JP2013033410A - Method and system for supporting pathological diagnosis - Google Patents

Abstract

It is an object of the present invention to provide a system and method for supporting pathological diagnosis in order to obtain more consistent results as compared with conventional methods.
The present invention provides systems and methods that support pathological diagnosis to obtain more consistent results compared to conventional methods. The present invention also provides a method for determining pathological diagnostic criteria to obtain more consistent results. The present invention also provides systems and methods that support pathological diagnosis based on the criteria obtained by the present invention.
[Selection figure] None

Description

  The present invention relates to the field of pathological diagnosis. The present invention provides methods and systems that support pathological diagnosis.

  Obtaining consistent results in pathological diagnosis is generally difficult. In particular, for idiopathic interstitial pneumonia (IIP), it is still difficult to obtain a consistent pathological diagnosis of IIP, despite scientific progress following the publication of the 2002 ATS / ERS classification. is there. Difficulties in pathological diagnosis of diseases such as IIP are due to the lack of a common view on what diagnostic criteria should be used for pathological diagnosis and differences in individual interpretation.

  Accordingly, it is an object of the present invention to provide a system and method for supporting pathological diagnosis to obtain consistent results. The present invention also provides a method for determining pathological diagnostic criteria. The present invention also provides systems and methods that support pathological diagnosis based on the criteria obtained by the present invention.

(Item 1)
The following steps:
(A) selecting criteria candidates for a given disease;
(B) A step of obtaining a judgment result of a plurality of pathologists for the criteria candidate for the same pathological diagnosis sample, wherein the plurality of pathologists are a group of experts or experts and A process consisting of a group of specialized pathologists;
(C) selecting a sample in which more than 60% of the plurality of pathologists showed the same diagnosis; and
(D) performing a statistical analysis on the judgment result of the selected sample and selecting a criterion that provides a parameter for the predetermined disease;
A method for determining criteria for pathological diagnosis of a given disease.
(Item 2)
The method according to item 1, wherein the plurality of pathologists are a group of experts.
(Item 3)
The method according to item 1, wherein the statistical analysis is discriminant analysis stepwise.
(Item 4)
The following steps:
(A) selecting criteria candidates for a given disease;
(B) obtaining the judgment results of a plurality of pathologists using the criteria candidates for the same pathological diagnosis sample;
(C) a step of grouping the determination results obtained in step (b) by hierarchical cluster analysis;
(D) excluding judgment results that were not grouped by hierarchical cluster analysis;
(E) determining the validity of the determination results grouped by hierarchical cluster analysis; and
(F) selecting criteria candidates that are common in 60% or more of the judgment results grouped by hierarchical cluster analysis as criteria for pathological diagnosis of a predetermined disease;
A method for determining criteria for pathological diagnosis of a given disease.
(Item 5)
Item 5. The method according to item 1 or 4, wherein the predetermined disease is idiopathic interstitial pneumonia.
(Item 6)
The criteria candidate is
1) Clear scar-like fibrosis 2) Inflammatory cell infiltration into the alveolar septum (IP) (region farther from the honeycomb lung)
3) Microscopic honeycomb lung formation 4) Distribution of lesions 5) Bias to lobular margin 6) Central bronchiole lesion 7) Fibroblast nest 8) Adjacent normal lung 9) Smooth muscle hyperplasia 10) Peripheral airway lesions 11) Lymphoid follicles with germinal centers 12) Significant changes due to smoking 13) Changes in acute lung injury 14) Granulomas / stromal giant cells 15) Organized pneumonia (Masson body)
16) Pleural abnormality 17) Vascular changes (pulmonary artery)
18) Destruction of the lung structure 19) Organized lesion in the lumen (Masson body)
20) The method according to item 5, which is selected from the group consisting of elastic fibrosis.
(Item 7)
This is a method to support the diagnosis of idiopathic interstitial pneumonia, including the following criteria: (1) Clear scar-like fibrosis (2) Microscopic honeycomb formation (3) Distribution of lesions (4) Leaflet margin (5) Adjacent normal lung (6) Inflammatory cell infiltration into alveolar septum (in areas away from honeycomb)
Providing a diagnostic indication based on the method.
(Item 8)
A system that supports pathological diagnosis of a given disease, including:
(K) a plurality of criteria relating to the predetermined disease, and a storage unit that stores parameters of each of the plurality of criteria;
(L) a finding data acquisition unit that acquires finding data of pathological diagnosis related to the plurality of criteria;
(M) based on the parameter and the finding data; (i) if the parameter and the finding data match, estimate the subject's likelihood of suffering the predetermined disease, or (ii) A determination unit that prompts the user to seek an expert's determination when the parameter does not match the finding data; and
(N) A system including an output unit that outputs the determination of the determination unit.
(Item 9)
9. The system of item 8, wherein the criteria and the parameters are as follows:
(A) selecting criteria candidates for a given disease;
(B) A step of obtaining a judgment result of a plurality of pathologists for the criteria candidate for the same pathological diagnosis sample, wherein the plurality of pathologists are a group of experts or experts and A process consisting of a group of specialized pathologists;
(C) selecting a sample in which more than 60% of the plurality of pathologists showed the same diagnosis; and
(D) performing a statistical analysis on the judgment result of the selected sample and selecting a criterion that provides a parameter for the predetermined disease;
Determined by the system.
(Item 10)
The disease is idiopathic interstitial pneumonia, the criteria and parameters are:
(1) Clear scar-like fibrosis (2) Microscopic honeycomb lung formation (3) Distribution of lesions (4) Bias to lobular margin (5) Adjacent normal lung (6) Alveolar septum Inflammatory cell infiltration (in areas away from honeycomb)
The system according to item 8, wherein
(Item 11)
A cloud computing system that supports pathological diagnosis of a given disease, including:
(P) a plurality of criteria relating to the predetermined disease, and a storage unit that stores parameters of each of the plurality of criteria;
(Q) a reception unit that receives findings data of pathological diagnosis regarding the plurality of criteria from a user;
(R) based on the parameter and the finding data; (i) if the parameter and the finding data match, estimate the subject's likelihood of suffering the predetermined disease, or (ii) A determination unit that prompts the user to seek an expert's determination when the parameter does not match the finding data; and
(S) a providing unit that provides the determination of the determining unit to the user;
A system comprising:
(Item 12)
12. The cloud computing system according to item 11, wherein the criteria and the parameter are the following steps:
(A) selecting criteria candidates for a given disease;
(B) A step of obtaining a judgment result of a plurality of pathologists for the criteria candidate for the same pathological diagnosis sample, wherein the plurality of pathologists are a group of experts or experts and A process consisting of a group of specialized pathologists;
(C) selecting a sample in which more than 60% of the plurality of pathologists showed the same diagnosis; and
(D) performing a statistical analysis on the judgment result of the selected sample and selecting a criterion that provides a parameter for the predetermined disease;
Determined by the system.
(Item 13)
The disease is idiopathic interstitial pneumonia, the criteria and parameters are:
(1) Clear scar-like fibrosis (2) Microscopic honeycomb lung formation (3) Distribution of lesions (4) Bias to lobule margin (5) Adjacent normal lung (6) Alveolar septum Inflammatory cell infiltration (in areas away from honeycomb)
The system according to item 11, wherein

