the reflected light, and displays an image of the inside of the patient's body on the display device 6.

[0038] The optical fiber scope 2 is inserted into the subject via the instrument channel of the endoscope 7, and the excitation light output by the light source device 3 irradiates the gastrointestinal mucosal tissue inside the subject, while also collecting Raman scattered light from the gastrointestinal mucosal tissue and guiding it to the photodetector 4.

[0039] Figure 3 shows the optical fiber scope shown in Figure 1 as viewed from the tip. As shown in Figure 3, a transmitting optical fiber 21 that irradiates the gastrointestinal mucosal tissue in the subject with excitation light is located at the center of the optical fiber scope 2, and three layers of receiving optical fibers 22 that receive Raman scattered light from the gastrointestinal mucosal tissue are arranged around the outer circumference of the transmitting optical fiber 21. This is just one example of the configuration of an optical fiber scope, and the number, arrangement, and diameter of the transmitting and receiving optical fibers can be selected in various ways. In addition, a lens may be placed at the tip of the optical fiber scope to refract the excitation light and the Raman scattered light from the gastrointestinal mucosal tissue, thereby controlling the light projection range and the Raman scattered light collection range in the gastrointestinal mucosal tissue to a desired range.

[0040] The light source device 3 irradiates the gastrointestinal mucosal tissue with excitation light of a predetermined wavelength. The predetermined wavelength is near-infrared light between 780 nm and 850 nm, for example, 785 nm, in order to reduce noise due to autofluorescence from the mucosal tissue and increase the signal-to-noise ratio (S / N ratio). However, the predetermined wavelength is not particularly limited and may be one or more monochromatic lights.

[0041] The photodetector 4 generates a Raman spectrum by spectrally measuring Raman scattered light from the gastrointestinal mucosal tissue. The photodetector 4 consists of a spectrometer, which includes, for example, a diffraction grating to spatially disperse the Raman scattered light according to its wavelength, and a photodetector that arranges photodetectors in a one-dimensional or two-dimensional array. The photodetector is a CCD (Charge Coupled Device) or a CMOS sensor. The photodetector 4 generates a Raman spectrum by spectrally measuring the Raman scattered light from the optical fiber scope 2. The judgment device 5 will be described later.

[0042] The display device 6 is a display using liquid crystal or organic EL (Electroluminescence) and displays endoscopic images captured by the endoscope 7.

[0043] [Configuration of the Determination Device] The determination device 5 uses the Raman spectrum acquired from the photodetector 4 to determine the future risk of relapse in gastrointestinal mucosal tissue that has achieved endoscopic remission in the treatment of inflammatory bowel disease. The determination device 5 corresponds to the processor according to this disclosure. This determination device 5 is implemented using a processor having hardware such as an FPGA (Field-Programmable Gate Array), GPU (Graphics Processing Unit), or CPU (Central Processing Unit), and memory, which is a temporary storage area used by the processor.

[0044] Figure 4 is a block diagram showing the configuration of the determination device shown in Figure 1. As shown in Figure 4, the determination device 5 comprises an acquisition unit 51, a calculation unit 52, a determination unit 53, an output unit 54, a control unit 55, and a storage unit 56.

[0045] The acquisition unit 51 irradiates the gastrointestinal mucosal tissue with excitation light of a predetermined wavelength from the light source device 3 via the optical fiber scope 2 and acquires the detected Raman spectrum from the photodetector 4.

[0046] The calculation unit 52 calculates a first feature quantity that represents the state or amount of substance contained in the mucosal epithelium from the Raman spectrum of the first wavenumber band. The first feature quantity represents, for example, the modification state of the O-linked glycans of mucin. The glycan modification state represents the amount of glycans, the composition of monosaccharide molecules contained in the glycans, the amount of sulfate groups attached to the glycan ends, etc. The first wavenumber band is set according to the monosaccharide molecules contained in the O-linked glycans of mucin and the types of functional groups attached to the glycan ends, such as sulfate groups. The first feature quantity may also represent the amount of phospholipids. In this case, the first wavenumber band is set according to the functional groups contained in the phospholipids. While it is possible to use the area intensity of the Raman spectrum of the first wavenumber band for the first feature quantity, it is more appropriate to use the amount obtained by dividing it by the area intensity of the Raman spectrum of the wavenumber band containing the Raman peak derived from aromatic amino acids that constitute the protein, in order to generate a highly accurate detector in the generation of the detector described later.

[0047] The determination unit 53 determines the future risk of deterioration of the gastrointestinal mucosal tissue based on the first feature. Specifically, the determination unit 53 determines the future risk of inflammatory bowel disease recurrence of the gastrointestinal mucosal tissue using a determination device generated with the first feature for multiple patients, each of which is labeled to indicate whether or not inflammation has recurred in the gastrointestinal mucosal tissue.

[0048] The output unit 54 outputs the judgment result. Specifically, the output unit 54 outputs an image signal that displays the judgment result on the display device 6.

[0049] The control unit 55 controls the entire determination device 5.

[0050] The memory unit 56 sequentially stores the Raman spectra acquired by the acquisition unit 51. The memory unit 56 also stores the determination device used by the determination unit 53 for determination.

[0051] [Determination Method] Next, a determination method using the determination system 1 will be described. Figure 5 is a flowchart showing an overview of the process performed by the determination device according to Embodiment 1. As shown in Figure 5, the acquisition unit 51 irradiates the gastrointestinal mucosal tissue with excitation light of a predetermined wavelength from the light source device 3 via the optical fiber scope 2 and acquires the detected Raman spectrum from the photodetector 4 (step S1). It is preferable to first confirm the areas where inflammation has occurred in the mucosal tissue using intracellular images taken with the endoscope 7 and acquire Raman spectra at several locations.

[0052] Figure 6 shows an example of a Raman spectrum in colonic mucosal tissue. As shown in Figure 6, the acquisition unit 51 acquires a Raman spectrum in which the vertical axis represents the light intensity of Raman scattered light and the horizontal axis represents the Raman shift.

[0053] Next, the calculation unit 52 calculates a first feature quantity representing the O-linked glycosylation state of mucin from the Raman spectrum of the first wavenumber band (step S2). The first wavenumber band can be selected from one or more wavenumber bands A, B, C, and D in Figure 6. The first feature quantity is obtained by dividing the area intensity of the Raman spectrum of any one of wavenumber bands A, B, C, and D by the area intensity of the Raman spectrum of the wavenumber band containing the Raman peak derived from the amino acids of the protein. However, multiple feature quantities may be calculated using multiple wavenumber bands.

