RU2814404C1 - Method for differential diagnosis of obstructive pathology of major duodenal papilla - Google Patents

Abstract

FIELD: medicine; surgery; endoscopic ultrasonic diagnostics.

SUBSTANCE: disclosed is a method for differential diagnosis of obstructive pathology of a major duodenal papilla, where endosonographic images of the patient are compared with six reference endosonograms and reference schemes, containing an image of a normal major duodenal papilla and the main pathological conditions, additional informative signs of malignant obstruction are measured, with the use of a fuzzy decision rule system, the confidence of the patients being assigned to the investigated pathology classes is determined.

EFFECT: invention provides reducing the number of errors in the differential diagnosis of the obstructive pathology of the major duodenal papilla by comparing the analyzed endosonograms with the reference, as well as with their reference schemes and pathological formations with additional measurement of informative signs of malignant obstruction using an adequate mathematical apparatus of research.

1 cl, 7 dwg

Description

The invention relates to the field of medicine, namely to surgery and endoscopic ultrasound diagnostics and can be used in the differential diagnosis of obstructive pathology of the major duodenal papilla (MDP).

Benign and malignant pathology of the hepatopancreatoduodenal zone (HPZ) is widespread in the world population and covers all age groups of the population. One of the current localizations of this pathology is the BDK and the parapapillary zone of the duodenum (DU). Instrumental visualization of the BD ampulla and papillary zone is a complex diagnostic task that requires a multidisciplinary approach due to the complex anatomical features of this area, as well as the diversity of pathologies encountered. Full visualization of the structures of the papillary zone of the duodenum, and especially the BDK, is hampered by the presence of air in the duodenum and the small size of both the papilla itself and the pathology developing in it. Direct examination of the BDDC during gastroduodenoscopy makes it possible to diagnose pathology only if the lesion is sufficiently large and extends echstrampullary. The introduction into clinical practice of endoscopic ultrasonography (EUS), the unique property of which is detailed visualization of organs in the immediate vicinity of the digestive tract, has significantly increased the possibilities of objective diagnosis of BDK diseases. Endosonography is an additional highly effective method of examining patients with BDK pathology, and for intraampullary cancer, EUS is the most effective than all others. According to research, EUS has the greatest sensitivity and specificity in assessing the size and local extent of tumors and detecting lymph node involvement compared to CT and MRI. At the same time, to date there are no clearly formulated sonographic criteria and algorithms that allow differentiating inflammatory and neoplastic changes in the BDDC, which is naturally explained by the inability to use widespread statistical methods of data analysis to identify and evaluate diagnostic criteria, since it is not possible to observe representative homogeneous samples of patients with a statistically significant difference for each diagnostic criterion. This circumstance significantly reduces the information content and objectivity of endosonographic diagnosis of the pathology of the BDK and papillary zone. Errors and untimely diagnosis of diseases of the major duodenal papilla are still the cause of unsatisfactory results of treatment of the pathology being studied (Diagnostics and results of endoscopic methods of treatment of diseases of the major duodenal papilla / A. G. Shuleshova, E. I. Brekhov, V. V. Dobrusin [et al. ] // Medical Bulletin of the Ministry of Internal Affairs. - 2012. - No. 1(56). - P. 8-17).

There is a known method for diagnosing diseases of the BDK with gastroduodenoscopy and endoscopic retrograde cholangiopancreatography (ERCP) (Manual on clinical endoscopy: Edited by V.S. Savelyev, V.M. Buyanov, S.I. Lukomsky. - M.: Medicine. - 1985.; 544 p. .). To diagnose the pathology of the papilla, gastroduodenoscopy and transpapillary interventions are performed, during which visual and radiological changes in the BDK and ductal systems of the liver and pancreas are assessed. Stenosis is characterized by a change in the shape and size of the papilla, difficulty in cannulating the ampulla, dilatation of the bile and/or main pancreatic ducts, stricture of the bile and/or pancreatic ducts, determined visually and on radiographs, simultaneous contrasting of both ductal systems, an increase in the time of removal of the contrast agent from the general bile (more than 45 minutes) and main pancreatic (more than 15 minutes) ducts. The main disadvantage of such diagnostics is the inability to establish a diagnosis based on visual criteria and the need to perform ERCP, and in some cases, endoscopic papillosphincterotomy, which are invasive interventions and are associated with a high risk of complications. To diagnose adenoma and cancer of the BDK during gastroduodenoscopy, the shape, size of the BDK, the nature of the surface, and the condition of the mucous membrane are assessed. For BDK cancer, an antegrade X-ray contrast study allows one to assess the extent of malignant infiltration of the bile ducts. If there are changes characteristic of BDK neoplasms, endoscopic papillotomy and biopsy are indicated, since changes in the papilla are diagnosed by histological examination. The disadvantages of gastroduodenoscopy are the difficulty of diagnosing these changes in the early stages of the disease when the pathology is localized intraampullary. In such a situation, there is no possibility of differentiating malignant and benign changes in the BDK.

