Systems and methods for providing real-time feedback during medical image delineation

Abstract

Systems and methods are disclosed for providing real-time feedback during image-based delineation of anatomical or pathological regions on medical images. Medical image data representing a volumetric anatomical region is presented via an interactive graphical user interface that enables a user to generate a contour. A plurality of clinician-generated contour datasets corresponding to the anatomical region is stored, and derived contour representations indicative of clinician agreement and variability are generated from the stored datasets. As the user generates or modifies the contour, spatial relationships between the user-generated contour and the derived contour representations are determined in real time. Based on the determined spatial relationships, one or more characteristics of a contouring tool or graphical indicator are dynamically modified to provide non-revealing feedback integrated into the graphical user interface. The disclosed systems and methods support image-based delineation across a range of medical imaging modalities and clinical workflows.

Description

Attorney Docket No. 24636.796WO1 SYSTEMS AND METHODS FOR PROVIDING REAL-TIME FEEDBACK DURING MEDICAL IMAGE DELINEATIONCross Reference to Related Applications

[0001] This patent application claims priority to, and thus the benefit of an earlier effective filing date from, U.S. Provisional Patent Application No. 63 / 790,886 (filed April 18, 2025), the contents of which are hereby incorporated by reference.Field

[0002] The present disclosure relates generally to computer-implemented systems and methods for medical image processing. More particularly, the disclosure relates to systems and methods for providing real-time, context-aware feedback during image-based delineation or contouring of anatomical or pathological regions on medical images, including within interactive graphical user interfaces used for clinical, diagnostic, planning, or training workflows.Background

[0003] Many areas of medicine rely on the accurate delineation of anatomical structures on medical images. In such workflows, a clinician reviews medical image data (e.g., computed tomography (CT) images, magnetic resonance imaging (MRI), dental images, pathology images, or other diagnostic or planning images) and manually defines boundaries corresponding to regions of clinical interest. These regions may include areas intended for treatment, analysis, or further evaluation, as well as regions that should be avoided or protected. The resulting delineations are used as inputs to downstream clinical, diagnostic, or planning processes, and inaccuracies in such delineations can adversely affect outcomes.

[0004] Image-based delineation tasks are commonly performed using computer-based contouring or annotation tools that present two-dimensional image slices assembled into a volumetric representation of an anatomical region. Clinicians typically generate contours by interacting with graphical tools to outline structures on individual image slices, with the system aggregating those outlines into a three-dimensional representation. While such tools are widely used, the task of accurately delineating anatomical boundaries remains complex, timeconsuming, and susceptible to error.Attorney Docket No. 24636.796WO1

[0005] A persistent challenge in medical image delineation is the inherent variability among qualified clinicians. Even when practitioners have comparable training and access to the same reference materials or guidelines, delineations of the same anatomical region may differ in extent, shape, or placement. In many clinical contexts, more than one delineation may be considered acceptable, reflecting differences in clinical judgment, interpretation of anatomical features, or balancing of competing considerations. As a result, there is often no single “correct” contour against which a user’ s work can be definitively evaluated.

[0006] Existing contouring and annotation systems typically provide limited support for addressing such variability. Feedback, if provided at all, is commonly delivered after a delineation has been completed and is often based on comparison to a single reference contour or guideline. This approach may obscure the range of acceptable practice and may fail to alert a user, in a timely manner, when a delineation deviates substantially from commonly observed patterns. Moreover, delayed or post-hoc feedback may allow errors or suboptimal decisions to propagate across multiple image slices before being identified.

[0007] Additionally, conventional systems generally do not provide context-aware feedback during the act of delineation itself. User interfaces are often static with respect to the delineation process, displaying contours without dynamically adapting visual cues based on how a user’s current input relates to previously observed delineations by other clinicians. As a result, users may lack real-time awareness of whether portions of a contour fall within regions of broad clinical agreement, within regions of known variability, or outside regions typically delineated by peers.