  The present invention provides systems and methods that support pathological diagnosis to obtain more consistent results as compared to conventional methods. The present invention also provides a method for determining pathological diagnostic criteria to obtain more consistent results. The present invention also provides systems and methods that support pathological diagnosis based on the criteria obtained by the present invention.

FIG. 1 shows a phylogenetic tree for cluster analysis without Supervise. Four clusters (A to D) are seen. Cluster A is the largest cluster containing 9 panels. Panels from each group are basically well mixed, but cluster D is composed only of JP panels. Most of the general pathologists are located outside the cluster. “EX” means 4 experts including ATS / ERS classification / review members, “UE” means 8 US and European lung pathologists, “JP” means 10 Japanese lung pathologists And “GP” refers to seven general pathologists. FIG. 2 shows the Kaplan Meier curve for survival without progression. If the percentage predicted FVC drop is equal to or greater than 10, the case is considered to have an event. Death within the follow-up period is also considered an event. A solid line corresponds to a case diagnosed as a normal interstitial pneumonia (UIP) pattern, and a broken line corresponds to a case diagnosed as a nonspecific interstitial pneumonia (NSIP) pattern. A consensus diagnosis was determined as the primary diagnosis between panels in each cluster. If not definitive, the case was excluded from the analysis. None of the groups showed statistical significance due to the small number of cases (p-values of 0.231 and 0.142, respectively). Note that the NSIP curve in cluster D is lower than that in cluster A.

  The description of the preferred embodiment is described below, but it should be understood that this embodiment is an illustration of the present invention and that the scope of the present invention is not limited to such a preferred embodiment. It should be understood that those skilled in the art can easily make modifications, changes and the like within the scope of the present invention with reference to the following preferred embodiments.

  The present invention will be described below. It is to be understood that the terms used in this specification are used in the meaning normally used in the art unless otherwise specified.

  Listed below are definitions of terms particularly used in the present specification.

(Definition)
As used herein, the term “criteria” refers to events utilized in pathological diagnosis. Criteria typically includes, but is not limited to, events observed on hematoxylin-eosin stained (H & E stained) slides and elastica-Wangeson stained (EVG stained) slides. Not.

Representative examples of criteria candidates and their scores in H & E staining of idiopathic interstitial pneumonia (IIP) include, for example,
1) Clear scar-like fibrosis Mild
Moderate
Severe
None
Skip without judging (Skip)
2) Inflammatory cell infiltration into the alveolar septum (IP) (region far from the honeycomb lung)
Mild
Moderate
Severe
None
Skip without judging (Skip)
3) Microscopic honeycomb lung formation Mild
Moderate
Severe
None
Skip without judging (Skip)
4) Distribution of lesions Clearly clear (Clearly patchy)
Relatively patchy
Diffuse
Skip without judging (Skip)
5) Deviation to the edge of leaflet Mild
Moderate
Severe
None
Skip without judging (Skip)
6) Bronchiolocentric lesion Mild
Moderate
Severe
None
Skip without judging (Skip)
7) Fibroblast nest Mild
Moderate
Severe
None
Skip without judging (Skip)
8) Transition between adjacent normal lung and steep lesion (With abduction transition)
Slow transition with lesions (With clinical transition)
None
Skip without judging (Skip)
9) Smooth muscle hyperplasia Mild
Moderate
Severe
None
Skip without judging (Skip)
10) Peripheral airway lesions Mild
Moderate
Severe
None
Skip without judging (Skip)
11) Lymphoid follicle with germinal center Mild
Moderate
Severe
None
Skip without judging (Skip)
12) Significant changes due to smoking RB / DIP-like reaction Emphysema
Stellate scar
Only a few discoveries (Slight findings only)
None
Skip without judging (Skip)
13) Change in acute lung injury Mild
Moderate
Severe
None
Skip without judging (Skip)
14) Granulomas / giant cells in the interstitium Mild
Moderate
Severe
None
Skip without judging (Skip)
15) Organized pneumonia (Masson body)
Mild
Moderate
Severe
None
Skip without judging (Skip)
16) Pleural abnormality Thoracic thickness (fibrotic)
Pleurisy (cell infiltration)
Pleurisy (fibrin)
No pleurodesis (None)
Skip without judging (Skip)
17) Vascular changes (pulmonary artery)
Serious intimal thickness Serious medial thickness Vasculitis Thrombus Slight discovery only
None
Skip without judging (Skip)
However, it is not limited to these.

As a representative example of criteria candidates in EVG staining of idiopathic interstitial pneumonia (IIP), for example,
1) Destruction of lung structure 2) Organized lesion in the lumen (Masson body)
3) Examples include but are not limited to elastic fibrosis.

  Other criteria for interstitial pneumonia include hyaline membrane formation, plasma cell infiltration, edema, type II epithelial hyperplasia, eosinophil infiltration, neutrophil infiltration, lymphocyte infiltration, pigmentation, hemosiderin deposition, histiosphere Type, squamous metaplasia, presence or absence of apoptosis, presence or absence of necrosis, and the like.

  In addition, as criteria for distinguishing cancer histology such as lung cancer, intercellular bridge, degree of keratinization, nucleolus, cytoplasmic acidophilicity, desmoplastic reaction, fission image, mucus production, intracellular nuclear localization, Stratification, nuclear shape, chromatin pattern, glandular formation, papillary structure, cord arrangement, alveolar formation and size, scaly arrangement, TTF1 staining, NapsinA staining, p63 staining, CK5 / 6 Staining, CK14 staining, rosette formation, neuroendocrine marker staining, and the like. The same applies to other tumors (such as brain tumors), which are characteristic of brain tumors such as the formation of perinuclear halos, presence or absence of protrusions, proliferation of vascular endothelium, cell density, staining of GFAP, staining of Olig2, MIB -1 positive rate, S-100 dyeability, presence of Rosenthal fiber, etc. are added. Similarly, criteria for cancerous and non-neoplastic lesions throughout the body are included.

