Method and system for assisting in the identification of a gastrointestinal pathology from endoscopic images

Abstract

The invention relates to a method for assisting in the identification of a gastrointestinal pathology from endoscopic images, comprising the steps of: -a) retrieving at least one endoscopic image of a surface of a wall of the digestive tract; -b) retrieving information on the position of the surface of the wall of the digestive tract present on the image; -c) retrieving information on the dimension of the surface present on the image; -d) displaying the endoscopic image to a user with an indication of the position of the surface and an indication of the dimension of this surface; -e) requesting the user to select a pathology from a list of predefined proposed pathologies; -f) storing the user's selection; -g) displaying the user's selection for another user. Figure to be published with the abstract: Fig. 1

Description

Title of the invention: Method and system for assisting in the identification of a gastrointestinal pathology from endoscopic images

[0001] The invention relates to gastroenterological diagnostics, and in particular to assistance in gastroenterological diagnostics carried out on the basis of an endoscopic examination.

[0002] Endoscopic examination proves to be particularly useful in the identification of pathologies in gastroenterology, making it possible to identify possible cancerous areas in the digestive tract.

[0003] The variety of pathologies and the visual similarities between certain pathologies make their identification particularly complicated, in particular because each practitioner is not confronted with all of the pathologies or cannot be confronted with certain pathologies with sufficient frequency to acquire significant experience.

[0004] The invention aims to resolve one or more of these drawbacks. The invention thus relates to a method for assisting in the identification of a gastrointestinal pathology from endoscopic images, comprising the steps of: -a) recover at least one endoscopic image of a surface of a wall of the digestive tract; -b) recover information on the position of the surface of the wall of the digestive tract present on the image; -c) recover information on the dimension of the surface present on the image; -d) displaying the endoscopic image to a user with an indication of the position of the surface and an indication of the size of this surface; -e) require the user to select a pathology from a list of predefined proposed pathologies; -f) remember user selection; -g) display user selection for another user.

[0005] The invention also relates to the following variants. Those skilled in the art will understand that each of the features of the following variants can be independently combined with the above features, without constituting an intermediate generalization.

[0006] According to a variant, during step e), a pictogram representative of the schematic aspect of the pathology is displayed for each of said pathologies in the list.

[0007] According to a further variant, the method comprises repeating steps d), e) and f). for several users, the method further comprising h) numerically calculating the selection percentage for each of the pathologies based on the users' selections, and further comprising i) displaying the calculated percentage for each of the pathologies proposed to an additional user during a step e).

[0008] According to another variant, the method further comprises a step j) of numerical calculation of the probability of a pathology as a function of the location, size, color and / or appearance of the surface of the wall of the digestive tract on said image, the method further comprising k) displaying the probability calculated for each of the pathologies proposed to an additional user during step j).

[0009] According to another variant, the method comprises the simultaneous display for each of said pathologies in the list, of a pictogram representative of the schematic aspect of this pathology, of the percentage of selection of this pathology by users, and of the calculated probability of pathology.

[0010] According to yet another variant, during step e), the user is successively invited to select a type of lesion, a color of lesion, an aspect of vascular relief of the lesion tissue, and / or an aspect of the mucosal relief of the lesion tissue, invitations proposing a list of choices defined according to a previous selection by the user.

[0011] According to one variant, the method comprises preliminary steps: -memorization of said endoscopic image of the surface of a wall of the digestive tract; -memorizing the position and size of said snapshot.

[0012] Other characteristics and advantages of the invention will emerge clearly from the description given below, for information purposes only and in no way limiting, with reference to the appended drawings, in which:

[0013] [Fig-1] is a schematic representation of an example of a sequence of steps implemented works in a method for assisting in the identification of a gastrointestinal pathology according to an exemplary embodiment of the invention;

[0014] [Fig.2] is an example of a communication architecture for the implementation of the invention;

[0015] [Fig.3] is an example of a display for authenticating a user;

[0016] [Fig.4] is an example of an interface for assisting in taking endoscopic images;

[0017] [Fig.5] is an example of a step of taking an endoscopic image during a step of the process;

[0018] [Fig.6] is an example of an endoscopic image documentation interface;

[0019] [Fig.7] illustrates an example of displaying a snapshot and information for a practitioner advice ;

[0020] [Fig.8] illustrates an interface for selecting a type of lesion;

[0021] [Fig.9] illustrates an interface for selecting a lesion color;

[0022] [Fig. 10] illustrates an interface for selecting a mucosal relief aspect of the lesion tissue;

[0023] [Fig. 11] illustrates an interface for selecting an aspect of the vascular relief of the lesion tissue;

[0024] [Fig. 12] illustrates an interface for displaying a summary of the opinions of consulting practitioners;