[0054] In the first waveband, the area intensity of the Raman spectrum in wavebands A, B, C, and D can vary depending on the type and composition of monosaccharide molecules and functional groups such as sulfate groups attached to the sugar chain ends of the mucin (i.e., the relative ratio of monosaccharides and functional groups such as sulfate groups attached to the sugar chain ends). Therefore, by calculating multiple features using multiple wavebands, it is possible to characterize the glycan modification state of mucin in the subject's gastrointestinal mucosal tissue from more multifaceted perspectives than by using only one type of feature, and a more discriminative classifier can be generated in the generation of the classifier described later. In this way, it may be possible to predict the future risk of deterioration of gastrointestinal mucosal tissue with high accuracy.

[0055] Waveband A is 780 cm －１ More than 900cm－１ The following wavenumber range includes contributions from Raman peaks of several glycan constituent molecules, such as N-acetylglucosamine (GlucNAc), N-acetylgalactosamine (GalNAc), galactose, and fucose.

[0056] Waveband B is 1020 cm －１ 1160cm or more －１ The following wavenumber range includes contributions from Raman peaks of several glycan constituent molecules, such as sulfate groups at the O-linked glycan termini, N-acetylglucosamine, N-acetylgalactosamine, galactose, glucose, fucose, mannose, and sialic acid.

[0057] Waveband C is 1320 cm －１ More than 1400cm －１ The following wavenumber ranges are observed, including contributions from the Raman peaks of N-acetylglucosamine and sialic acid.

[0058] Waveband D is 1400 cm －１ More than 1500cm －１ The following wavenumber ranges are observed, including contributions from the Raman peaks of N-acetylglucosamine, N-acetylgalactosamine, sialic acid, and fucose.

[0059] Furthermore, as a first feature, the area intensity of one of the Raman spectra in wavenumber bands A, B, C, and D is calculated, and this area intensity is then normalized by dividing it by the area intensity of the Raman spectrum in wavenumber band E, which originates from mucosal components different from the O-linked glycans of mucin. As the area intensity of the Raman spectrum in wavenumber band E, it is appropriate to use the area intensity of the Raman spectrum in a wavenumber band that includes Raman peaks derived from amino acids of proteins constituting the mucosa. Wavenumber band E in Figure 6 shows an example of a wavenumber band set in this way. Here, wavenumber band E is 990 cm⁻¹. －１ 1010cm or more －１ The following wavenumber ranges are present, and include Raman peaks derived from the aromatic ring respiratory oscillations of amino acids (aromatic amino acids) that make up proteins.

[0060] One way to set the frequency band E in this way is 1200 cm. －１ More than 1300cm －１The following wavenumber band may be set to include the Raman peak derived from the amide III vibration of the polypeptide backbone of the amino acids that make up the protein, and 1620 cm². －１ More than 1700cm －１ The following wavenumber band may be defined as a wavenumber band that includes the Raman peaks originating from the amide I vibrations of the polypeptide backbone of the amino acids that make up the protein.

[0061] By the method described above, the area intensity of one of the wavenumber bands A, B, C, or D, which includes the Raman peak of the monosaccharide component constituting the O-linked glycans of mucin, can be used to accurately evaluate the O-linked glycan modification state of mucin by correcting for variations in the area intensity of the Raman spectrum caused by differences in the measurement area of ​​the Raman spectrum on the mucosa, differences in the relative spatial arrangement between the optical fiber scope 2 for acquiring Raman scattered light and the mucosa, such as the measurement conditions of the Raman spectrum caused by the angle at which the optical fiber scope 2 is applied to the mucosa, or individual differences in the amount of mucin secreted from the mucosa.

[0062] Furthermore, using Raman microscopy analysis, it is possible to obtain a two-dimensional distribution image (Raman image) and Raman spectra of the chemical components constituting the gastrointestinal mucosal tissue by applying non-negative constraint matrix factorization analysis to numerous Raman spectral data (hyperspectral data) obtained from observation points at different locations in thinly sectioned gastrointestinal mucosal tissue. The inventors have confirmed that, in the Raman spectra of predetermined components of colonic mucosal tissue obtained by the above-described Raman microscopy analysis, there are Raman peaks in wavenumber bands A, B, C, and D that are attributed to monosaccharide molecules constituting the O-linked glycans of mucin and functional groups at the glycan terminals, and that the distribution of these predetermined components in the Raman image substantially matches the distribution of mucin stained with AB (Alucian Blue) staining in colonic mucosal tissue. Therefore, the Raman peaks in wavenumber bands A, B, C, and D can be used as markers to evaluate the O-linked glycosylation state of mucin (the amount of glycans modifying mucin and the components of those glycans).

[0063] Subsequently, the determination unit 53 uses a determination device stored in the memory unit 56 to determine, based on the first characteristic quantity, the future risk of inflammation recurrence in the gastrointestinal mucosal tissue that has achieved endoscopic remission in the treatment of inflammatory bowel disease (step S3).

[0064] For example, with respect to the discrimination boundary set in the classifier, if the feature exists in the positive region of the discrimination boundary, there is no risk of recurrence; if the feature exists in the negative region of the discrimination boundary, there is a risk of recurrence. This allows for a binary determination of the risk of recurrence.

[0065] Alternatively, if there are one or two types of features, the spatial distance between the feature corresponding to a point in the one-dimensional or two-dimensional feature space and the discrimination boundary corresponding to a line is calculated. If there are three types of features, the spatial distance between the feature corresponding to a point in the three-dimensional feature space and the discrimination boundary corresponding to a plane is calculated. Here, the relapse risk may be stratified into multiple levels according to the spatial distance between the feature and the discrimination boundary, such as low relapse risk if there is a spatial distance of a predetermined value or more between the feature on the positive side of the discrimination boundary and the discrimination boundary, medium relapse risk if the spatial distance between the feature and the discrimination boundary is less than or equal to a predetermined value, and high relapse risk if there is a spatial distance of a predetermined value or more between the feature on the negative side of the discrimination boundary and the discrimination boundary, and the risk may be determined step by step. The classifier will be described later.

[0066] The output unit 54 then outputs the determination result as an image signal to be displayed on the display device 6 (step S4).

[0067] According to Embodiment 1 described above, the calculation unit 52 calculates a first feature quantity, the determination unit 53 determines the future risk of inflammatory bowel disease recurrence of gastrointestinal mucosal tissue based on the first feature quantity, and the output unit 54 outputs the determination result. Therefore, the future risk of deterioration of gastrointestinal mucosal tissue can be determined with high accuracy, and the diagnosis by a physician can be assisted.

[0068] [Method for generating the judgment device] Next, the method for generating the judgment device used by the judgment unit 53 to determine the risk of relapse of inflammatory bowel disease will be explained. Figure 7 is a flowchart showing the method for generating the judgment device. As shown in Figure 7, if a subject (patient) is diagnosed with ulcerative colitis by a physician, remission induction treatment with drugs, etc., is performed to induce remission of ulcerative colitis (step S11).