  As used herein, the term “sample for pathological diagnosis” refers to a sample that can be used for diagnosis in pathology. This sample may be a pathological specimen itself such as a slide of H & E staining and a slide of EVG staining, or may be image data of these pathological specimens.

  As used herein, the term “pathologist” includes (1) “experts” who classify, create or review criteria for a given disease, and (2) “specialized” who specialize in a given disease. Pathologists ”and (3)“ general (non-specialist) pathologists ”(also referred to as“ general pathologists ”) who perform comprehensive pathological diagnosis without specializing in a predetermined disease. It is not limited to these.

  As used herein, the term “pathologist's judgment result” refers to a judgment result derived by a pathologist based on findings on criteria or criteria candidates. The “pathologist's judgment result” includes, but is not limited to, whether or not the patient is afflicted with a specific disease and the possibility of being afflicted with a specific disease.

  As used herein, the term “parameter” refers to information that associates criteria with a given disease.

For example, criteria parameters available in idiopathic interstitial pneumonia (IIP) are as follows:
・ When "clear scar fibrosis" is moderate to severe, it suggests normal interstitial pneumonia (UIP), and when none or mild, nonspecific interstitial pneumonia (NSIP) Or other interstitial pneumonia.
・ When “microscopic honeycomb lung formation” is moderate to severe, it suggests normal interstitial pneumonia (UIP), otherwise, nonspecific interstitial pneumonia (NSIP) or other Suggest interstitial pneumonia.
・ If the distribution of lesions is relatively patchy or clearly patchy, it usually indicates interstitial pneumonia (UIP), and if it is diffusely distributed, nonspecific interstitial pneumonia (NSIP) Other interstitial pneumonia is suggested.
・ When "bias to lobular margin" is moderate to severe, it suggests normal interstitial pneumonia (UIP), and when none or mild, nonspecific interstitial pneumonia (NSIP) Or other interstitial pneumonia.
・ When “adjacent normal lung” suddenly moves to the lesion, it suggests normal interstitial pneumonia (UIP), and when it gradually moves to the lesion, nonspecific interstitial pneumonia (NSIP) ) And other interstitial pneumonia.
・ When inflammatory cell infiltration into alveolar septum is moderate to severe, it suggests nonspecific interstitial pneumonia (NSIP) or other interstitial pneumonia. Suggests interstitial pneumonia (UIP).

  As used herein, a criterion having a parameter is information about a given disease (eg, “none”, “mild”, “moderate”, or “serious”) (eg, “none”, “mild”, or “severe”). Criteria that provide UIP, or not UIP, NSIP, or NSIP).

(1. Criteria determination method)
Criteria candidates that can be used in the pathological diagnosis of a given disease are well known. It is important to select criteria that are useful to support pathological diagnosis of a given disease from among the known criteria candidates. In the present invention, such criteria are selected as follows using the results of pathological diagnosis of a plurality of pathologists.

(1.1. Method of determining criteria from statistical processing of findings of multiple pathologists)
Criteria can be determined by statistically processing the knowledge of multiple pathologists (preferably, experts or specialized pathologists, more preferably experts). The plurality of pathologists may be a plurality of pathologists that belong to the same group by cluster analysis described in “1.2” below.

  Regarding determination results using criteria candidates by a plurality of pathologists, 50% or more, 60% or more, 70% or more, 75% or more, 80% or more, 85% or more, 90% or more, or 95 of a plurality of pathologists Select samples (samples) that showed the same diagnosis by more than%, calculate the average value of the items, perform statistical processing, and provide “parameters” (ie, information that associates criteria with a given disease) from these Select the criteria to be used.

  Statistical processing includes, but is not limited to, discriminant analysis stepwise, Fisher exact test, chi-square test, log rank test, Wilcoxon test, Cox test, logistic regression analysis, and combinations thereof. Further, as the statistical processing, for example, statistical analysis software such as JMP (registered trademark) can be used.

(1.2. Method using cluster analysis)
Using cluster analysis, criteria can be determined as follows:
(A) selecting criteria candidates for a given disease;
(B) obtaining the judgment results of a plurality of pathologists using the criteria candidates for the same pathological diagnosis sample;
(C) a step of grouping the determination results obtained in step (b) by hierarchical cluster analysis;
(D) excluding judgment results that were not grouped by hierarchical cluster analysis;
(E) determining the validity of the determination results grouped by hierarchical cluster analysis; and
(F) A step of selecting criteria candidates common to the determination results grouped by hierarchical cluster analysis as criteria for pathological diagnosis of a predetermined disease.

  The hierarchical cluster analysis in the step (c) can be performed using various well-known techniques regarding the relationship between the determination results of each pathologist and the course of the subject. For example, when software is divided into clusters by diagnosis using SPSS or other software, a consensus diagnosis for each cluster (diagnosis selected by the most pathologists in the cluster) and correlation between changes in the lung function of the patient can be seen. it can. For example, it is divided into an improvement group, an exacerbation group, and a stable group by the change by 12 months of lung functions (FVC, DLco, etc.). Whether the diagnosis is correlated with the distinction between the improvement group and the exacerbation group can be examined by chi-square test or Fisher test.

  In the step (d), the judgment result that is not grouped by the cluster analysis in the step (c) deviates from the judgment result of other pathologists, and therefore it is considered that the diagnosis accuracy is impaired. exclude. However, in a specific aspect of the present invention, the exclusion step (d) is not essential, and the present invention can be implemented without performing the exclusion step.

  In the exclusion step (d), for example, the number of persons included in the group and the low clustered hierarchy can be used as an index to exclude the diagnosis results that were not included there.

  The judgment of validity in the step (e) can be determined by comparing the judgment result with the course of the subject and whether the course of the subject matches the judgment result. This validity determination can be made by creating a determination result as to whether or not a given disease is lethal and comparing it with the subsequent survival rate, survival time, etc. of the subject. This validity can be determined by Fisher's exact test, chi-square test, log rank test, Wilcoxon test, Cox test, logistic regression analysis, and the like.