[0025] [Fig. 13] illustrates a set of pictograms that can be displayed to provide a selection of categories of lesion types to be selected by a practitioner;

[0026] [Fig. 14], [Fig. 15] and [Fig. 16] illustrate a set of pictograms that can be displayed to suggest types of lesions to be selected by a practitioner;

[0027] [Fig. 17] illustrates a set of pictograms that can be displayed to suggest lesion colors for a practitioner to select;

[0028] [Fig. 18] illustrates a set of pictograms that can be displayed to suggest mucosal reliefs to be selected by a practitioner;

[0029] [Fig. 19] illustrates a set of pictograms that can be displayed to suggest vascular reliefs for a practitioner to select.

[0030] [Fig.l] schematically illustrates a set of steps implemented in a method for assisting in the identification of a gastrointestinal pathology from endoscopic images, according to an exemplary implementation of the invention.

[0031] The invention aims to enable a practitioner to have assistance from a community of practitioners and computer processing tools in order to have more criteria before being able to decide on a precise diagnosis. Indeed, different gastrointestinal pathologies may present relatively similar aspects when examining endoscopic images while the associated therapeutic decisions may prove to be radically different. In addition, certain pathologies may only be encountered at a low frequency, making their identification more difficult for the practitioner who is confronted with them.

[0032] For this purpose, a communication architecture 1 as illustrated in [Fig.2] can be implemented. This communication infrastructure comprises a server 2 accessing a database 21 and having processing capabilities. User terminals 4 can communicate with the server 2 via a telecommunications network 3.

[0033] The method for assisting in the identification of a gastrointestinal pathology from endoscopic images of the wall of the digestive tract can be implemented with the following steps:

[0034] -a) recovering at least one endoscopic image of a surface of a wall of the tube digestive; -b) recover information on the position of the surface of the wall of the digestive tract present on the image; -c) recover information on the dimension of the surface present on the image; -d) displaying the endoscopic image to a user with an indication of the position of the surface and an indication of the size of this surface; -e) require the user to select a pathology from a list of predefined proposed pathologies; -f) remember user selection; -g) display user selection for another user.

[0035] Such a method allows a remote practitioner to have a maximum of information in order to enable him to make the most accurate diagnosis possible despite his distance from the place where the endoscopic image was taken. Such a practitioner thus has additional information and not just a simple image. The diagnostic suggestion made by this practitioner thus allows another user to consult this diagnosis and thus be able to refine his own diagnosis, by having a good histological prediction in order to be able to make the right therapeutic choice.

[0036] A more detailed example of the method will be described. During a step 100, a user connects via a terminal 4 to the server 2 and authenticates himself with a graphical interface, as illustrated in [Fig. 3]. This graphical interface can be offered via an internet browser or via a terminal application. The display can be optimized for a PC or for a smartphone. The user is assigned a profile based on his recognized experience recorded on the server 2. Depending on the level of intervention during use, the user is assigned either a beginner practitioner profile, an advanced practitioner profile or an expert practitioner profile. The practitioners can either intervene as an interviewer or as an advisor.A practitioner will act as an interviewer if they are performing an endoscopic examination and wish to have assistance in identifying gastrointestinal pathology. A beginner practitioner profile may be assigned by a practitioner specializing in gastroenterology who does not yet have recognized expertise. An advanced practitioner profile may be assigned to a practitioner with already recognized expertise regarding certain pathologies. An expert practitioner profile may be assigned to a practitioner with even more recognized expertise. The change of profile may be validated following ratings by other practitioners, following the validation of exercises (for example implemented by server 2) or following the validation of expertise by other practitioners. A greater number of expertise levels for the profiles may of course be considered.User profiles can also be selected later, after authentication, depending on the server 2 applications used.

[0037] The start of the method corresponds to steps implemented for an interviewing practitioner. If the connection to the server 2 precedes the endoscopic examination itself, the interviewing practitioner can advantageously benefit from assistance in taking images during the endoscopic examination, corresponding to step 101.

[0038] The server 2 manages the interfaces displayed to one or more users during the course of the process, as well as the digital processing of the responses provided by the users. An example of a succession of displays during assistance in taking endoscopic images of the wall of the digestive tract is illustrated in [Fig. 4]. The display 1011 invites the questioning practitioner to take an endoscopic image of the lesion in white light and in global view. The practitioner can either select an endoscopic image already taken by the button 10111, or take the endoscopic image by means of the button 10112. The display 1012 invites the questioning practitioner to specify whether he has identified an area of ​​interest, specifying the direction: an area of ​​interest corresponds to an area of ​​focal degeneration, an absence of an area of ​​interest corresponds to a homogeneous lesion.The practitioner can either mention the identification of an area of ​​interest using button 10121, or the absence of an area of ​​interest using button 10122. Display 1013 invites the questioning practitioner to take an endoscopic image using chromoendoscopy on any area of ​​a lesion. The practitioner can either select an endoscopic image already taken using button 10131, or take the endoscopic image using button 10132. Other sequences of displays may of course be proposed depending on the specificities of the pathology being sought.

