Presenting dental ai findings

Abstract

The present invention provides a system, method and accompanying user-interface (UI) for reviewing artificial intelligence-based findings for dental images, ensuring that clinically relevant findings are accurately identified, documented, and acted upon without unnecessary repetition and delay.

Description

[0001] Presenting Dental Al Findings

[0002] Field of the invention

[0003] The present invention relates to a system, method and accompanying user-interface (Ul) for reviewing artificial intelligence-based findings for dental images, ensuring that clinically relevant findings are accurately identified, documented, and acted upon without unnecessary repetition and delay.

[0004] Background

[0005] Dental clinicians regularly take x-ray images of the teeth of their patients as a part of routine diagnostics. During a single session, a series of x-ray images may be taken, some of which overlap and thus contain common diagnostic information. Each image must be meticulously inspected by the dentist to identify and document both clinical findings (such as pathologies or injuries) and non-clinical findings.

[0006] These x-ray images can include periapical radiographs, which provide detailed views of individual teeth and their surrounding bone structure, and panoramic radiographs, which offer a comprehensive view of the entire mouth, including the teeth, jaws, and surrounding tissues. Bitewing radiographs, which capture the crowns of the upper and lower teeth simultaneously, are also commonly employed to detect cavities and monitor the health of existing fillings. Moreover, dentists may combine intraoral x-rays with panoramic x-rays to get a more comprehensive view, and corresponding Al findings can be identified in both image types. Furthermore, dental clinicians often take Full Mouth Series (FMX) of x-rays, which consist of multiple images providing a detailed examination of the entire mouth. Some of the images in an FMX series overlap and may thus comprise corresponding Al findings.

[0007] Recently, Artificial Intelligence (Al) has become available to support the identification of pathological and non-pathological findings in dental images. However, not all Al-generated findings hold clinical relevance. Some Al findings can be false positives or inconsequential. Moreover, among the clinically relevant findings identified by Al, certain findings may require treatment while others may not. This necessitates that the clinician reviews the Al-generated findings for each image to determine their relevance and decide which findings should be kept in the patient's file and which may be ignored. For the releva nt findings, the dentist may further want to document which of these findings necessitate treatment or should be monitored in future visits. This review process and acceptance of the releva nt findings advantageously allows to build a well-documented patient file.

[0008] This process demands that the dentist inspects each Al finding individually and decides on whether to ignore the finding or retain it in the patient's file, possibly accompanied by clinician’s comments. Clinicians often find this review process to be tedious and time-consuming, especially as it is often done during the patient visit. Consequently, it happens that a clinician only glances at an Al-annotated image to identify findings immediately relevant to the treatment session, without specifically accepting the clinically relevant Al finding into the patient file.

[0009] Valuable updates to the patient file may be missed, leading to potential missed treatment opportunities or the unnecessary repetition of diagnostic imaging. Therefore, there is a need for strategies and workflows that facilitate the Al findings review by the clinician. These strategiesshould also entice and motivate the clinician to accept the clinically relevant Al findings into the patient file.

[0010] It is an object of the present invention to provide a method and accompanying user- interface (U I) enabling an efficient review process of Al findings for dental images, ensuring that clinically relevant findings are accurately identified, documented, and acted upon without unnecessary repetition and delay.

[0011] Summary of the Invention

[0012] The present invention relates to a system and computer computer-based method for reviewing annotated dental images. The method involves retrieving a dataset of dental images of the same patient with annotations generated by one or more Al algorithms. Each annotation comprises a type label indicating the type of Al finding, a bounding shape and / or an image mask with image coordinates for overlaying the bounding shape and / or image mask over the dental image at the position of the Al finding. Preferably, the boundingshape encloses a dental image area comprising the Al finding, while the image mask specifically covers this area. Each annotation further comprises a label indicating the anatomical position of the Al finding. In an embodiment, the anatomical position refers to an identification of the tooth position, such as the tooth number, that comprises or is closestto the Al finding. However, for selected Al findings, such as caries, the anatomical position label may be complemented with further information on the actual region of the tooth wherein the Al finding is found. For instance, next to the tooth number, the anatomical position label may indicate whether an Al finding is found in the ‘distal’, ‘mesial’, ‘occlusal’ or ‘bucco lingual’ region of the tooth. Further, each annotation can be set to ‘accepted’, ‘rejected’, or ‘undefined’ or an equivalent status. The method also includes displaying a current dental image with annotations via a graphical user interface (U I), which has a window for presenting the dental image and a text panel wherein the type and anatomical position labels of the annotations can be listed. The method further comprises setting or resetting the status of an annotation to accepted or rejected based on user input.