  The determination of the validity in the step (e) is not necessarily essential, and the present invention can be implemented without performing the step (e).

  The step (f) may be performed using all of the determination results grouped by the cluster analysis or may be performed using a part thereof. For example, in the step (e), validity may be determined, a group with high validity may be selected, and the step (f) may be performed.

  In the above step (f), “common in the judgment results grouped by hierarchical cluster analysis” means that, for example, the findings regarding the criteria are 60% or more, 65% or more, 70% or more of the group , 75% or more, 80% or more, 85% or more, 90% or more, or 95% or more.

(2. Use of selected criteria)
A user who makes a diagnosis obtains a score for the selected criteria. The score is converted into a numerical value from 0 (none) to 1 (mild), 2 (moderate), 3 (serious), for example. The sum total of the scores is calculated for each case by multiplying the numerical scores by the weighting coefficients of each criterion (for example, coefficients obtained by statistical analysis described in 1.1. Above). Diagnosis by the user can be supported by this score.

  For example, in the diagnosis of UIP or NSIP, a case with a high score is a case with a high probability of making a diagnosis as UIP, and a case with a low score is a case with a high probability of making a diagnosis with NSIP. A case showing an intermediate score can be output to the user as a case that is difficult to diagnose and a case where consultation with an expert is recommended.

  In addition, the user's score and the findings on which the scoring is based are acquired simultaneously as digital images, and these images and score values are provided to the expert so that the expert can set the criteria for scoring the user. It is also possible to judge whether it is right or wrong. For example, if the score value and finding image acquired by the user are transferred to the expert as a consultation recommended case, quicker and simpler consultation can be performed. This expert judgment or feedback to the consultation user can also be used as a diagnostic aid.

(3. System of the present invention)
Although the system of the present invention can be executed using a single computer, a computer connected to a network or a computer connected to a cloud computing system is preferable.

(3.1. Single computer system)
The present invention provides a computer system that supports pathological diagnosis related to a predetermined disease and that is not connected to a network. The system is as follows:
(K) a plurality of criteria relating to the predetermined disease, and a storage unit that stores parameters of each of the plurality of criteria;
(L) a finding data acquisition unit that acquires finding data of pathological diagnosis related to the plurality of criteria;
(M) based on the parameter and the finding data; (i) if the parameter and the finding data match, estimate the subject's likelihood of suffering the predetermined disease, or (ii) A determination unit that prompts the user to seek an expert's determination when the parameter does not match the finding data; and
(N) An output unit that outputs the determination of the determination unit is provided. The determination unit that is the main body of this system may be configured by a computer, for example, a general-purpose computer.

  The storage unit may store finding data as necessary. Further, the output unit may output (display on the screen and / or print) the image information of the observation data as necessary.

A typical processing procedure of this system is as follows.
1) A plurality of criteria related to a predetermined disease and parameters of the plurality of criteria are stored in the storage unit, or a plurality of criteria related to the predetermined disease and parameters of the plurality of criteria are stored. Providing a storage unit;
2) Acquire findings data of pathological diagnosis on multiple criteria via the findings data acquisition unit;
3) Based on the parameters and the findings data, the judgment unit (or the computer main body) (i) estimates and outputs the possibility of the subject's affliction when the parameters match the findings data Or (ii) prompt the user to seek expert judgment if the parameters do not match the findings data; and
4) The determination unit outputs the determination to the output unit.

The criteria are typically criteria determined according to the method of the present invention. If the disease is idiopathic interstitial pneumonia, typical criteria,
(1) Clear scar-like fibrosis (2) Microscopic honeycomb lung formation (3) Distribution of lesions (4) Bias to lobular margin (5) Adjacent normal lung (6) Alveolar septum Inflammatory cell infiltration (in areas away from honeycomb)
It is.

(3.2. Cloud computing system)
The present invention provides a cloud computing system that supports pathological diagnosis related to a predetermined disease. The system is as follows:
(Q) a reception unit that receives findings data of pathological diagnosis regarding the plurality of criteria from a user;
(R) based on the parameter and the finding data; (i) if the parameter and the finding data match, estimate the subject's likelihood of suffering the predetermined disease, or (ii) A determination unit that prompts the user to seek an expert's determination when the parameter does not match the finding data; and
(S) a providing unit that provides the determination of the determining unit to the user;
Is provided. The determination unit that is the main body of this system may be configured by a computer, for example, a general-purpose computer.

  The storage unit may store finding data as necessary.

A typical processing procedure of this system is as follows.
1) A plurality of criteria related to a predetermined disease and parameters of the plurality of criteria are stored in the storage unit, or a plurality of criteria related to the predetermined disease and parameters of the plurality of criteria are stored. Providing a storage unit;
2) Accepting pathological diagnosis findings data from a user via a reception unit;
3) Based on the parameters and the findings data, the judgment unit (or the computer main body) (i) estimates and outputs the possibility of the subject's affliction when the parameters match the findings data Or (ii) prompt the user to seek expert judgment if the parameters do not match the findings data; and
4) Provide the judgment of the judgment unit to the user.

The criteria are typically criteria determined according to the method of the present invention. If the disease is idiopathic interstitial pneumonia, typical criteria are
(1) Clear scar-like fibrosis (2) Microscopic honeycomb lung formation (3) Distribution of lesions (4) Bias to lobular margin (5) Adjacent normal lung (6) Alveolar septum Inflammatory cell infiltration (in areas away from honeycomb)
It is.

  The system of the present invention preferably prompts the user to seek expert judgment if the decision does not match typical parameters.

(Example 1: Analysis of pathological diagnosis of idiopathic interstitial pneumonia)
Hereinafter, the present invention will be described in detail by way of examples of pathological diagnosis of idiopathic interstitial pneumonia, but the present invention is not limited to idiopathic interstitial pneumonia.

(design)
We have 20 consecutive video-assisted thoracic cavities clinically judged to be chronic interstitial lung biopsied between November 2008 and May 2009 from a public Tosei hospital. Mirror surgery (VATS) was selected. 29 panelists (pathologists) observed slides of hematoxylin and eosin staining (H & E staining) and elastica-Wangieson staining (EVG staining) to diagnose histological patterns without information on clinical and radiological data went. For each region and category, a generalized kappa coefficient was calculated for agreement between observers. The panelists were grouped by cluster analysis. Predictive data from lung function tests were obtained, and disease progression was defined by changes in forced vital capacity (FVC) and carbon monoxide diffusing capacity at 1 year follow-up. The association between disease progression and pathological diagnosis was analyzed using Fisher's exact test and data were compared between clusters.