[0069] During the continuation of treatment after achieving clinical remission through induction therapy, the subject is subjected to an endoscopy 7, and the physician diagnoses that the subject has achieved endoscopic remission (mucosal healing) of ulcerative colitis (Step S12).

[0070] At this time, the area where the inflammation of the mucosa due to ulcerative colitis has subsided is identified using intracellular imaging, and the Raman spectrum of the identified area in the colonic mucosa of the subject is measured. Alternatively, a tissue biopsy may be performed on the identified area in the colonic mucosa of the subject, and the Raman spectrum of the excised biopsy tissue may be measured outside the body (step S13).

[0071] Then, the acquisition unit 51 acquires the measured Raman spectrum, and the calculation unit 52 calculates a first feature from the acquired Raman spectrum (step S14). The calculated first feature is stored in the storage unit 56.

[0072] The colonic mucosa, the gastrointestinal mucosal tissue of the large intestine, has a mucus and mucosal barrier function that protects the mucosa from bacteria and foreign substances present in the intestinal lumen. When this barrier function breaks down, bacteria and harmful chemicals in the intestinal tract invade the mucosa, which is considered one of the causes of inflammatory bowel disease. Mucin, which is secreted as mucus from mucosal epithelial cells, is a protein containing O-linked glycans, and it has been reported that the amount and modification state of these glycans greatly affect the expression of the mucus-mediated mucosal barrier function. Therefore, the calculation unit 52 calculates a first feature quantity that represents the O-linked glycan modification state of mucin and evaluates the state of the mucus-mediated mucosal barrier function in the colonic mucosa.

[0073] Subsequently, the doctor conducts follow-up observations for X years (for example, 1 to 3 years) (Step S15), and the doctor diagnoses whether or not the ulcerative colitis has relapsed (Step S16).

[0074] Then, when the acquisition unit 51 acquires the diagnostic result, a first feature quantity is generated (step S17) to which a label indicating whether or not inflammatory bowel disease has relapsed has been assigned to the gastrointestinal mucosal tissue, and it is stored in the memory unit 56.

[0075] These processes are performed on multiple subjects, and once labeled first feature quantities for multiple subjects are accumulated in the storage unit 56, a classifier can be generated using the labeled first feature quantities for multiple subjects (step S18).

[0076] (Example 1) Figure 8 shows the feature data calculated from the accumulated Raman spectra. As shown in Figure 8, if M features are obtained from K subjects for each subject, this data can be represented as a K × M matrix. Each of the M features corresponds to a quantity normalized by dividing the area intensity of different wavenumber bands (wavenumber bands A, B, C, D) by the area intensity of wavenumber band E, and there may be one or more. As mentioned above, multiple types of wavenumber bands, including the Raman peaks of protein amino acids, can be set for wavenumber band E. In addition, for each data point, a label is assigned to each of the K subjects from which M features have been obtained, indicating whether or not ulcerative colitis has relapsed. This label is, for example, represented as 1 for relapse and -1 for non-relapse. By performing machine learning using this labeled data as training data, a classifier for determining the risk of relapse can be generated. The machine learning method is not particularly limited, and supervised machine learning methods such as Fisher's linear discriminant analysis, logistic regression analysis, and support vector machines can be used.

[0077] Furthermore, the feature XA (first feature) in Figure 8 is set to wavenumber band 1112 cm. －１ 1160cm or more －１ The area intensity of the Raman spectrum below (part of waveband B) is measured in waveband 997 cm⁻¹. －１ 1005cm －１The following (part of waveband E) Raman spectrum is divided by the area intensity to obtain the feature quantity XB (first feature quantity), which is calculated for waveband 1463 cm⁻¹. －１ 1498cm －１ The area intensity of the Raman spectrum below (part of waveband D) is measured in waveband 997 cm⁻¹. －１ 1005cm －１ The following quantities were used, divided by the area intensity of the Raman spectrum. Labeled features indicating whether ulcerative colitis relapsed or not were used as training data to determine the discrimination boundary using Fisher's linear discriminant.

[0078] Figure 9 shows the wavenumber band used to calculate features in the Raman spectrum of the colonic mucosal tissue. The wavenumber band BA1 shown in Figure 9 is 997 cm⁻¹. －１ 1005cm －１ Below, the wavenumber band BA3 is 1112 cm. －１ 1160cm or more －１ The following waveband BA5 is 1463 cm. －１ 1498cm －１ The following applies: Feature XA is the amount obtained by dividing the area intensity of waveband BA3 by the area intensity of waveband BA1. Feature XB is the amount obtained by dividing the area intensity of waveband BA5 by the area intensity of waveband BA1.

[0079] Here, we will explain the method for selecting the wavenumber band when calculating the feature quantities, and what information the feature quantities calculated based on the wavenumber band selected by this method reflect regarding the state or amount of substances contained in the gastrointestinal mucosa. Figure 10 shows a microscopic image of mucosa collected from a subject. Microscopic image I1 is an image taken from colon mucosal tissue collected from the subject. An observation area A1 is then set within microscopic image I1.

[0080] Next, a Raman image of observation region A1 is acquired using a Raman microscope. In Raman imaging analysis, based on the molecular-specific Raman spectrum and the spatial distribution information of the components corresponding to that Raman spectrum, substances contained in observation region A1 can be identified and their spatial distribution obtained. Figure 11 shows the distribution of aromatic amino acids in observation region A1 and approximately represents the distribution of proteins in observation region A1. In the protein distribution DP shown in Figure 11, the areas closer to white represent areas containing a large amount of protein. Figure 12 shows the distribution of mucin in observation region A1. In the mucin distribution DM shown in Figure 12, the areas closer to white represent areas containing a large amount of mucin.

[0081] Furthermore, the mean Raman spectrum (average value of the Raman spectrum) in the mucin-containing region can be obtained from the mucin distribution DM. In addition, by using hyperspectral imaging analysis methods such as non-negative constraint matrix factorization, the Raman spectrum of mucin in the mucosa can be extracted from the mean Raman spectrum in the mucin-containing region, with the contribution of Raman spectra of principal components other than mucin removed. Figure 13 shows the Raman spectrum of the colonic mucosa and the Raman spectrum of mucin included in the Raman spectrum of the colonic mucosal tissue obtained by the method described above. The solid line in Figure 13 represents the Raman spectrum of the entire observation region A1, i.e., the colonic mucosa, and the dashed line represents the Raman spectrum of mucin included in the Raman spectrum of the colonic mucosa. From the Raman spectrum of mucin contained in the colonic mucosa, represented by the dashed line, the area intensity in wavenumber band BA3 (BA3-mucin) is calculated. From the Raman spectrum of the colonic mucosa, represented by the solid line, the area intensity in wavenumber band BA1 (BA1-mucosa) is calculated. Then, the former area intensity (BA3-mucin) is divided by the latter area intensity (BA1-mucosa) to obtain a feature quantity (i.e., a feature quantity calculated using the Raman spectrum of mucin contained in the colonic mucosa, represented by the dashed line, that corresponds to feature quantity XA calculated using the Raman spectrum of the colonic mucosa, represented by the solid line). This feature quantity YA is defined as such. Since this feature quantity YA changes directly depending on the amount of sugar chains bound to the mucin contained in the mucosa and the composition of monosaccharide molecules constituting the sugar chains, it can be treated as a quantity representing the sugar chain modification state of mucin.