[0013] Typically, the respective annotations either comprise non-pathology or pathology findings. Nonpathology findings relate to observations, which do not require treatment, while pathology findings may require treatment. Considering that non-pathology findings may be more robust and / or less critical, the annotations comprising non-pathology findings can be automatically set to accepted. In this case, the user- interface may comprise functionality allowing the user to overrule the automatic acceptance and set any one of the annotations comprising non-pathology findings to rejected. In an embodiment the text panel of the Ul only lists the type and anatomical position labels of annotations set to accepted, which allows the user to easily identify accepted annotations. In this embodiment, the automatically accepted annotations are immediately listed in the text panel when loading a current image into the user-interface.

[0014] In an embodiment, annotations comprising pathology findings are set to undefined until the current dental image is reviewed by a user, typically a dental clinician. In an embodiment, when loading a current dental image, undefined annotations comprising pathology findings may be indicated in the user interface by overlaying said bounding shapes over the displayed current dental image. In this way the user’s attention is directed to the areas of the dental image comprising pathology findings, while these image areas can be viewed unobstructed. In a further embodiment, when hovering over a bounding shape of an undefined annotation, the image mask is displayed within the boundingshape overlaying the actual image area comprising the Al finding area. In an embodiment, when the user accepts an annotation with a pathology finding, thebounding shape is deleted from the user interface and the image mask is displayed. In this way the user can clearly differentiate between accepted and undefined annotations comprising pathology findings. In addition, the type and anatomical position labels of an accepted annotations comprising a pathology finding is listed in said text panel of the Ul. It is preferred that any bounding shape and / or image mask overlaying the current dental image in the user interface are deleted when an annotation is set to ignored. Typically, rejected annotations are deleted. In a further embodiment, when displaying a current dental image, it is automatically verified whether undefined annotations of the current image, in particular such annotations comprising pathology findings, correspond to annotations of other dental image in the dataset that were accepted. Correspondence of annotations is determined based on matching type and anatomical position labels. If corresponding accepted annotations are found, it is indicated to the user. In addition, a user-interface element may be generated, which allows the user to jointly accept annotations of the current dental image that correspond to accepted annotations of another image.

[0015] List of Figures

[0016] Figure 1: Generating annotations for a received dental image

[0017] Figure 2: A first workflow for presenting a current dental image to the user for reviewing the annotations

[0018] Figure 3: A second workflow for presenting a current dental image to the user for reviewing the annotations

[0019] Figure 4: A workflow comprising auto-accepting annotations for which corresponding accepted annotations were identified for one or more dental images in the image dataset for a given patient Figure 5: Illustration of the review and acceptance of annotations comprising pathology findings according to the first workflow (see scheme Figure 2).

[0020] Figure 6: Illustration of the acceptance and review of annotations comprising pathology findings according to the second workflow (see scheme Figure 3).

[0021] Figure 7: Illustration of the review and acceptance of annotations comprising pathology findings according to the workflow comprising auto-accepting annotations for which corresponding accepted annotations were identified for one or more dental images in the image dataset for a given patient (see scheme Figure 4).

[0022] Figure 8: Presentation of annotations comprising non-pathology findings.

[0023] Figure 9: Illustration of Al based identification of a tooth surface in a dental image and the segmentation of the identified to tooth surface in an enamel (E), dentin (D) and pulp (P) region Definitions

[0024] Pathology findings: refers to Al based observations, which may require treatment. These observations may relate but are not restricted to areas of the dental image featuring: caries, bone loss, calculus, root canal defects, apical defects or marginal defects. In the present invention an Al based pathology finding is processed into a dental image annotation comprising a type label indicating the type of Al finding (caries, bone loss, ...), a bounding shape and / or an image mask with image coordinates for overlaying the bounding shape and / or image mask over the dental image at the position of the Al finding. Each annotation may further comprise a label indicatingthe anatomical position of the Al finding. This anatomical position label may be obtained by combining information of the Al algorithm used for identifying the pathology finding with information from one or more other Al algorithms identifying anatomical structures. In an embodiment, the anatomical position refers to an identification of the tooth position, such as the tooth number, thatcomprisesorisclosestto the Al finding. However, for selected Al findings, such as caries, the anatomical position label may be complemented with further information on the actual region of the tooth wherein the Al finding is found. For instance, next to the tooth number the anatomical position label may indicate whether an Al finding is found in the ‘distal’, ‘mesial’, ‘occlusal’ or ‘bucco lingual’ region of the tooth. For certain pathology findings the annotation may be further complemented with information on the tooth layers wherein the finding is observed. For instance, running an Al algorithm for segmenting the tooth surfaces in the image into enamel, dentin and pulp layers (Figure 9) allows to indicate whether a pathology finding, such as a caries is situated in one more of these tooth layers.