(result)
The overall consistency of the 20 cases among all pathologists was 0.22. If limited to pulmonary pathologists, κ was 0.27. Panelists were further divided into four groups by cluster analysis. Patients' prognosis is divided into improvement group, exacerbation group, and stable group based on changes in FVC and DLco based on changes in respiratory function tests at 12 months, where only the largest cluster is statistically determined as disease progression A significant association was shown (p = 0.045).

(Conclusion)
The consistency of pathological diagnosis in IIP is low. There were several different subgroups among pathologists, and only the largest cluster showed an association with disease progression. These findings can be found on the website http: // ws022. u-toyama. ac. Through jp (ID: ILD, pass: UIP), it can be used as educational material for the standardization of pathological diagnosis in clinical practice, and practice cases and model solutions can be shown in the same way.

(Introduction)
Contrary to the prediction after the publication of the ATS / ERS classification of idiopathic interstitial pneumonia (IIP) in 2002, interobserver variability in pathological diagnosis of IIP is still a problem (Nicholson et al, Franks). , Et al, Flahrty, Thorax 2003). Surgical lung biopsy is an invasive procedure for patients, and surgical biopsy for chronic interstitial lung disease is an inevitable process for the diagnosis of interstitial lung disease. This procedure is extremely invasive, and obtaining a VATS biopsy is critical for the patient, and there is also some risk of sudden changes. In view of the risk of sudden changes (also referred to as acute exacerbations) after VATS biopsy, this treatment should be sufficiently beneficial to the patient. In this regard, pathological diagnosis should have a prognostic and / or predictive value that is clinically available. One problem that is not clearly known is how consistent pulmonary and non-pulmonary pathologists have obtained years after the ATS / ERS consensus statement.

  In addition, an accurate diagnosis for chronic interstitial pneumonia, given the recent advances in new specific therapeutic options such as pirfenidone or NAC, which are used exclusively in patients with idiopathic pulmonary fibrosis (IPF) Is gaining importance.

  Now, in order to know how well we can be internationally consistent in pathological diagnosis, and to know what is the best suggestive diagnosis for prognosis prediction, the authors We investigated the consistency between observers for chronic interstitial pneumonia and compared its significance to patient outcomes.

(Materials and methods)
(Case selection and clinical evaluation)
The authors selected 20 consecutive video-assisted thoracoscopic surgery (VATS) from a single hospital (Public Tosei Hospital) where interstitial fibrosis was diagnosed as a major histological finding in a pathology report did. Patient's age, gender, background systemic disease, drug history, high-resolution CT (HRCT) findings, occupational exposure, treatment type, smoking calendar, survival status, and pulmonary function test (PFT) clinical data Obtained at several consecutive time points including before and 12 months after biopsy. The PFT data included in this study were vital capacity (VC), forced vital capacity (FVC), forced expiratory volume for 1 second (FEV1), and carbon monoxide diffusion capacity (Dlco).

(Scoring of cases by panelists)
Twenty-nine panelists who did not see any clinical or radioactivity data were asked to review all cases. The panelists were: 4 experts (EX) including ATS / ERS classification / review members, 8 US / EU pulmonary pathologists (USEP), 10 Japanese pulmonary pathologists (JP), and 7 Classified as a human general pathologist (GP).

  The panelists reviewed slides stained with hematoxylin and eosin staining and EVG staining and classified them by ATS / ERS pattern. All scoring data was obtained within 11 months from the VATS biopsy.

(Statistical analysis)
Most pulmonary pathologists categorized cases as “normal interstitial pneumonia (UIP)”, “nonspecific interstitial pneumonia (NSIP)” and “others”, so other ATS / ERS patterns Cases diagnosed below were put together in the “Other” subgroup.

  A generalized weighted κ coefficient was calculated for the agreement between observers for each category. The kappa consistency between the two observers in all combinations was also calculated. Consistency of the observer is not acceptable for the κ values of less than 0.20, 0.20 to 0.39, 0.40 to 0.59, 0.60 to 0.79, and 0.80 or more. ), Fair, moderate, good, and excellent.

  Supervised cluster analysis was performed on pathological diagnostic data. An association between pathological diagnosis and disease progression was made using a chi-square test. Due to the importance of histological diagnosis between the UIP and NSIP patterns and the uncertain clinical meaning of “other”, only cases of diagnosis of UIP or NSIP patterns were further analyzed. A Mantel-Haenszel test was used to compare group results between clusters. The numbers of improved and deteriorated patients were entered into a table divided into UIP patterns and NSIP patterns, and the average and standard deviation of the numbers in each column were calculated. Fisher's exact test was used to analyze the significance of the association between UIP vs NSIP discrimination and disease progression. The association between pathological diagnosis and progression-free survival (survival analysis by progression-free period) was also analyzed using the log rank test and the Kaplan Meier curve was plotted. A case was considered to have an exacerbation when the percent predicted FVC dropped by 10 or more. A p value <0.05 was considered significant. Statistical analysis was performed using STATVIEW (Computing Resource Center, Santa Monica, CA, USA) for the kappa coefficient, SPSS for cluster analysis, and JMP® (SAS Institute, Cary, for other analyses). NC, USA).

(Prognostic significance)
Lung function tests were used to determine disease progression and survival status. PFT data was obtained every 2-5 months until death in all patients. One patient refused PFT for complete recovery and was transferred to another clinic.

  Patients were further divided into three groups (improvement, stability, and deterioration) based on changes in PFT data at two different time points (before and after 12 months (± 2 months)). According to the study of Latsi et al. And Collard et al., Improved cases are defined as cases with an increase in the percent predicted FVC of 10 or greater, and worsened cases are a decrease in the percent predicted FVC of 10 or greater or time points before VATS Any case of worse than 15% of Dlco was defined. A stable case was defined as a case in which Dlco had not dropped by 15% from the pre-VATS time point and the percent predicted FVC change was less than 10.

  Informed consent was obtained for all patients and the institutional ethics committee of the public Tohsei Hospital approved this protocol.

(result)
(Group data and outcome determinants)
A clinical summary of all cases included in the study is shown in Table 1.