[0082] Figure 14 shows the correlation between feature quantity XA and the glycosylation state of mucin (feature quantity YA). The horizontal axis of Figure 14 represents feature quantity XA calculated from the Raman spectrum of a predetermined volume of colonic mucosa obtained by endoscopic observation, and the vertical axis represents feature quantity YA, which represents the glycosylation state of mucin (feature quantity XA calculated from the Raman spectrum of mucin included in the Raman spectrum of the colonic mucosa, which can be said to be feature quantity XA calculated from the Raman spectrum of mucin contained in the mucosa obtained by Raman microscopy analysis). From Figure 14, the correlation coefficient between feature quantity XA and the glycosylation state of mucin is 0.91, indicating that feature quantity XA calculated from the Raman spectrum of colonic mucosa obtained by endoscopic observation has a remarkably strong correlation with the glycosylation state of mucin.

[0083] Similar to feature vector XA, calculations were also performed for feature vector XB. Figure 15 shows the correlation between feature vector XB and the glycosylation state of mucin. The horizontal axis of Figure 15 represents feature vector XB calculated from the Raman spectrum obtained by endoscopic observation, and the vertical axis represents feature vector YB (feature vector XB calculated from the Raman spectrum of mucin obtained by Raman microscopy), which represents the glycosylation state of mucin. From Figure 15, the correlation coefficient between feature vector XA and the glycosylation state of mucin is 0.86, indicating that feature vector XB has a strong correlation with the glycosylation state of mucin. Thus, by appropriately setting the wavenumber band for calculating features in the Raman spectrum of the colonic mucosa, as shown by the solid line in Figure 9 or Figure 13, it is possible to calculate feature vectors XA and XB that have a strong correlation with the glycosylation state of mucin.

[0084] Furthermore, similar to mucin, Raman imaging analysis can be used to obtain Raman spectra of phospholipid-containing regions from the distribution of phospholipids. Figure 16 shows the Raman spectra of phospholipids included in the Raman spectrum of colonic mucosal tissue. The solid line in Figure 16 represents the Raman spectrum of the entire observation region A1, and the dashed line represents the Raman spectrum of phospholipids contained in the colonic mucosa. From the Raman spectrum of phospholipids included in the Raman spectrum of the colonic mucosa, represented by the dashed line, the area intensity in wavenumber band BA3 (BA3-phospho) is calculated, and from the Raman spectrum of the colonic mucosa, represented by the solid line, the area intensity in wavenumber band BA1 (BA1-mucosa) is calculated. Then, the feature quantity XA obtained by dividing the former area intensity (BA3-phospho) by the latter area intensity (BA1-mucosa) is defined as feature quantity YA. Since this feature variable YA changes directly in response to the amount of phospholipids relative to the amount of protein contained in the colonic mucosa, it can be treated as a relative amount of phospholipids.

[0085] Figure 17 shows the correlation between feature quantity XA and the relative amount of phospholipids (feature quantity YA). The horizontal axis of Figure 17 represents feature quantity XA calculated from Raman spectra obtained by endoscopic observation, and the vertical axis represents feature quantity YA (feature quantity XA calculated from Raman spectra of phospholipids obtained by Raman microscopy), which represents the amount of phospholipids. From Figure 17, the correlation coefficient between feature quantity XA and phospholipid amount is 0.57, indicating that the correlation between feature quantity XA and phospholipid amount is far weaker than the correlation between feature quantity XA and the mucin glycosylation state.

[0086] Calculations were performed on feature XB in the same way as on feature XA. Figure 18 shows the correlation between feature XB and phospholipid content. The horizontal axis of Figure 18 represents feature XB calculated from Raman spectra obtained by endoscopic observation, and the vertical axis represents feature YB (feature XB calculated from Raman spectra of phospholipids obtained by Raman microscopy), which represents phospholipid content. From Figure 18, the correlation coefficient between feature XB and phospholipid content is 0.47, indicating that the correlation between feature XB and phospholipid content is far weaker than the correlation between feature XB and mucin glycosylation status.

[0087] As explained above, the wavenumber bands BA3 and BA5 in Example 1 were selected such that there is a strong correlation between feature quantities XA and XB and the modification state of mucin, and a weak correlation between feature quantities XA and XB and the amount of phospholipids. In this way, by selecting an appropriate wavenumber band in the calculation of feature quantities, it is possible to emphasize the contribution from specific substances (mucin glycans) contained in mucosal tissue.

[0088] (Specific example of a classifier 1) Using features XA and XB, a classifier can be generated to determine the risk of ulcerative colitis relapse in colonic mucosal tissue. Figure 19 shows a scatter plot in a two-dimensional feature space of features XA and XB calculated from the Raman spectra of the ulcerative colitis relapse group (19 cases) and the non-relapse group (45 cases). Furthermore, Figure 19 shows the discrimination boundary obtained by performing Fisser's linear discriminant method on the labeled features superimposed on the scatter plot. In Figure 19, the relapse group is shown as a square and the non-relapse group as a circle. Then, an ROC (Receiver Operating Characteristic) curve was created by changing the threshold of this discrimination boundary.

[0089] Figure 20 shows the ROC curves obtained by varying the threshold using Fisher's linear discriminant method. As shown in Figure 20, an AUC (Area Under the Curve) value of 0.843 was obtained for the created ROC curve. On the other hand, based on histological diagnostic evaluation indices such as the Geboes Score and the PiCasso Histological Remission Index (PHRI), an AUC value of 0.7 to 0.8 for the ROC curve has been reported as a predictive performance against relapse in ulcerative colitis patients who have achieved endoscopic remission (see, for example, Gui X, et al. Gut 2022;71:889-898). Therefore, this method yielded a sufficiently high AUC value that is comparable to or surpasses the predictive ability of histological methods for predicting ulcerative colitis relapse, demonstrating that it is possible to accurately determine the risk of ulcerative colitis relapse in colorectal mucosal tissue.

[0090] (Specific Example of a Classifier 2) Figure 21 shows a table for classification. As shown in Figure 21, a table may be generated to determine the risk of relapse from features as a classifier. Specifically, if threshold L1 ≤ feature XA, the score is SA1, and the risk of relapse is determined to be low according to this score SA1. Similarly, if threshold L2 ≤ feature XA < threshold L1, the score is SA2, and the risk of relapse is determined to be medium according to this score SA2. Similarly, if feature XA < threshold L2, the score is SA3, and the risk of relapse is determined to be high according to this score SA3. These thresholds L1 and L2 are predetermined values ​​using labeled first features for multiple subjects.