[0025] Non-pathology finding refers to Al based observations, which do not require treatment. These observations may relate but are not restricted to areas of the dental image featuring: fillings (amalgam, composite), bridge, implant, prosthetic crown post and root canal treatment. In the present invention an Al based non-pathology finding is processed into a dental image annotation comprising a type label indicating the type of Al finding (caries, bone loss, ...), a bounding shape and / or an image mask with image coordinates for overlaying the bounding shape and / or image maskoverthe dental image at the position of the Al finding. Each annotation may further comprise a label indicating the anatomical position of the Al finding. This anatomical position label may be obtained by combining information of the Al algorithm used for identifying the pathology finding with information from one or more other Al algorithms identifying anatomical structures. In an embodiment, the anatomical position refers to an identification of the tooth position, such as the tooth number, that comprises or is closest to the Al finding. However, for selected Al findings, such as fillings, the anatomical position label may be complemented with further information on the actual region of the tooth wherein the Al finding is found.

[0026] Detailed Description

[0027] The method and system of the present invention involves a dental image management system comprising one or more computers running a database for receiving and storing dental image data. The dental image management system further comprises a screen and dental image software for displaying in a user-interface (Ul) one or more dental images permitting a user to review, analyze, and / or annotate the presented dental image. The database and dental image software can be installed on the same computing device or on separate devices connected in a network. These computing devices may be connected to one or more dental imaging devices, such as intraoral or extraoral x-ray sensors, a panoramic dental x-ray device, a CBCT device, an intraoral camera, or an intraoral scanner. Images received from any of these imaging devices are stored in the database along with patient identification and other relevant metadata, including amongst others the age and gender of the patient, date on which the image was taken, type of x-raysensor used. The image information may be in the form of digital imaging and communications in medicine (DICOM) or other image formats (TIF, PNG, JPG, etc.).

[0028] The dental image management system may include algorithms for analyzing dental images, which may allow for the automatic identification of certain anatomical landmarks or structures represented in the images. This identification enables automatic positioning and orientation of the images according to a predefined desired view. For example, identifying the tooth numbers in a series of intraoral x-rays allows the system to automatically position the x-rays in an FMXsetup. The automatic identification of anatomical features, such as tooth numbers, further facilitates grouping images of a patient with common features. For instance, a user inspecting a particular tooth in a current x-ray image may in this way request the dental image analysis software to also present previously acquired images of the same tooth for the same patient. Typically, the identification of anatomical features is performed using artificial intelligence (Al) techniques, such as machine learning techniques. Neural networks and convolutional neural networks have shown to be particularly useful in the analysis of dental images. Many of these Al algorithms for detecting anatomical features are robust and can run in the background without needing user verification or confirmation. Consequently, this Al-driven functionality is valued by dental clinicians as it simplifies workflows without requiring user effort. However, Al-based automatic detection and indication of image areas containing observations that may require treatment, typically require inspection and verification by the user. These pathology findings may include false positives, and not all detected pathologies are clinically relevant. Additionally, there is a possibility that the algorithm might miss clinically relevant findings. Therefore, the user still needs to independently review dental images for the presence of relevant findings. As such, while Al assisted image inspection provides valuable second opinion information, dental clinicians experience the task of reviewing and individually accepting or rejecting annotations comprising pahtology findings as burdensome. Especially, as this inspection is typically done during the patient’s visitand each additional task lengthens the chair-time of the patient.