As the closest analogue, we chose the method for diagnosing the pathology of IVH according to EUS data using fuzzy mathematical models, described in the work of Belozerov V.A., Korenevsky N.A., Korzhuk N.L. Diagnosis of pathology of the extrahepatic bile ducts according to endoscopic ultrasonography using fuzzy mathematical models // Proceedings of the South-West State University. Series: Management, computer technology, computer science. Medical instrumentation. 2022. T. 12, No. 2. P. 149-164. https://doi.org/10.21869/ 2223-1536-2022-12-2-149-164.

In this method, differentiation of IVH obstruction is based on comparison of endosonograms of the observed pathology with reference patterns of reference endosonograms. Reference diagrams of endosonograms are used, where for each disease two variants of images are presented: with longitudinal visualization of the structure of the formation and with transverse one: C1 - normal CBD; C2-echographic structure of the CBD in choledocholithiasis; C3-echographic structure of the CBD containing a “loose” calculus; C4-echographic structure of the CBD in benign scar-inflammatory stricture; C5-echographic structure of the CBD in cholangiocarcinoma; C6- CBD with cholangiocarcinoma with invasion into surrounding structures. Using the scale of names C1, ..., C6, a scale of intervals S was constructed, displaying on it the measure of proximity between typical reference echo structures of the CBD, on which the functions of membership in the classes malignant obstruction - ω _P and benign obstruction - ω _D with the base variable S are determined. In this interpretation, the values membership functions are interpreted as confidence values in the studied classes of states. The values of the membership functions are calculated using the formulas:

$${\mu }_{{\omega }_D}\left(S\right)=\{\begin{array}{l}\text{0}\text{, if S}\le \text{1;}\\ \text{0}\text{.98S - 0}\text{.98}\text{, if 1<S}\le \text{2;}\\ \text{-0}\text{.245S+1}\text{.47}\text{, if 2<S}\le \mathrm{6,}\end{array}$$

$${\mu }_{{\omega }_P}\left(S\right)=\{\begin{array}{l}\text{0}\text{, if S}\le 2;\\ \text{0}\text{.242S - 0}\text{.484}\text{, if 2<S}\le 5.8;\\ \text{0}\text{.92}\text{, if 5}\text{.8<S}\le 6.\end{array}$$

Private certainties $$U_{{\omega }_{\mathcal{l}}}$$ in the studied classes of states l=D, P is determined by the corresponding membership functions:

The decision rules work in such a way that, based on the selected sonogram, the value of S is determined with an accuracy of 0.1, with the calculation of the corresponding values of the membership functions _, which determine the confidence in the studied classes of states.

The decision about the nature of the disease is made based on the maximum value $$U_{{\omega }_{\mathcal{l}}}$$ .

However, this method was developed for the diagnosis and differentiation of IVH pathology and cannot be used to interpret images of pathological changes in the BDK. The disadvantage of this method is also low diagnostic sensitivity and diagnostic specificity, especially in areas with similar values of confidence coefficients for differentiated classes _, which can lead to diagnostic errors. In addition, in this work, for the differential diagnosis of IVH pathology, idealized images are used as an object of comparison, namely reference diagrams of reference endosonograms, which, without demonstrating real reference images that are familiar to the perception of a practitioner of ultrasound diagnostics, leads to his partial disorientation and reduces the effectiveness of the method . Extrapolation of options for longitudinal and transverse visualization of reference sonograms to display the pathology of the BDK has no practical significance due to the small size of the diagnosed structures and the close values of their measurements in different scanning planes. In practical terms, the characteristic image of the pathological focus is often maximally manifested in one of the sections, or these images are very identical.

The technical objective of the proposed method is to reduce the number of errors in the differential diagnosis of obstructive pathology of the BDK.