[0091] (Specific example of a classifier 3) Figure 22 shows the scores of two features. As shown in Figure 22, the risk of recurrence may be determined using two features XA and XB. The method for calculating the score for feature XA may be the same as in the example shown in Figure 19. For feature XB, scores SB1 to SB3 are assigned according to the size of feature XB for thresholds M1 and M2. The total score T is the sum of the scores of feature XA and feature XB.

[0092] Figure 23 shows a table for making a determination. As shown in Figure 23, if threshold Z1 ≤ total score T, the risk of relapse is determined to be low; if threshold Z2 ≤ total score T < threshold Z1, the risk of relapse is determined to be medium; and if total score T < threshold Z2, the risk of relapse is determined to be high. A similar method may be used to generate a determination table from three or more features.

[0093] (Example 2) Figure 24 shows the wavenumber band used to calculate features in the Raman spectrum of colonic mucosal tissue. The wavenumber band BA1 shown in Figure 24 is 997 cm⁻¹. －１ 1005cm －１ Below (part of waveband E), waveband BA13 is 1050 cm. －１ 1089cm or more －１ Below (part of waveband B), waveband BA15 is 1427 cm. －１ 1462cm or more －１The following represents a portion of waveband D. In Example 2, feature quantity XA is the amount obtained by dividing the area intensity of waveband BA13 by the area intensity of waveband BA1. Feature quantity XB is the amount obtained by dividing the area intensity of waveband BA15 by the area intensity of waveband BA1.

[0094] Figure 25 shows the Raman spectrum of mucin contained in the Raman spectrum of the colonic mucosal tissue. The solid line in Figure 25 represents the Raman spectrum of the entire observation area A1 (i.e., the colonic mucosa), and the dashed line represents the Raman spectrum of mucin contained in the entire observation area A1.

[0095] Figure 26 shows the correlation between feature quantity XA and the glycosylation status of mucin. The horizontal axis of Figure 26 represents feature quantity XA calculated from Raman spectra obtained by endoscopic observation, and the vertical axis represents feature quantity YA representing the glycosylation status of mucin (feature quantity XA calculated from Raman spectra of mucin contained in the colonic mucosa, which can be obtained by Raman microscopy analysis). From Figure 26, the correlation coefficient between feature quantity XA and feature quantity YA representing the glycosylation status of mucin is 0.76, indicating that feature quantity XA has a weaker correlation with the glycosylation status of mucin than in Example 1.

[0096] Figure 27 shows the Raman spectrum of phospholipids contained in the Raman spectrum of the colonic mucosal tissue. The solid line in Figure 27 represents the Raman spectrum of the entire observation area A1 (i.e., the colonic mucosa), and the dashed line represents the Raman spectrum of phospholipids contained in the entire observation area A1.

[0097] Figure 28 shows the correlation between feature quantity XA and phospholipid quantity (the amount of phospholipid relative to the proteins in the colonic mucosa). The horizontal axis of Figure 28 represents feature quantity XA calculated from the Raman spectrum of the colonic mucosa obtained by endoscopic observation, and the vertical axis represents feature quantity YA (feature quantity XA calculated from the Raman spectrum of phospholipids obtained by Raman microscopy), which represents the amount of phospholipid. From Figure 28, the correlation coefficient between feature quantity XA and phospholipid quantity is 0.62, indicating that feature quantity XA has a stronger correlation with phospholipid quantity than in Example 1.

[0098] As explained above, the wavenumber bands BA13 and BA15 in Example 2 show a weaker correlation between feature quantities XA and XB and the modification state of mucin, and a stronger correlation between feature quantities XA and XB and the amount of phospholipids, compared to the wavenumber bands BA3 and BA5 in Example 1. In other words, wavenumber bands BA13 and BA15 are selected to further weaken the correlation between feature quantities XA and XB and the glycosylation state of mucin, and to further strengthen the correlation between feature quantities XA and XB and the amount of phospholipids.

[0099] Figure 29 shows a scatter plot in a two-dimensional feature space of feature quantities XA and XB, calculated from the Raman spectra of the ulcerative colitis relapse group (19 cases) and the non-relapse group (45 cases). Furthermore, Figure 29 shows the discrimination boundary obtained by performing Fisser's linear discriminant method on the labeled features, superimposed on the scatter plot. In Figure 29, the relapse group is shown as a square and the non-relapse group as a circle. Then, ROC curves were created by changing the threshold of this discrimination boundary.

[0100] Figure 30 shows the ROC curves obtained by varying the threshold using Fisher's linear discriminant method. As shown in Figure 30, an AUC value of 0.856 was obtained for the created ROC curve. In Example 2, a higher AUC value was obtained than in Example 1, demonstrating that the risk of ulcerative colitis recurrence in the colonic mucosal tissue can be determined with high accuracy.

[0101] (Example 3) Figure 31 is a diagram showing the wavenumber band used to calculate feature quantities in the Raman spectrum of colonic mucosal tissue. The wavenumber band BA1 shown in Figure 31 is 997 cm⁻¹. －１ 1005cm －１ Below (part of waveband E), waveband BA23 is 1049 cm. －１ 1160cm or more －１ Below (part of waveband B), waveband BA25 is 1408 cm. －１ 1498cm －１ The following represents a portion of waveband D. In Example 3, feature quantity XA is the amount obtained by dividing the area intensity of waveband BA23 by the area intensity of waveband BA1. Feature quantity XB is the amount obtained by dividing the area intensity of waveband BA25 by the area intensity of waveband BA1.

[0102] Figure 32 shows the Raman spectrum of mucin included in the Raman spectrum of colonic mucosal tissue. The solid line in Figure 32 represents the Raman spectrum of the entire observation region A1, and the dashed line represents the Raman spectrum of mucin included in the Raman spectrum of the entire observation region A1.

[0103] Figure 33 shows the correlation between feature quantity XA and the glycosylation status of mucin. The horizontal axis of Figure 33 represents feature quantity XA calculated from Raman spectra obtained by endoscopic observation, and the vertical axis represents feature quantity YA, which represents the glycosylation status of mucin (feature quantity XA calculated from Raman spectra of mucin contained in mucosal tissue obtained by Raman microscopy). From Figure 33, the correlation coefficient between feature quantity XA and feature quantity YA, which represents the glycosylation status of mucin, is 0.84, indicating that feature quantity XA has a stronger correlation with the glycosylation status of mucin than in Example 2.

[0104] Figure 34 shows the Raman spectrum of phospholipids included in the Raman spectrum of the colonic mucosal tissue. The solid line in Figure 34 represents the Raman spectrum of the entire observation region A1, and the dashed line represents the Raman spectrum of phospholipids included in the Raman spectrum of the entire observation region A1.