[0029] Consequently, it occurs that a clinician only glances at an Al-annotated image to identify findings immediately relevant to the treatment session, without specifically accepting all clinically relevant Al-based annotations generated for the dental image. So, there is a need for strategies and workflows that facilitate the review of these annotations by the clinician. These strategies should motivate the clinician to review and accept the clinically relevant Al findings into the patient file. In the context of the present invention, it was found that a clinician may be motivated to do this when an output of the workflow for reviewing and accepting or rejecting the Al-based findings is an annotated view of the dental image that is particularly suitable for presenting and discussing the observed pathologies and treatment options with the patient. Further, dental clinicians appreciated workflows wherein they were alerted that annotations in a current dental image corresponded to annotations that had been accepted in a previously reviewed dental image of the same patient. Mostly they felt comfortable to accept, without detailed verification, annotations in a current image for which such corresponding accepted annotations were identified, which expediates the review process.

[0030] Figure 1 illustrates a possible workflow for generating and storing annotations for a dental image received in the dental image management system according to the present invention. A dental image, for instance an intraoral x-ray, may be received (100) from an imaging device operationally linked to any one of the computers of the dental image management system. After receiving the dental image the computer may automatically start the execution of one or more operations so one or more artificial intelligence techniques can be used to analyse the image and generate annotations (110). The artificial intelligence techniques, such as machine learning techniques, may proces dental image and optional associated information (such as for instance age and gender of the patient) to generate information on the teeth or tooth number positions represented in the image, to segment the areas of the images representing teeth, to segment tooth areas of the image into enamel, dentin and pulp tooth layers (Figure 8), to identify different regions in tooth areas of the image (for instance distal, mesial, occlusal and / or bucco lingual tooth regions), to identify image regions representing possible pathology findings, and / or to identify image regions representing non-pathology findings. Preferably, the different outputs of the Al-based analysis iscombined to identify and discard erroneous observations (120). For instance, a caries found outside of image areas representing teeth can be ignored as a false positive. Similarly, a caries found in an image region representing a dental implant can be ignored. Non-discarded pathology and non-pathology findings and associated information can then be stored (150) as annotations linked to the dental image complementing a set of annotated dental images for a given patient. Each annotation typically comprises a type label indicating the type of Al finding, a bounding shape and / or an image mask (140) with image coordinates for overlaying the bounding shape and / or image mask over the dental image at the position of the Al finding. Preferably, the bounding shape encloses a dental image area comprising the Al finding, while the image mask covers this image area. Each annotation preferably further comprises a label (130) indicating the anatomical position of the Al finding. In an embodiment, the anatomical position refers to an identification of the tooth position, such as the tooth number, that comprises or is closest to the Al finding. However, for selected Al findings, such as caries, the anatomical position label may be complemented with further information on the actual region of the tooth wherein the Al finding is found. For instance, next to the tooth number, the anatomical position label may comprise an indication of whether an Al finding is found in the ‘distal’, ‘mesial’, ‘occlusal’ or ‘bucco lingual’ region of the tooth.

[0031] To allow a user to review the dental image and associated annotations different workflows are proposed. In a first workflow presented in the scheme of Figure 2 a current dental image together with its annotations is loaded into the image display window of the Ul. Preferably, when loading a current dental image (200), the undefined annotations, for instance those comprising pathology findings, may be indicated in the user interface by overlaying said bounding shapes over the current dental image (220). In this way the user’s attention is directed to the areas of the dental image comprising pathology findings, while the original image data in these areas can be viewed unobstructed. When the user hoovers over a bounding shape of an undefined annotation (230), the image mask is displayed within the bounding shape overlaying the actual image area comprising the Al finding area (240). This permits the user to verify whether the image area he considered to comprise a finding within the bounding shape coincides with the area that was identified with the Al technique. Based on this the user may decide to accept or reject an annotation. Upon acceptance of an annotation its bounding shape overlaying the current dental image is deleted from the user interface and the image mask is displayed overlaying the relevant image area. In this way the user can easily differentiate between accepted and undefined annotations. In addition, the type and anatomical position labels of an accepted annotation is listed in said text panel of the Ul. When an annotation is rejected its bounding shape and / or image mask overlaying the current dental image are deleted. So, after processing the originally undefined annotations, the Ul displays the current dental image wherein only the image areas of the accepted annotations are indicated by the overlaying image masks. Particularly, when the annotations comprise pathology findings, this view allows the clinician to clearly and convincingly communicate the need for treatment to the patient.