Abbreviations:
UCTD, unclassified connective tissue disease; RA, rheumatoid arthritis;
M, male; F, female; Ex, former smoker;
S, steroids; C, cyclosporin A; P, pirfenidone;
FVC, forced vital capacity; Dlco, carbon monoxide diffusion capacity; f / u, follow-up months;
*: Functional test data obtained immediately before VATS biopsy
**: A percentage predicted FVC reduction equal to or greater than 10 was observed and the case was considered to have an event. Deaths within the follow-up process are also counted as events.
***: Case 11 was concluded as rapidly progressive interstitial pneumonia by multi-region CRP discussion. No evidence of chronicity was found clinically or radiologically. This patient died in one month. For the purpose of this study to identify a meaningful prognosis in chronic fibrosis interstitial pneumonia, this case was excluded from the prognosis and survival analysis.

  The patient's age ranged from 31 to 75 years with a median of 62. As for the sex of patients, 13 were men, 7 were women, and there were many men. The smoking situation was that 4 were still smokers, 9 were former smokers, and 7 had no smoking experience. In 19 cases, PFT follow-up was completed until death, or to the present if surviving, but one had full recovery and the patient was transferred to a personal clinic without disease progression. Thus, complete PFT data is available for 19 cases. The follow-up period was 1-25 months after VATS, with a median of 19 months. Three patients died during follow-up. One patient (Patient 2) experienced an acute exacerbation immediately after VATS biopsy but improved with steroid pulse therapy. Two cases (patients 2 and 6) experienced acute exacerbations during the follow-up course. Patient 2's symptoms and PFT data improved again after treatment with steroid pulses, but patient 6 died shortly after exacerbation.

  Eight cases (patients 3, 5, 6, 8, 10, 12, 19 and 20) used pirfenidone in the course, among which cases 5, 8 and 20 were treated with pirfenidone alone.

  Seven cases (patients 2, 3, 7, 13, 14, 16 and 17) were treated with steroids and cyclosporin A. Five cases (patients 1, 4, 9, 15 and 18) were only being observed and received no treatment.

  Two cases had connective tissue disease (case 7 as systemic sclerosis (SSc), case 17 as rheumatoid arthritis (RA)) as a background disease. Four cases (patients 6, 11, 13, and 20) were detected to have a condition known as unclassified connective tissue disease (UCTD), but none of these four patients were It did not progress to defined connective tissue disease.

  Case 11 had a histological description of fibrosis in the pathology report and was included in this study. However, this patient died one month after the onset of the disease, and this case was actually concluded to be rapidly progressive interstitial pneumonia. This was concluded as acute interstitial pneumonia as an ATS / ERS consensus diagnosis after CRP discussion, excluded from chronic interstitial pneumonia, and not included in prognostic and survival analyses.

(Consistency of pathological diagnosis)
VATS biopsy was taken from 3 segments (S5, 8 and 9) for 17 cases, 2 segments (S5 and 8) for 2 cases, and 1 segment (S8) for 1 case. For all cases, one EVG stained slide and multiple H & E stained slides were prepared per case.

  All diagnostic data is shown in Table 2.

EX: ATS / ERS Committee / Review Member,
USEP: US and EU pulmonary pathologists,
JP: Japanese lung pathologist,
GP: General pathologist The numbers correspond separately as follows:
1, UIP pattern;
2, NSIP pattern;
3, DIP pattern;
4, RB pattern;
5, OP pattern;
6, DAD pattern;
7, LIP pattern;
Note that most of the columns in 0, other pulmonary pathologists are filled with 1, 2 or 0, whereas the GP column shows a mixture of different histological patterns. This is because most pulmonary pathologists believe that either chronic or fibrotic interstitial pneumonia is either a UIP or NSIP pattern, and if this histological pattern is in either the UIP or NSIP pattern If not, it indicates that other histological patterns other than those described in the ATS / ERS classification are being considered.

  Most pathologists who specialize in pulmonary pathology diagnosed under UIP, NSIP and “Other”. “Other” includes chronic interstitial pneumonia (NOS), hypersensitivity pneumonia pattern, collagen vascular disease-related interstitial lung disease pattern, bronchiolocentric interstitial fibrosis pattern, smoking-related interstitial lung disease pattern, And various states such as “Other” as specified in the ATS / ERS consensus statement. None of the specific patterns in “Other” reached a match across the three panelists. Several pulmonary pathologists used COP, DIP and RB diagnoses, but their use is rare (12 calls out of a total of 4400 scores, all consistent across 3 panelists) Compared to pulmonary pathologists, general pathologists (6 from Japan and 1 from the US) used COP, RB, DIP, AIP, and LIP diagnoses more frequently (30 calls out of a total of 210 scores).

  The weighted kappa coefficient for 20 cases for all pathologists was 0.22 and was considered “fair”. If limited to pulmonary pathologists, the weighting κ was 0.27 and still “fair”.

  For diagnostic consistency analyzed by subcategory, the weight κ is 0.41 (normal) for EX, 0.30 (slight) for USEP, 0.18 (bad) for JP, and The GP was 0.13 (bad).

  K values between two observers among 22 pulmonary pathologists were also calculated (Table 3).

The gray column corresponds to a kappa value higher than 0.6. Most of the highlighted columns are combinations for ATS / ERS committee / review members (Panels 1-3).

  Some high agreement was observed mainly among members of the ATS / ERS committee / review. Of the 231 pattern combinations, only 15 matched good matches, and 12 out of 15 were combinations that actually included ATS / ERS committee / review members. Except for ATS / ERS committee / review members, only three combinations resulted in good agreement.

(Cluster analysis)
Cluster analysis without hierarchical Supervise was performed using software and the phylogenetic tree was plotted. As shown in FIG. 1, four clusters (clusters A to D) were observed, and nine scattered panels remained outside the clustering. Most of the panels outside the cluster were general pathologists, which was not surprising. Neither category nor region showed a strong association for each cluster, except that cluster D was formed only by Japanese lung pathologists. Cluster A is the largest cluster with 9 panels, including 4 JPs, 3 USEPs, and 2 EXs.

(Relationship with disease progression)
From the PFT data, 6 cases, 5 cases and 8 cases were considered as improved cases, stable cases and worsened cases, respectively. Due to prognostic uncertainty in stable cases, only improved cases and worsened cases (14 cases in total) were used for Fisher's test.
Individuals in the four clusters were analyzed separately, and the number of cases under diagnosis with UIP and NSIP patterns was separately applied to a 2 × 2 column. Average numbers and standard deviations were calculated for each cluster and column, and the numbers were fit to Fisher's exact test (Table 4a).