[0105] Figure 35 shows the correlation between feature quantity XA and phospholipid quantity. The horizontal axis of Figure 35 represents feature quantity XA calculated from Raman spectra obtained by endoscopic observation, and the vertical axis represents feature quantity YA (feature quantity XA calculated from Raman spectra of phospholipids obtained by Raman microscopy), which represents phospholipid quantity. From Figure 35, the correlation coefficient between feature quantity XA and phospholipid quantity is 0.62, indicating that feature quantity XA has a similar correlation with phospholipid quantity as in Example 2.

[0106] As explained above, the wavenumber bands BA23 and BA25 in Example 3 show a stronger correlation between feature quantities XA and XB and the glycosylation state of mucin than the wavenumber bands BA13 and BA15 in Example 2, and a stronger correlation between feature quantities XA and XB and phospholipid content than the wavenumber bands BA3 and BA5 in Example 1. In other words, wavenumber bands BA23 and BA25 are selected to show a strong correlation between feature quantities XA and XB and both the glycosylation state of mucin and the phospholipid content.

[0107] Figure 36 shows a scatter plot in a two-dimensional feature space of feature quantities XA and XB, calculated from Raman spectra of the colonic mucosa in the relapse group (19 cases) and the non-relapse group (45 cases) of ulcerative colitis. Furthermore, Figure 36 shows the discrimination boundary obtained by performing Fisser's linear discriminant method on the labeled features, superimposed on the scatter plot. In Figure 36, the relapse group is shown as a square and the non-relapse group as a circle. Then, ROC curves were created by changing the threshold of this discrimination boundary.

[0108] Figure 37 shows the ROC curves obtained by changing the threshold using Fisher's linear discriminant method. As shown in Figure 37, an AUC value of 0.859 was obtained for the created ROC curve. In Example 3, a higher AUC value was obtained than in Example 2, demonstrating that the risk of ulcerative colitis relapse in the colonic mucosal tissue can be determined with high accuracy. Thus, by appropriately setting the wavenumber band in the calculation of feature quantities XA and XB based on the Raman spectrum of the colonic mucosa, the correlation between feature quantities XA and XB and both the glycosylation state of mucin and the amount of phospholipids can be strengthened, thereby enabling more accurate determination of the risk of ulcerative colitis relapse.

[0109] As in Examples 2 and 3, the wavenumber band may be selected so as to increase the correlation between feature quantities XA and XB and the amount of phospholipids. In other words, the first wavenumber band may be set according to the functional groups contained in the phospholipids. To put it another way, the first feature quantities, XA and XB, may represent the amount of phospholipids. In this case as well, the AUC value of the ROC curve is high, and the risk of ulcerative colitis relapse in the colonic mucosal tissue can be determined with high accuracy.

[0110] (Modification) In Embodiment 1, the calculation unit 52 calculates the area intensity by selecting one or more wave number bands A, B, C, and D in Figure 6 as the first feature quantity, but is not limited to this. The calculation unit 52 may calculate the first feature quantity using a part of wave number bands A, B, C, and D in Figure 6.

[0111] Figure 38 shows the Raman spectrum and modified wavenumber bands in the colonic mucosal tissue. As shown in Figure 38, the calculation unit 52 may calculate the area intensity of wavenumber band B1 as the first feature quantity. Wavenumber band B1 is 1120 cm⁻¹ －１ 1160cm or more －１ The following wavenumber range primarily includes contributions from monosaccharide molecules that constitute mature O-linked glycans, such as N-acetylglucosamine and sialic acid. By selecting this range, it is possible to obtain information that emphasizes a specific state (e.g., the degree of glycan modification) in the O-linked glycan modification state of mucin, enabling judgments based on more detailed information about the mucin glycan modification state.

[0112] (Embodiment 2) In Embodiment 2, the calculation unit 52 calculates a second characteristic quantity representing the amount of neutrophil mucosal infiltration from the Raman spectrum of the second wavenumber band. The second wavenumber band is wavenumber band F in Figure 6, which is 1550 cm⁻¹. －１ More than 1600cm －１ The following wavenumber bands include the Raman peak of myeloperoxidase contained in neutrophils. The second feature is the amount obtained by dividing the area intensity of the Raman spectrum in wavenumber band F by the area intensity of the Raman spectrum in the wavenumber band that includes the Raman peak derived from the amino acids that make up the protein.

[0113] According to Raman microscopy analysis, by applying non-negative matrix factorization analysis to numerous Raman spectral data (hyperspectral data) obtained from different observation points in thinly sectioned gastrointestinal mucosal tissue, it is possible to obtain a two-dimensional distribution image (Raman image) and Raman spectra of the chemical components constituting the gastrointestinal mucosal tissue. The inventors have confirmed that the Raman spectra of predetermined components of colonic mucosal tissue obtained by the above-described Raman microscopy analysis method match the Raman spectra of neutrophils isolated from blood, and that signals originating from myeloperoxidase, which is abundant in neutrophils, are observed in wavenumber band F of the Raman spectrum. Furthermore, the distribution of the predetermined component in the Raman image substantially matches the distribution of neutrophils stained by myeloperoxidase staining in colonic mucosal tissue. Therefore, the Raman peak in wavenumber band F can be used as a marker to evaluate neutrophil mucosal infiltration.

[0114] Furthermore, as a second feature, the area intensity of the Raman spectrum in waveband F is calculated and normalized by dividing it by the area intensity of the Raman spectrum in waveband G. Waveband G is 1630 cm⁻¹. －１ 1680cm or more －１ The following wavenumber bands are present, including contributions from Raman peaks originating from amide I vibrations of the protein polypeptide backbone. The wavenumber band G is 990 cm⁻¹. －１ 1010cm or more －１ It is also possible to select a wavenumber band. This wavenumber band includes contributions from Raman peaks originating from aromatic ring respiratory oscillations of aromatic amino acids that make up proteins.

[0115] The determination unit 53 then determines the future risk of deterioration of the gastrointestinal mucosal tissue based on the first and second features. Specifically, the determination unit 53 sets a predetermined threshold for the Raman spectral feature XB (second feature) that reflects the amount of neutrophil infiltration. Figure 39 shows how the threshold is set. In Figure 39, the relapse group is shown as a square and the non-relapse group as a circle. As shown in Figure 39, data where feature XB is above the threshold TH is excluded. The Raman spectrum of neutrophils has a signal intensity in waveband B that is about the same as that of mucin glycans. Therefore, when neutrophil mucosal infiltration is significant, the area intensity of the Raman spectrum of neutrophils is superimposed on the area intensity of the Raman spectrum in waveband B. As a result, there is a disadvantage in that the mucin glycan modification state cannot be properly evaluated. The reason for excluding data where feature XB is above the threshold TH is to reduce the influence of the Raman spectrum of neutrophils as noise on the first feature. Subsequently, the determination unit 53 generates a classifier using the remaining data, similar to the first embodiment. Furthermore, it creates an ROC curve for the generated classifier.