To solve this problem, the methodology for the synthesis of hybrid fuzzy decision rules (MFDR) was used as a mathematical research tool, which is based on the symbiosis of the intelligence of a medical specialist, a cognitive engineer and artificial intelligence. The intelligence of clinicians compensates for the lack of statistics and allows, together with a cognitive engineer, to build formal models based on poorly formalized data. A cognitive engineer conducts exploratory analysis and selects mathematical models that correspond to the structure of the problems being solved. A rational combination of intelligent components is provided by the corresponding MSGNRP algorithm. Such interaction provides the opportunity to generate new information systems that solve problems of image interpretation and ensure the correctness of decisions made in conditions of uncertainty in the source data. To improve the quality of differential diagnosis of endosonograms, studies have been organized with the involvement of experts who are focused on diagnosing the pathology of BDK. The results of treatment of 188 patients with BDK pathology were subjected to prospective and retrospective analysis. Among them, 134 (71.3%) patients with benign stenosis of the spinal cord, 24 (12.8%) patients with benign neoplasms, 31 (16.5%) patients with cancer of the spinal cord. The age of the patients ranged from 24 to 82 years, the average age was 69+6.3 years. The majority of patients were women - 63.4%. Among this group of patients, 104 (55.4%) people suffered from cholelithiasis, 32 (30.8%) of whom had previously undergone cholecystectomy at various times; accordingly, 84 (44.6%) patients did not have cholelithiasis.

The essence of the invention is that for the differential diagnosis of obstructive pathology of BDDC according to classes ω _P - “malignant pathology of BDDC” and ω _D - “benign pathology of BDDC”, EUS of the hepatopancreatoduodenal zone is performed. Video images of EUS of patients with BDK pathology are analyzed, image scales are formed that reflect the echographic structure of the studied objects, consisting of a set of endosonograms of all examined patients. From the videos, the optimal images are selected that most accurately convey the anatomical structures being studied and the identified pathology. Sonographic images on a scale of echographic structures are ranked in order from normal BDK to papillary cancer with malignant infiltration into surrounding structures. The lack of a formal procedure for the formation of the initial system of signs used to solve the problem of differential diagnosis of the pathology of the BDK encourages one to rely on the experience of ultrasound diagnostic experts focused on the pathology of the pancreaticobiliary zone organs. To optimize differentiation, the pathology of the BDK is presented in the form of two images: the first image is a reference sonogram, and the second is an artificially created, idealized image (reference), displaying only significant structures isolated from the reference sonogram, which allows focusing on the actual pathology and providing more accurate diagnosis.

From a mathematical point of view, the problem of classifying the studied classes of diseases using endosonograms belongs to the class of poorly formalized problems that are advisable to solve using the theory of fuzzy decision logic and, in particular, MSGNRP, which was developed to solve problems with a data structure similar to our problem. Recommendations for the practical use of MSGNRP, including solving problems similar to the proposed method, are described in the work of Korenevsky N.A. Methodology for the synthesis of hybrid fuzzy decision rules for medical intelligent decision support systems NA Korenevsky, SN Rodionova, AI Khripina. - Stary Oskol: TNT. - 2019. - 472 p.

Analysis of reference endosonograms of the echographic structure allowed us to conclude that specialists confidently differentiate the observed images and correlate them with reference sonograms and reference diagrams. This determines the possibility of significantly increasing the accuracy in decisions made by moving from the order scale (number of sonograms) to a more powerful interval scale S with displaying a measure of proximity between reference endosonograms of echostructures of obstructive pathology of the IVH. In the course of the research, taking into account the opinions of highly qualified experts and the data structure of the problem being solved, the interval scale S was constructed according to the following rules. The numbers of the main reference sonograms were assigned integers from 1 to 6, taking into account the fact that with the growth of these numbers, the confidence in the decisions made in the class ω _P through the class ω _D increases. A scale segment of 0.1 was selected as the unit of measurement. The mechanism for constructing the S scale is described in detail in the works:

Belozerov V.A., Korenevsky N.A. Differential diagnosis of pancreatic pathology according to endoscopic ultrasonography based on fuzzy mathematical models // System analysis and management in biomedical systems, volume 20, no. 1, 2021.- P. 102-110.

Belozerov V.A., Korenevsky N.A. Fuzzy models of differential diagnosis of focal formations of the pancreas according to ultrasonography based on the analysis of reference patterns of reference endosonograms //. News of the Southwestern State University. Series: Management, computer technology, computer science. Medical instrumentation, volume 10 No. 3/4, 2020._С102-117.

Belozerov V.A., Korenevsky N.A., Korzhuk N.L. Diagnosis of pathology of the extrahepatic bile ducts according to endoscopic ultrasonography using fuzzy mathematical models // Proceedings of the South-West State University. Series: Management, computer technology, computer science. Medical instrumentation. 2022. T. 12, No. 2. P. 149-164. https://doi.org/10.21869/ 2223-1536-2022-12-2-149-164.