[0039] During step 102, the questioning practitioner is therefore invited to take one or more endoscopic images. [Fig.5] illustrates an example of an endoscopic image that can be taken.

[0040] During a step 103, the questioning practitioner is invited to specify context information for the endoscopic image, with a view to proposing information usually unavailable on a simple image: in particular the location of the lesion in the digestive tract, the size of the lesion or the image scale, an inflammatory context, the brand and model of the endoscope, and / or the zoom magnification used to produce the endoscopic images. [Fig. 6] illustrates an example of a practitioner questioning interface. An image of the lesion may be displayed as a reminder, as well as a schematic representation of an intestinal duct. The user may be asked to specify the position of the lesion, for example via a drop-down menu 1031. The position of the lesion may be illustrated graphically by highlighting or an arrow on the schematic representation of the intestinal duct.The user may be asked to specify the size of the lesion or the scale of the image, for example as illustrated by the 1032 drop-down menu. The size entry can be . performed by a selection of ranges or by a selection on a graduated scale. This information may optionally be retrieved automatically by communication with the endoscope or the extraction of metadata from the image. The user may be prompted to enter comments, for example in an input field 1033. The user may also be prompted to specify the model and brand of the endoscope, for example via a drop-down menu 1034. This information may optionally be retrieved automatically by communication with the endoscope or the extraction of metadata from the image. The user may be prompted to specify the zoom magnification used for the endoscope image, for example on a graduated scale 1035. This zoom magnification may optionally be retrieved automatically by communication with the endoscope or the extraction of metadata from the image.

[0041] In step 104, the various context information is associated with the snapshot or series of snapshots and is stored in the database 21.

[0042] In step 105, the snapshots and the associated context information are made accessible to a community of practitioners having terminals 4.

[0043] In step 106, the endoscopic images and the associated context information, in particular the information on the position of the surface of the wall of the digestive tract present on the image or the dimension of the surface present on the image, are displayed to a consulting practitioner. [Fig.7] illustrates an example of image and information display for a consulting practitioner, with a beginner, advanced or expert profile, by displaying an endoscopic image, a location of the endoscopic image, text added by the questioning practitioner, an indication of the dimensions of the image and a magnification corresponding to this image.

[0044] In step 107, the consulting practitioner (beginner, advanced or expert) is invited to make a selection from proposals for characterizing the lesion. Figures 8 to 11 illustrate an example of a succession of different steps of such an invitation. Such an invitation can also be submitted in advance to the questioning practitioner and his responses can be submitted to the consulting practitioners to facilitate their selection.

[0045] In [Fig.8], the location and size of a lesion image are recalled to the consulting practitioner. The lesion image can be displayed to facilitate selection. A list of predefined lesion types with pictograms can be displayed to the consulting practitioner. The pictograms are advantageously as illustrative as possible of the associated lesions, to facilitate selection. The list can be accessible via a drop-down menu 1050. The list of lesions can be dependent on the location and size displayed. The published Paris and LST classifications can be used to qualify the type of lesions. A confirmation invitation 1051 is displayed to allow confirmation of the selection made.

[0046] In [Fig.9], the selected lesion type is recalled in the cartridge 1052. The practitioner is invited to select a lesion color type from a predefined list, for example by means of a drop-down menu 1053. The predefined list can be proposed according to the previous selection. A confirmation invitation 1054 is displayed to allow confirmation of the selection made.

[0047] In [Fig. 10], the type of lesion is recalled in the cartridge 1052 and the type of color is recalled in the cartridge 1055. The practitioner is invited to select a vascular relief aspect of the lesion tissue from a predefined list, for example by means of a drop-down menu 1056. The predefined list can be proposed according to the previous selections. The published Sano classification can be used to qualify the appearance of the vascular relief of the lesion tissue. A confirmation invitation 1057 is displayed to allow confirmation of the selection made.

[0048] In [Fig. 11], the type of lesion is recalled in the cartridge 1052, the type of color is recalled in the cartridge 1055 and the appearance of the vascular relief of the lesion tissue is recalled in the cartridge 1058. The practitioner is invited to select details on the appearance of the mucosal relief of the lesion tissue from a predefined list, for example by means of a drop-down menu 1059. The predefined list can be proposed according to the previous selections. The published Kudo classification can be used to qualify the appearance of the mucosal relief of the lesion tissue. A confirmation invitation 1060 is displayed to allow confirmation of the selection made.