[0116] Figure 40 shows the ROC curves obtained by varying the threshold using Fisher's linear discriminant method. As shown in Figure 40, an AUC value of 0.95 was obtained for the created ROC curve. The fact that an even higher AUC value was obtained than in Embodiment 1 indicates that by considering the amount of neutrophil mucosal infiltration, the future risk of deterioration of gastrointestinal mucosal tissue can be determined with even greater accuracy.

[0117] (Embodiment 3) In Embodiment 3, the determination unit 53 determines the future risk of cancer in the gastrointestinal mucosal tissue using a determination unit generated with first feature quantities for a plurality of patients who have been assigned labels indicating whether or not the gastrointestinal mucosal tissue of the stomach or esophagus has become cancerous.

[0118] In the gastric or esophageal mucosa, prolonged inflammation can cause a change in the mucosa, transforming it into a mucosa similar to that of the intestinal mucosa (intestinal metaplasia). Intestinal metaplasia mucosa is known to be a mucosa with a high risk of becoming cancerous. The type of mucin secreted by mucosal epithelial cells differs between intestinal metaplasia mucosa and normal mucosa.

[0119] The mucins secreted from normal gastric mucosa are MUC5AC and MUC6, which are neutral mucins (gastric mucins) with sugar chains containing few sialic acid and sulfate groups. On the other hand, the mucins secreted from intestinal metaplastic mucosa mainly contain MUC2, which is an acidic mucin (intestinal mucin) with sugar chains containing many sialic acid and sulfate groups.

[0120] Furthermore, it has been reported that changes in the O-linked glycosylation state of mucin may occur due to or during carcinogenesis, caused by abnormalities in the function of the Golgi apparatus in cells that attach sugar chains to mucin. For example, changes in the amount of sialic acid or sulfate groups at the end of the sugar chain, changes in the monosaccharide composition constituting the sugar chain, or the development of immature sugar chains may occur.

[0121] Based on the above facts and considerations, we conceived the idea that a first feature representing the O-linked glycosylation state of mucin can be calculated using Raman spectroscopy, and that this first feature can be used to determine the future risk of cancer in gastrointestinal mucosal tissue.

[0122] In Embodiment 3, the first waveband can be selected from one or more wavebands A, B, C, and D in Figure 6. Furthermore, a different waveband may be selected as the first waveband compared to Embodiment 1.

[0123] Figure 41 shows an example of a Raman spectrum in the mucosa of the large intestine. The wavenumber band D2 shown in Figure 41 is 1455 cm⁻¹. －１ More than 1500cm －１ The following wavenumber range includes contributions from the Raman peaks of N-acetylglucosamine and sialic acid, monosaccharide molecules that constitute the O-linked glycans of mucin. The intensity of the Raman peak in this wavenumber range increases as the O-linked glycans of mucin mature.

[0124] It is preferable to calculate the area intensity of the Raman spectrum in wave number band D2 as the first feature and normalize it by dividing it by the area intensity of the Raman spectrum in wave number band D1. Wave number band D1 is 1425 cm⁻¹. －１ 1455cm －１This wavenumber band includes contributions from Raman peaks of N-acetylgalactosamine, a monosaccharide molecule that constitutes O-linked glycans, and lipids such as phospholipids, which are mucus components other than mucin glycans, but does not include significant contributions from the Raman peaks of N-acetylglucosamine and sialic acid. Therefore, the amount obtained by dividing the area intensity of the Raman spectrum in wavenumber band D2 by the area intensity of wavenumber band D1 shows a larger value as the O-linked glycans of mucin mature. This is thought to be a marker that reflects abnormalities in mucin glycosylation induced by functional abnormalities of goblet cells during the carcinogenesis process of mucosal tissue.

[0125] The first feature is the wavenumber band B1 of 1120 cm in Figure 38. －１ 1160cm or more －１ The area intensity of the Raman spectrum included below may be divided by the area intensity of the Raman spectrum in wave number band E. Since wave number band B1 includes contributions from monosaccharide molecules that constitute mature O-linked glycans such as N-acetylglucosamine and sialic acid, this characteristic value will be larger as the O-linked glycans of mucin mature.

[0126] In Embodiment 3, as in Embodiment 1, first feature quantities are accumulated for multiple patients who have been labeled to indicate whether or not their gastrointestinal mucosal tissue of the stomach or esophagus has become cancerous. A classifier can then be generated using machine learning or the like with these labeled first feature quantities for multiple patients. The classification unit 53 can then use the generated classifier to determine the future risk of cancer in the gastrointestinal mucosal tissue, thereby assisting a physician's diagnosis.

[0127] (Embodiment 4) In Embodiment 4, the determination unit 53 uses a determination unit generated using a first feature quantity for a plurality of patients to which a label indicating whether or not gastroesophageal reflux disease has recurred has been assigned to the gastrointestinal mucosal tissue to determine the future risk of recurrence of gastroesophageal reflux disease in the gastrointestinal mucosal tissue.

[0128] In gastroesophageal reflux disease (GERD), acidic contents from the stomach reflux into the esophagus and remain there, resulting in damage and inflammation of the esophageal mucosa. Proton pump inhibitors (PPIs) are used to treat GERD, but in cases of refractory GERD, they need to be taken for a long period of time.

[0129] Proton pump inhibitors have the effect of suppressing gastric acid secretion, but this has been linked to risks of side effects such as intestinal infections due to disruption of the gut microbiota, changes in drug absorption in the stomach due to drug interactions, and vitamin and mineral deficiencies. Therefore, it is considered desirable to change, reduce, or temporarily discontinue the medication at an appropriate time during long-term maintenance treatment. On the other hand, there is a possibility that inflammation may recur as a result of changing, reducing, or temporarily discontinuing the medication.

[0130] Therefore, by calculating a first feature representing the O-linked glycosylation state of mucin using Raman spectroscopy, and using this first feature to determine the future risk of recurrence of gastroesophageal reflux disease in the gastrointestinal mucosal tissue, it is possible to assist in decisions regarding changes, reductions, or even temporary discontinuation of medications.

[0131] In Embodiment 4, the first wavenumber band can be selected from one or more wavenumber bands A, B, C, and D in Figure 6.

[0132] In Embodiment 4, as in Embodiment 1, first feature quantities are accumulated for multiple patients whose gastrointestinal mucosal tissue is labeled to indicate whether or not gastroesophageal reflux disease has recurred. A classifier can then be generated using machine learning or the like with the labeled first feature quantities for multiple patients. The classification unit 53 can then use the generated classifier to determine the future risk of gastroesophageal reflux disease recurrence in the gastrointestinal mucosal tissue, thereby assisting the physician's diagnosis.