The resulting scales S allow us to judge how close the observation object with the reference image is to the studied classes of states ω _Р and ω _Д according to the membership functions constructed on them $${\mu }_{{\omega }_D}(S)$$ And $${\mu }_{{\omega }_R}(S)$$ . The sought membership functions were constructed in accordance with the recommendations of the work of Korenevsky N.A. Methodology for the synthesis of hybrid fuzzy decision rules for medical intelligent decision support systems. NA Korenevsky, SN Rodionova, II Khripina. - Stary Oskol: TNT. - 2019. - 472 p.

Analysis of the obtained membership functions showed that their use in decision rules gives an unacceptably wide zone of uncertain solutions and low sensitivity for the class ω _Р.

Taking this into account, we organized special studies, during which, in addition to the main sonographic characteristics of differentiated pathology reflected in the reference sonograms, we studied the diagnostic capabilities of a group of sonographic signs, the most pronounced characteristics of which cannot be displayed within a single reference sonogram. This group also includes features with quantitative characteristics, since their maximum values are also not displayed within the reference sonogram. In addition, when implementing a mechanism for differential diagnosis of IVH pathology, the clinician inevitably encounters a pathology whose confidence coefficients are close, which can negatively affect the diagnostic sensitivity and specificity of the method. The area with similar confidence coefficients includes the most difficult to differentiate pathology with similar echographic signs. It primarily includes benign strictures of the bile ducts and malignant ones, up to 90% of which are cholangiocarcinomas. Differentiation of this pathology presents the greatest difficulties, since cholangiocarcinoma in most cases is characterized by an infiltrative growth pattern, and manifests itself in the form of focal thickening of the duct wall and stricture of the bile ducts without the formation of a tumor mass. At the same time, various benign strictures, including iatrogenic trauma resulting from surgical interventions, as well as scar-inflammatory strictures, mimic infiltrative cholangiocarcinomas. Taking into account what has been said at the expert level, to increase confidence in the diagnosis ω _P - malignant obstruction of the CBD, a space of informative features Xi was obtained consisting of: X1 - CBD diameter over 14 mm; X2 - symptom of two ducts; X3 - signs of metastatic lymphadenopathy; X4 - signs of invasion into the great vessels. This group of informative signs includes signs that are not directly characteristic of the pathological focus, but significantly influence the differential diagnosis of benign and malignant lesions of the BDDC. All signs are binary encoded: 0 - the sign is absent (does not correspond to the selected condition); 1 - the sign is present (corresponds to the selected condition). Using the recommendations of the MSGNRP for each of the informative signs Xi, confidence coefficients CIi were obtained that the patient has a malignant CBD obstruction (class ω _P ). For zero values of Xi - KUi=0. For single values of Xi the following set of confidence coefficients was obtained: KU1=0.45; KU2=0.45; KU3=0.7; KU4=0.7.

If there are several informative features from this group, confidence in ω _P is determined by the aggregation of KUi.

The essence of the invention is illustrated by drawings.

Figure 1 shows an image of the echographic structure of a normal BDK: a - endosonogram, b - reference diagram;

Figure 2 shows an image of the echographic structure of the BDK with sclerosing papillitis; a - endosonogram, b - reference diagram;

Figure 3 shows an endosonogram of the structure of the BDK in hyperplastic stenotic papillitis; a - endosonogram, b - reference diagram;

Figure 4 shows an image of the structure of the BDK adenoma; a - endosonogram, b - reference diagram;

Figure 5 shows an image of the echographic structure of BDK cancer without infiltration of surrounding tissues; a - endosonogram, b - reference diagram;

Figure 6 shows an image of the echographic structure of BDK cancer with infiltration of surrounding tissues; a - endosonogram, b - reference diagram;

Figure 7 shows graphs of the functions of membership in classes ω _Р and ω _Д on the S scale.

Analytically, the graphs of membership functions shown in Figure 7 are described by the expressions:

Confidence in the studied classes of diseases on the S scale is determined by the expressions:

Confidence in class ω _P for a group of additional features is determined by the expression:

Where $$U{D}_{{\omega }_P}(1)=\mathrm{TO}{U}_1$$ .

Taking into account the recommendations of the work Korenevsky N.A. Methodology for the synthesis of hybrid fuzzy decision rules for medical intelligent decision support systems NA Korenevsky, SN Rodionova, AI Khripina. - Stary Oskol: TNT. - 2019. - 472 p. the final confidence for class ω _D is determined by expression (3), and for class ω _P by an expression of the form:

In differential diagnosis, the decision is made in favor of the class with greater final confidence.