Only cluster A showed a statistically significant association with disease progression (p value = 0.045) (Table 4b).

Improved cases were defined by an increase in percentage predicted FVC equal to or greater than 10. Exacerbated cases were defined by either a decrease in percentage equal to or greater than 10 or a greater than 15 percent decrease in carbon monoxide diffusing capacity from the time immediately prior to VATS biopsy.
Cluster A shows a smaller range of standard deviations and a statistically significant association with disease progression (p value 0.045).

(Survival analysis by progression-free period)
Since the consensus diagnosis of both clusters B and C did not satisfy a sufficient number of cases for analysis of the Log Rank test, the Kaplan Meier curve could not be plotted. Both clusters A and D showed some separation of the two curves between the UIP and NSIP patterns, but no statistically significant difference was observed as the number of cases analyzed was small (The p values were 0.231 and 0.142, respectively) In cluster D, NSIP was diagnosed, but it was confirmed that the group included a poor prognosis group (FIG. 2).

(Example 2: Selection of criteria for idiopathic interstitial pneumonia)
Regarding the results of Example 1, JMP (registered trademark) ver. Discriminant analysis stepwise using 9 and scoring and diagnosing all sections individually by 4 experts for the 17 criteria examined, 3 or more of the 4 experts have the same diagnosis The average value of the item was calculated in the specimens that showed the above, and six criteria providing parameters were selected from these. They are "clear scar fibrosis", "microscopic honeycomb lung formation", "lesion distribution", "bias to lobular margin", "adjacent normal lung", "inflammatory cell infiltration into alveolar septum""Met.

  Among these, it was found that discrimination based on “microscopic honeycomb lung formation”, “lesion distribution”, and “bias to the edge of leaflet” is the most useful criterion. In the discriminant analysis, “granuloma” was also recognized as a molecule that regulates the diagnosis of UIP, but the score in 90% or more of the cases was “0” and the most characteristic observed disease was Since it was not idiopathic interstitial pneumonia but hypersensitivity pneumonitis, it was considered inappropriate for the evaluation of idiopathic interstitial pneumonia, and was excluded from the judgment results.

(Example 3: Use of selected criteria)
By using the criteria selected in the second embodiment, a method or system that supports the diagnosis of idiopathic interstitial pneumonia by the user can be provided.

  A user who makes a diagnosis obtains a score for the selected criteria. The score for each criterion is determined in advance. And a score is converted into the numerical value from 0 (none) to 1 (mild), 2 (moderate), 3 (serious), for example. The sum total of the scores is calculated for each case by multiplying these numerical scores by the weighting coefficient of each criterion. Based on this sum, the user's diagnosis can be supported.

  For example, in the diagnosis of UIP or NSIP, a case with a high score is a case with a high probability of making a diagnosis as UIP, and a case with a low score is a case with a high probability of making a diagnosis with NSIP.

  A case showing an intermediate score can be output to the user as a case that is difficult to diagnose and a case where consultation with an expert is recommended. In cases where diagnosis is difficult, the score obtained by the user and the findings on which the scoring is based are acquired simultaneously as digital images, and these images and score values are provided to the expert. It is also possible to determine whether the scoring criteria are correct or incorrect. For example, if the score value and finding image acquired by the user are transferred to the expert as a consultation recommended case, quicker and simpler consultation can be performed.

(Consideration of Examples 1 to 3)
The consistency of pathological diagnosis in IIP is still low. In particular, the diagnosis by JP and GP showed poor agreement. Judging from the low κ values in the non-ATS / ERS committee / review panel, the low degree of consistency is probably not related to unusual regional criteria, but probably to differences in individual interpretation. doing.

  Diagnostic criteria that are only available to limited experts cannot contribute to future science. On the other hand, the expert obtains the score for the diagnostic criteria that the user has attached and the findings on which the scoring is based on digital images at the same time, and provides these images and score values to the expert so that the expert It is also possible to judge whether the criteria for scoring is correct or incorrect. For example, if the score value and finding image acquired by the user are transferred to the expert as a consultation recommended case, quicker and simpler consultation can be performed.

  One of the important points of the present invention is that the main answer is always correct and the rare answer is not necessarily wrong, and proves whether there is a correlation between pathological diagnosis and clinical prognosis It is in. Another important point of the present invention is that it provides specific criteria that are important, for example, in the diagnosis of interstitial pneumonia.

  Given the degree of invasiveness of surgical lung biopsies, primarily VATS biopsies, pathological diagnosis is clinical by predicting prognosis or by predicting favorable effects and / or side effects of a particular treatment. Should be meaningful. Considering the significant therapeutic impact on patients, including lung transplantation, steroid therapy with or without cytoxane, and application to new drugs such as pirfenidone and bosentan, prognostic decisions, especially in chronic interstitial pneumonia, are is important.

  This study is meaningful in terms of identifying pathological diagnoses that predict patient outcomes within the variability of pathological diagnoses by supervised cluster analysis. It is important to convince that surgical lung biopsy is clinically meaningful. Within the cluster A panel, the cases are highly consistent, and the expected degree of improvement / progression from diagnosis can be used as educational material through the Internet. Such cases can be found at http: // ws022. u-toyama. ac. available from jp / address (type ILD as user ID and UIP as password).

  In the above analysis, there is a debate about whether to include cases that are already known to be connective tissue diseases. However, some reports show that UIP patterns in RA behave like IPF, and NSIP patterns in other connective tissue diseases, including SSc, behave like idiopathic NSIP. Therefore, there is no problem in being included in this case group.

  The present invention has an impact on clinical diagnosis and contributes to standardization of pathological diagnosis in, for example, chronic interstitial pneumonia. According to the present invention, for example, a user-friendly, simple and simple criterion for classifying a disease, which reveals a highly difficult disease such as interstitial pneumonia, is provided. Scores are obtained for these criteria and converted to numerical values from 0 (none) to 1 (mild), 2 (moderate), and 3 (severe). This number will be obtained by the user who will perform diagnosis in the future. The sum total of the scores is calculated for each case by multiplying these numerical scores by the weighting coefficient of each criterion. A case with a high value of this score is a case with high probability of diagnosing as UIP, and a case with a low score is a case with high probability of diagnosing as NSIP. A case showing an intermediate score can be treated as a case in which consultation with an expert is recommended as a case that is difficult to diagnose.