[0032] Figure 5 illustrates the review and acceptance of annotations comprising pathology findings according to the first workflow (Figure 2). In Figure 5A the user has accepted the finding of bone loss at tooth position 12 indicated by the image mask 501 , while the distal and mesial caries in tooth 12 and the caries in tooth 11 are still undefined and indicated on the dental image by the overlaying bounding shapes 502. The user subsequently accepts both caries in tooth 12 resulting in the replacement of the bounding shapes 502 with the image masks 503 (Figure 5B). After rejecting the finding of the caries in tooth 11, its bounding shape 502 is deleted resulting in a dental image wherein the bone loss at tooth position 12 and the distal and mesial caries in tooth12 are indicated by the image masks 501 and 503. The type and anatomical position labels of the accepted annotations are listed in the text panel 504. This view allows the dentist to clearly explain to the patient which tooth requires treatment and why.

[0033] A second possible workflow for presenting a current dental image to the user for reviewing the annotations is presented in Figure 3. In this workflow before or upon loading a current image into the Lil all undefined annotations are set to accepted 310. The type and anatomical position labels of the accepted annotations are listed in the text panel of the Ul 320. These annotations may be further indicated on the dental image by overlaying bounding shapes and / or image masks in the Ul. Alternatively, the annotations may only be indicated in the dental image when the user selects the type label of an annotation in said text panel. Based on the information presented in the Ul, the user may decide to reject a selected annotation because it is not correct or of little clinical significance 330. Following such rejection the type and anatomical position label of the annotation is deleted from the text paneland any indications of the annotation in the dental image are removed 340.

[0034] Figure 6 illustrates the acceptance and review of annotations comprising pathology findings according to the second workflow (Figure 3). Upon loading the current dental image all caries findings 601 and the bone loss findings 602 are automatically accepted and indicated in the dental image by the image mask 601 , 602. In addition, the type and anatomical position labels of the annotation are listed in the text panel 603. Preferably, the Ul comprises a toggle element permitting to remove and recall the image masks 601, 602 allowing the clinician to inspect the original image data at the location of the image masks. Using this toggle function the clinician may eventually decide to reject one or more of the annotations. When an annotation is rejected, its image mask 601 , 602 and its listing in the text panel 603 are deleted.

[0035] In an embodiment, above-mentioned workflows 1 and 2 are used in combination. For example, workflow 1 is used for reviewing annotations comprising pathology findings, while workflow 2 is used for reviewing annotations comprising non-pathology findings. In another embodiment, the user is given the option whether to use workflow 1 or workflow 2 for reviewing annotations of a current dental image. For instance, if based on an initial inspection the clinician considers that most of the bounding shape indicated pathology findings appear clinically relevant he may decide to use workflow 2 for processing the annotations comprising pathology findings. Alternatively, when some initial inspection reveals that one or more of the indicated pathology findings seem of little clinical significance, he may decide to individually review the annotations using workflow 1. Figure 4 illustrates a workflow offering the option to semi-automatically accept annotations for which corresponding accepted annotations were identified for one or more dental images in the image dataset of a given patient. When loading a current dental image of a patient 400 and associated annotations 410 in the Ul, the computer automatically verifies for undefined annotations whether corresponding annotations were accepted in previously reviewed images of the image dataset of the patient 420. Annotations of two different dental images are considered to correspond when their type and anatomical position labels match. If such corresponding accepted annotations are found, it may be indicated to the user 430. In addition, a user- interface element may be generated 440, which allows the user to jointly accept the annotations of the current dental image that correspond to accepted annotations of another image. If these annotations are accepted they are listed in the text panel of the Ul and / or they may be indicated in the image with said bounding shapes and / or image masks. Furthermore, each accepted annotation may be linked to a record grouping corresponding annotations. Alternatively, the user may opt to individually inspect each of the annotations of the current image for whichcorresponding annotations were identified. This workflow is particularly useful for reviewing annotations comprising pathology findings.

[0036] Figure 7 illustrates the review and acceptance of an annotations comprising pathology findings according to the workflow involving the option to semi-automatically accept annotations for which corresponding accepted annotations were identified in one or more dental images of the image dataset for a given patient. When loading the dental image in the Ul, the computer automatically identifies that the mesial caries in tooth 12 (701) was previously observed and accepted in other dental images of the same patient and generates a timer associated Ul element 702 informing the user thereof. By clicking this Ul element 702, the user can accept the an notation for the current image (Fig. 7A, Fig. 7B). Further, the text panel indicates that images with a corresponding or relating finding are available 705. By selecting the relevant relating finding indication 705 in the text panel the dental images 703, 704 with corresponding annotations are presented in the Ul together with the current dental image (Fig. 7C).