Specific examples of method implementation.

Patient A

Based on the analysis of the endosonogram in patient A, the doctor determined: S = 2.1. There are no additional signs.

For this value S the membership function is: $${\mu }_{{\omega }_D}(S)=0.85;{\mu }_{{\omega }_P}(S)=\mathrm{0.12.}$$

Diagnosis:

With a confidence of 0.85, Patient A has a benign BDSM obstruction, the confidence of a malignant BDSB obstruction is extremely low, so the final diagnosis is benign SBDC obstruction.

Patient B

Based on the analysis of the endosonogram in patient B, the doctor determined: S = 4.5 There are no additional signs.

For this value S the membership function is: $${\mu }_{{\omega }_D}(S)=0.43;{\mu }_{{\omega }_P}(S)=\mathrm{0.58.}$$

Conclusion. Confidence in diagnoses is low. Further examination or follow-up is required.

Patient B

Based on the analysis of the endosonogram in patient B, the doctor determined: S = 5.8. Additional features include features X2 and X3.

For this value S the membership function is: $${\mu }_{{\omega }_D}(S)=0.06;{\mu }_{{\omega }_P}(S)=\mathrm{0.9.}$$

Confidence in ω _P based on additional criteria: $$U{D}_{{\omega }_P}=\mathrm{0.84.}$$

Final confidence in ω _Р $$U{F}_{{\omega }_P}=\mathrm{0.98.}$$

Diagnosis:

With a confidence of 0.98, patient B has malignant obstruction of the BDK.

Patient G.

Based on the analysis of the endosonogram of patient G, the doctor determined: S = 5.1. Additional features include features X2 and X4.

For this value S the membership function is: $${\mu }_{{\omega }_D}(S)=0.26;{\mu }_{{\omega }_P}(S)=\mathrm{0.77.}$$

Confidence in ω _Р based on additional criteria: $$U{D}_{{\omega }_P}=\mathrm{0.84.}$$

Final confidence in ω _Р $$U{F}_{{\omega }_P}=\mathrm{0.96.}$$

Diagnosis:

With a confidence of 0.96, patient G has malignant obstruction of the BDK.

Claims (10)

A method for differential diagnosis of obstructive pathology of the major duodenal papilla, which consists in the fact that, according to endoscopic ultrasonography data, six pairs of endosonograms are constructed, and each pair is assigned integers that form a naming scale, from which they move to an interval scale displaying on it a measure of proximity to the studied echo structures, and perform endoscopic ultrasonography of the hepatopancreatoduodenal zone, characterized in that the papillary region of the duodenum is visualized and the patient’s endosonogram is compared with six reference endosonograms and reference diagrams: C1 - normal major duodenal papilla (MDP); C2 - echo of the graphic structure of the BDK in sclerosing papillitis; C3 - image of the echographic structure of the BDK with stenosing hyperplastic papillitis; C4 - echographic structure for adenoma of the BDDC; C5 - echographic structure for BDDC cancer without infiltration of surrounding tissues; C6 - echographic structure for BDK cancer with infiltration of surrounding tissues, while a reference diagram is added to each reference sonogram, and the ordinal scale of sonogram numbers is replaced by the interval scale S, on which the class membership functions are determined: malignant obstruction and benign obstruction , which are calculated using the formulas Where - membership function of the class benign obstruction with the base variable S, - function of membership in the class of malignant obstruction with the basic variable S, S - the basic variable determined by the doctor according to the scale of patterns of the echographic structure, and confidence in the presence of benign obstruction in the patient is determined by the expression and to assess the patient’s confidence in the presence of malignant obstruction additionally, informative signs are measured: X ₁ - CBD diameter over 14 mm; X ₂ - symptom of two ducts; X ₃ - signs of metastatic lymphadenopathy; X ₄ - signs of invasion into the great vessels, and these signs are binary coded: 0 - the sign is absent and 1 - the sign is present, and each sign is associated with the coefficient of confidence KU _i that the patient has a malignant obstruction of the BDK, and at the same time for zero values X _i - KU _i =0, and for single values X _i confidence coefficients are specified with the following values: KU ₁ =0.45; KU ₂ =0.45; KU ₃ =0.7; KU ₄ =0.7, and confidence in the class of malignant obstruction according to a group of additional signs is determined by the expression Where and taking into account the value and additional signs, the final confidence that the patient has a malignant obstruction is determined by the expression of the form and in differential diagnosis, the decision is made in favor of the class with greater final confidence.