[0133] Further effects and modifications can be readily derived by those skilled in the art. Therefore, broader aspects of the present invention are not limited to the specific details and representative embodiments expressed and described above. Accordingly, various modifications are possible without departing from the spirit or scope of the overall concept of the invention as defined by the appended claims and their equivalents.

[0134] 1. Judgment system 2. Optical fiber scope 3. Light source device 4. Photodetector 5. Judgment device 6. Display device 7. Endoscope 21. Transmitting optical fiber 22. Receiving optical fiber 51. Acquisition unit 52. Calculation unit 53. Judgment unit 54. Output unit 55. Control unit 56. Storage unit 71. Treatment instrument protrusion port

Claims

1. A determination device comprising a processor that irradiates gastrointestinal mucosal tissue with excitation light of a predetermined wavelength and acquires a detected Raman spectrum, calculates a first characteristic quantity representing the state or amount of substances contained in the mucosal epithelium from the Raman spectrum in the first wavenumber band, determines the future deterioration risk of the gastrointestinal mucosal tissue based on the first characteristic quantity, and outputs the determination result.

2. The determination device according to claim 1, wherein the first feature quantity represents the glycosylation state of mucin.

3. The determination device according to claim 1, wherein the first characteristic quantity represents the amount of phospholipids.

4. The determination device according to claim 1, wherein the first characteristic quantity represents the glycosylation state of mucin and the amount of phospholipids.

5. The determination device according to claim 2, wherein the first wavenumber band is set according to the type of monosaccharide molecule contained in the sugar chain of the mucin and the type of functional group attached to the end of the sugar chain.

6. The determination device according to claim 1, wherein the first wavenumber band is set according to the functional groups contained in the phospholipid.

7. The determination device according to claim 1, wherein the processor determines the future risk of deterioration of the gastrointestinal mucosal tissue by determining the risk of future deterioration of the gastrointestinal mucosal tissue using a determination device generated using the first feature quantities for a plurality of patients to which labels indicating whether or not the gastrointestinal mucosal tissue has deteriorated in the future, relative to the time of Raman spectrum measurement of the gastrointestinal mucosal tissue.

8. The determination device according to claim 1, wherein the processor calculates a second feature quantity representing the amount of neutrophil mucosal infiltration from the Raman spectrum in the second wavenumber band, and determines the future risk of deterioration of the gastrointestinal mucosal tissue based on the first feature quantity and the second feature quantity.

9. The determination device according to claim 1, wherein the processor determines the future risk of inflammatory bowel disease recurrence of the gastrointestinal mucosal tissue by a determination device generated using the first feature quantities for a plurality of patients to whom the gastrointestinal mucosal tissue has been labeled to indicate whether or not inflammatory bowel disease has recurred.

10. The determination device according to claim 1, wherein the processor determines the future risk of cancer in the gastrointestinal mucosal tissue by a determination device generated using the first feature quantities for a plurality of patients to whom labels indicating whether or not the gastrointestinal mucosal tissue of the stomach or esophagus has become cancerous.

11. The determination device according to claim 1, wherein the processor determines the future risk of recurrence of gastroesophageal reflux disease in the gastrointestinal mucosal tissue using a determination device generated using the first feature quantities for a plurality of patients to which the gastrointestinal mucosal tissue has been labeled to indicate whether or not gastroesophageal reflux disease has recurred.

12. The determination device according to claim 8, wherein the processor determines the future risk of inflammatory bowel disease recurrence of the gastrointestinal mucosal tissue by a determination device generated using the first and second features of a plurality of patients to whom the gastrointestinal mucosal tissue has been labeled to indicate whether or not inflammatory bowel disease has recurred.

13. The determination device according to claim 1, wherein the first wavenumber band is a wavenumber band containing the Raman peak of N-acetylglucosamine, galactose, fucose, mannose, glucose, sulfate group, N-acetylgalactosamine, or sialic acid.

14. The aforementioned first waveband is 780 cm. －１ More than 900cm －１ The following wavenumber band: 1020 cm －１ 1160cm or more －１ The following waveband: 1320 cm －１ More than 1400cm －１ The following wavebands, or 1400 cm －１ More than 1500cm －１ The determination device according to claim 1, wherein the wavenumber band is as follows.

15. The determination device according to claim 8, wherein the second wavenumber band is a wavenumber band that includes the Raman peak of myeloperoxidase contained in neutrophils.

16. The second waveband is a waveband of 1550 cm －１ or more and 1600 cm －１ or less. The determination device according to claim 8 17. The determination device according to claim 1, wherein the first feature quantity is the amount obtained by dividing the area intensity of the Raman spectrum in the first wavenumber band by the area intensity of the Raman spectrum in the wavenumber band that includes a Raman peak derived from an amino acid constituting the protein.

18. The determination device according to claim 8, wherein the second feature quantity is the amount obtained by dividing the area intensity of the Raman spectrum in the second wavenumber band by the area intensity of the Raman spectrum in the wavenumber band that includes the Raman peak derived from the amino acids constituting the protein.

19. A determination system comprising: a light source device that irradiates gastrointestinal mucosal tissue with excitation light of a predetermined wavelength; a photodetector that spectrally measures Raman scattered light from the gastrointestinal mucosal tissue to generate a Raman spectrum; and a determination device that calculates a first characteristic quantity representing the state or amount of substance contained in the mucosal epithelium from the Raman spectrum in the first wavenumber band, determines the future deterioration risk of the gastrointestinal mucosal tissue based on the first characteristic quantity, and outputs the determination result.

20. The determination system according to claim 19, comprising: an endoscope inserted into a subject; and an optical fiber scope inserted into the subject via a treatment instrument channel of the endoscope, which irradiates the gastrointestinal mucosal tissue in the subject with the excitation light output by the light source device, and collects the Raman scattered light from the gastrointestinal mucosal tissue and guides it to the light detection device.

21. A determination method comprising: irradiating gastrointestinal mucosal tissue with excitation light of a predetermined wavelength and obtaining a detected Raman spectrum; calculating a first characteristic quantity representing the state or amount of substances contained in the mucosal epithelium from the Raman spectrum in the first wavenumber band; determining the future deterioration risk of the gastrointestinal mucosal tissue based on the first characteristic quantity; and outputting the determination result.

22. A determination program that causes a processor to perform the following actions: irradiate gastrointestinal mucosal tissue with excitation light of a predetermined wavelength to obtain a detected Raman spectrum; calculate a first feature quantity representing the state or amount of substances contained in the mucosal epithelium from the Raman spectrum in the first wavenumber band; determine the future deterioration risk of the gastrointestinal mucosal tissue based on the first feature quantity; and output the determination result.

Images