[0049] The use of pictograms associated with predefined lesion characterizations makes it possible to obtain an efficient and rapid characterization of lesions on a single solution, without requiring recourse to medical examination guidelines or having knowledge of existing classifications.

[0050] Advantageously, the server 2 offers an interface further enabling a practitioner to add annotations or notes, either for one or more characterizations, or for the lesion itself, indicating for example its final histology determined after resection.

[0051] The practitioner's various selections and annotations are stored in the database 21 at step 108.

[0052] Advantageously, steps d), e) and f) are repeated for several users. In step 109, the server 2 numerically calculates the percentage of responses from the different users for the lesion characterizations, for example the mucosal relief aspect of the lesion tissue, the type of lesion, the color of the lesion, and / or the vascular relief aspect of the lesion tissue. These percentages are then stored by the server 2 and made accessible to practitioners with the required rights. Thus, in step 110, the responses and / or the percentages are made accessible to authorized users. In step 111, the questioning practitioner can thus consult the responses or the summary of the responses provided by the consulting practitioners.

[0053] It can be expected that these percentage calculations will be weighted. It can thus be considered that the selection of an expert practitioner will have more weight than the selection of an advanced practitioner, and even more than the selection of a beginner practitioner.

[0054] Advantageously, the percentage calculated for the lesion characterizations can be displayed to users. Advantageously, this percentage can be displayed simultaneously with an associated pictogram. [Fig. 12] illustrates an example of such a display. The invention allows efficient and rapid visualization on a single interface of the practitioners' responses. It is thus possible to have assistance in evaluating the concordance between an initial characterization proposed by the questioning practitioner and a majority characterization of the consulting practitioners.

[0055] Depending on the rights assigned to a user, the user will also be able to consult the responses of each of the other practitioners, for example to be able to consult the associated notes.

[0056] According to another aspect of the invention, the probability of a pathology is calculated digitally by the server 2 via an algorithm, based on the location, size, color and / or appearance of the surface of the wall of the digestive tract on said image. Such a calculation may for example be based on artificial intelligence learning, from images, positions, dimensions of lesions, and based on corresponding responses provided in the past. Such a probability may be returned to the users. The restitution of this probability may be separate from the display of the percentages calculated based on the responses of the practitioners. The restitution of this probability may be simultaneous with the display of the percentage calculated on the basis of the responses of the users. The results of the artificial intelligence may be presented in the form of a specific user, with a specific avatar.

[0057] Advantageously, the server 2 can provide a training interface for practitioners. This training interface can sequentially display images with their respective position and size. For each of these images, the user can be offered the sequence of steps illustrated in Figures 8 to 11. Once the user's various selections have been retrieved, the server 2 can display the satisfactory responses for the pathology associated with each of the images. The satisfactory responses can be displayed side by side with the user's responses, as well as with comments associated with the pathology and the criteria for its selection. Thus, practitioners can progress in their practice and experience without having to encounter a large number or a wide variety of clinical cases.A user's level of expertise may be modified based on the answers provided during training sessions on server 2.

[0058] Both artificial intelligence learning and the variety of practitioner training will benefit from a progressive enrichment of a database of clinical cases already submitted.

[0059] Advantageously, the method comprises a step of assistance for the questioning practitioner. This assistance step may in particular comprise a list of advice on taking photographs. This advice may, for example, indicate preferred positions to photograph, indicate a type of light to use for taking the photograph, indicate a type of contrasting product to use.

Claims

Claims

1. Method for assisting in the identification of a gastrointestinal pathology from endoscopic images, characterized in that it comprises the steps of: -a) recovering at least one endoscopic image of a surface of a wall of the digestive tract; -b) recovering information on the position of the surface of the wall of the digestive tract present on the image; -c) recovering information on the dimension of the surface present on the image; -d) displaying the endoscopic image to a user with an indication of the position of the surface and an indication of the dimension of this surface; -e) requesting the user to select a pathology from a list of predefined proposed pathologies; -f) storing the user's selection; -g) displaying the user's selection for another user.

2. Method for assisting in the identification of a gastrointestinal pathology according to claim 1, in which during step e), a pictogram representative of the schematic aspect of the pathology is displayed for each of said pathologies in the list.

3. A method for assisting in the identification of a gastrointestinal pathology according to any one of the preceding claims, comprising repeating steps d), e) and f) for several users, the method further comprising h) numerically calculating the percentage of selection for each of the pathologies based on the users' selections, and further comprising i) displaying the percentage calculated for each of the pathologies proposed to an additional user during a step e).