[0037] Figure 8 illustrates an embodiment for the presentation of non-pathology findings. Upon loading a current dental image undefined annotations comprising non-pathology findings are automatically set to accepted. Further, the type and anatomical position label of these annotations are listed in text panel 800 but the corresponding findings are not indicated on the image representation in the Ul (Figure 8A). However, the user can visualise a non-pathology finding, such as the prosthetic crown at tooth position 24 (801). When selectingthis annotation in the text panel, a bounding shape (802) and image mask (803) is overlaid on the current dental image indicating the corresponding Al finding (Figure 8B). A selection of multiple non-pathology annotations can be visualized using the visibility button 804 generating visibility switches 805 next to each of the annotations in the text panel 800 (Figure 8C). The annotations of which the visibility switch was activated are indicated in the dental image by an overlaying image mask 806. Figure 8 further shows bounding shapes (807) overlaying image areas indicating the pathology findings of annotations that have not yet been accepted or rejected by the user (undefined annotations).

Claims

Claims1. A computer-based method for reviewing annotated dental images, the method comprising:• retrieving a dataset of dental images of a same patient with annotations, which have been generated based on the findings of one or more Al based algorithms,• wherein each annotation comprises a type label indicating the type of the Al finding, a boundingshape and / or image masks togetherwith image coordinates permitting to overlay said bounding shape and / or mask over the dental image region of the Al finding and wherein each annotation further comprises an anatomical position label and wherein each said annotation can be set to accepted, rejected or undefined;• displaying, via a graphical user interface, at a given instance, a current dental image from the dataset of medical images with an indication of the annotations for said current dental imageo wherein said user interface comprises a window for displaying said current dental image and a text panel wherein the type labels of the annotations can be listed togetherwith the anatomical position label, ando wherein each of the annotations is indicated in the user-interface by listing its type and anatomical position label in said text panel or by overlaying said bounding shape or mask over said current dental image;• receiving a user input to set an annotation to accepted or rejected.

2. The computer-based method according to claim 1 wherein said annotations either comprise• Al findings that are not related to observations that may require treatment, referred to as ‘non-pathology findings’, or• Al findings that are related to observations that may require treatment, referred to as ‘pathology findings’.

3. The computer-based method according to claims 1 or 2 wherein for a displayed current dental image said text panel only lists the type and anatomical position labels of accepted annotations.

4. The computer-based method according to claims 1 to 3 wherein the annotations comprising non-pathology findings of a current dental image are preset to accepted when displaying said current dental image in the user interface.

5. The computer-based method according to claims 1 to 4, wherein undefined annotations comprising pathology findings are indicated in the user interface by overlaying said bounding shapes over the displayed current dental image.

6. The computer-based method according to claim 5 wherein during hovering a said bounding shape of a said undefined annotation comprising a pathology finding, a said image mask is displayed overlaying the current dental image indicating the area of the Al findingwithin the boundingshape.

97. The computer-based method according to claim 6 wherein upon accepting an annotation comprising a pathology finding, the bounding shape of said annotation is deleted from the user interface and said image mask is displayed.

8. The computer-based method according to claim 7 wherein the type label and anatomical position label of an accepted annotation comprising a pathology finding are listed in said text panel.

9. The computer-based method according any of the preceding claims further comprising:• upon displaying a current dental image in the user interface, it is automatically verified whether any of the undefined annotations of said current dental image correspond to an accepted annotations of another dental image of said dataset of dental images of a same patient, o wherein annotations of separate dental images are considered to correspond when both their type label and anatomical position label are the same;• if for one or more annotations of the current image such corresponding accepted annotations are identified, indicate to the user that such accepted corresponding annotations were found.

10. The computer-based method according to claim 10 wherein, upon identification of accepted corresponding annotations for one or more undefined annotations of the current dental image, a user-interface element is generated allowing a user to jointly accept the current dental image annotations for which such corresponding accepted annotations were identified.

11. The computer-based method according to claims 1 to 3 wherein annotations comprising pathology findings of a current dental image are preset to accepted when displaying said current dental image in the user interface.

12. The method according to claim 11 wherein the user can select annotations comprising pathology findings of a current dental image and set the selected annotation to rejected.

Images