  The present invention provides systems and methods that support pathological diagnosis to obtain more consistent results as compared to conventional methods. The present invention also provides a method for determining pathological diagnostic criteria to obtain more consistent results. The present invention also provides for the determination of parameters that define the pathological diagnosis based on the criteria obtained by the present invention. The present invention provides a system and method for supporting pathological diagnosis. The present invention provides training cases and exemplary solutions for diagnosing using parameters in diseases that are difficult to diagnose and have low agreement.

Claims (13)

The following steps:
(A) selecting criteria candidates for a given disease;
(B) A step of obtaining a judgment result of a plurality of pathologists for the criteria candidate for the same pathological diagnosis sample, wherein the plurality of pathologists are a group of experts or experts and A process consisting of a group of specialized pathologists;
(C) selecting a sample in which more than 60% of the plurality of pathologists showed the same diagnosis; and
(D) performing a statistical analysis on the judgment result of the selected sample and selecting a criterion that provides a parameter for the predetermined disease;
A method for determining criteria for pathological diagnosis of a given disease.   The method of claim 1, wherein the plurality of pathologists is a group of experts.   The method of claim 1, wherein the statistical analysis is discriminant analysis stepwise. The following steps:
(A) selecting criteria candidates for a given disease;
(B) obtaining the judgment results of a plurality of pathologists using the criteria candidates for the same pathological diagnosis sample;
(C) a step of grouping the determination results obtained in step (b) by hierarchical cluster analysis;
(D) excluding judgment results that were not grouped by hierarchical cluster analysis;
(E) determining the validity of the determination results grouped by hierarchical cluster analysis; and
(F) selecting criteria candidates that are common in 60% or more of the judgment results grouped by hierarchical cluster analysis as criteria for pathological diagnosis of a predetermined disease;
A method for determining criteria for pathological diagnosis of a given disease.   5. The method according to claim 1 or 4, wherein the predetermined disease is idiopathic interstitial pneumonia. The criteria candidate is
1) Clear scar-like fibrosis 2) Inflammatory cell infiltration into the alveolar septum (IP) (region farther from the honeycomb lung)
3) Microscopic honeycomb lung formation 4) Distribution of lesions 5) Bias to lobular margin 6) Central bronchiole lesion 7) Fibroblast nest 8) Adjacent normal lung 9) Smooth muscle hyperplasia 10) Peripheral airway lesions 11) Lymphoid follicles with germinal centers 12) Significant changes due to smoking 13) Changes in acute lung injury 14) Granulomas / stromal giant cells 15) Organized pneumonia (Masson body)
16) Pleural abnormality 17) Vascular changes (pulmonary artery)
18) Destruction of the lung structure 19) Organized lesion in the lumen (Masson body)
20) The method of claim 5 selected from the group consisting of elastic fibrosis. This is a method to support the diagnosis of idiopathic interstitial pneumonia, including the following criteria: (1) Clear scar-like fibrosis (2) Microscopic honeycomb formation (3) Distribution of lesions (4) Leaflet margin (5) Adjacent normal lung (6) Inflammatory cell infiltration into alveolar septum (in areas away from honeycomb)
Providing a diagnostic indication based on the method. A system that supports pathological diagnosis of a given disease, including:
(K) a plurality of criteria relating to the predetermined disease, and a storage unit that stores parameters of each of the plurality of criteria;
(L) a finding data acquisition unit that acquires finding data of pathological diagnosis related to the plurality of criteria;
(M) based on the parameter and the finding data; (i) if the parameter and the finding data match, estimate the subject's likelihood of suffering the predetermined disease, or (ii) A determination unit that prompts the user to seek an expert's determination when the parameter does not match the finding data; and
(N) A system including an output unit that outputs the determination of the determination unit. 9. The system of claim 8, wherein the criteria and the parameters are the following steps:
(A) selecting criteria candidates for a given disease;
(B) A step of obtaining a judgment result of a plurality of pathologists for the criteria candidate for the same pathological diagnosis sample, wherein the plurality of pathologists are a group of experts or experts and A process consisting of a group of specialized pathologists;
(C) selecting a sample in which more than 60% of the plurality of pathologists showed the same diagnosis; and
(D) performing a statistical analysis on the judgment result of the selected sample and selecting a criterion that provides a parameter for the predetermined disease;
Determined by the system. The disease is idiopathic interstitial pneumonia, the criteria and parameters are:
(1) Clear scar-like fibrosis (2) Microscopic honeycomb lung formation (3) Distribution of lesions (4) Bias to lobular margin (5) Adjacent normal lung (6) Alveolar septum Inflammatory cell infiltration (in areas away from honeycomb)
9. The system of claim 8, wherein A cloud computing system that supports pathological diagnosis of a given disease, including:
(P) a plurality of criteria relating to the predetermined disease, and a storage unit that stores parameters of each of the plurality of criteria;
(Q) a reception unit that receives findings data of pathological diagnosis regarding the plurality of criteria from a user;
(R) based on the parameter and the finding data; (i) if the parameter and the finding data match, estimate the subject's likelihood of suffering the predetermined disease, or (ii) A determination unit that prompts the user to seek an expert's determination when the parameter does not match the finding data; and
(S) a providing unit that provides the determination of the determining unit to the user;
A system comprising: 12. The cloud computing system of claim 11, wherein the criteria and the parameters are the following steps:
(A) selecting criteria candidates for a given disease;
(B) A step of obtaining a judgment result of a plurality of pathologists for the criteria candidate for the same pathological diagnosis sample, wherein the plurality of pathologists are a group of experts or experts and A process consisting of a group of specialized pathologists;
(C) selecting a sample in which more than 60% of the plurality of pathologists showed the same diagnosis; and
(D) performing a statistical analysis on the judgment result of the selected sample and selecting a criterion that provides a parameter for the predetermined disease;
Determined by the system. The disease is idiopathic interstitial pneumonia, the criteria and parameters are:
(1) Clear scar-like fibrosis (2) Microscopic honeycomb lung formation (3) Distribution of lesions (4) Bias to lobular margin (5) Adjacent normal lung (6) Alveolar septum Inflammatory cell infiltration (in areas away from honeycomb)
The system of claim 11, wherein