[0008] These limitations are not confined to a single medical specialty or imaging modality. Similar challenges arise in radiation therapy planning, dental and maxillofacial imaging, pathology image annotation, surgical planning, and other domains that involve manual delineation of anatomical or pathological features on medical images. Across these contexts, there remains a need for improved computer-based systems that support image-based delineation by providing timely, context-sensitive feedback that reflects patterns observed across multiple clinicians, without constraining users to a single reference solution.Attorney Docket No. 24636.796WO1Summary

[0009] The present disclosure provides computer-implemented systems and methods for supporting image-based delineation of anatomical or pathological regions on medical images through real-time, context-aware feedback. In various embodiments, the disclosed systems operate on medical image data representing a volumetric anatomical region and enable a user to generate contours or other delineations via an interactive graphical user interface.

[0010] In accordance with the disclosure, the system stores a plurality of cliniciangenerated contour datasets corresponding to a common anatomical region. From these datasets, the system generates one or more derived spatial representations, including a consensus contour representation indicative of regions of agreement among clinicians and a composite contour representation indicative of spatial variability across the clinician-generated datasets.

[0011] As a user generates or modifies a contour on the medical image data, the system determines spatial relationships between the user-generated contour and the derived contour representations. Based on the determined spatial relationships, the system dynamically modifies one or more visual characteristics of a contouring tool or graphical indicator presented within the graphical user interface. The visual modifications provide real-time, non-revealing feedback that reflects how portions of the user-generated contour relate to regions of clinician agreement and variability, without displaying reference contours or prescribing a single solution.

[0012] The disclosed systems and methods integrate the feedback directly into the contouring interface and operate contemporaneously with user interaction, thereby providing contextual awareness during the delineation process rather than after completion. The feedback may be presented in a manner that preserves user autonomy while conveying whether portions of a contour fall within regions commonly delineated by clinicians, within regions of known variability, or outside regions typically delineated.

[0013] Although described in the context of medical image contouring, the disclosed techniques are not limited to a particular medical specialty or imaging modality. The systems and methods may be applied to a wide range of image-based delineation tasks involving two-dimensional and three-dimensional medical images, including but not limited to radiationAttorney Docket No. 24636.796WO1 therapy planning, dental and maxillofacial imaging, pathology image annotation, surgical planning, and other clinical or diagnostic workflows that involve manual delineation of regions of interest.

[0014] Other illustrative embodiments (e.g., methods and computer-readable media relating to the foregoing embodiments) may be described below. The features, functions, and advantages that have been discussed can be achieved independently in various embodiments or may be combined in yet other embodiments, further details of which can be seen with reference to the following description and drawings.Brief Description of the Drawings

[0015] Some embodiments of the present disclosure are now described, by way of example only, and with reference to the accompanying drawings. The same reference number represents the same element or the same type of element on all drawings.

[0016] FIG. 1 is a block diagram of a contouring support system configured to provide real-time feedback during medical image delineation, in one exemplary embodiment.

[0017] FIG. 2 is a flowchart illustrating a process for providing real-time, context-aware feedback during medical image delineation, in one exemplary embodiment.

[0018] FIG. 3 is a flowchart illustrating a decision process for dynamically modifying at least one visual characteristic of a contouring tool based on determined spatial relationships between a user-generated contour and consensus and composite contour representations, in one exemplary embodiment.

[0019] FIG. 4 is a diagram of a graphical user interface during active contouring of a medical image slice, in one exemplary embodiment.

[0020] FIG. 5 is a diagram illustrating a contour refinement interface including magnified views of control point placement and a medical image view depicting a user-generated contour with discrete control points and boundary segments, in one exemplary embodiment.Attorney Docket No. 24636.796WO1

[0021] FIG. 6 is a diagram illustrating a contextual feedback interface presented in association with a medical image during contouring, including structured guidance and clinical consequence information, in one exemplary embodiment.

[0022] FIG. 7 is a conceptual representation illustrating relationships among a consensus contour, a composite contour, and a learner contour, along with associated expert notes, in one exemplary embodiment.

[0023] FIG. 8 is a block diagram of a computing environment operable to perform operations associated with the systems and methods disclosed herein, in one exemplary embodiment.Detailed Description

[0024] The figures and the following description depict specific illustrative embodiments of the disclosure. It will thus be appreciated that those skilled in the art will be able to devise various arrangements that, although not explicitly described or shown herein, embody the principles of the disclosure and are included within the scope of the disclosure. Furthermore, any examples described herein are intended to aid in understanding the principles of the disclosure, and are to be construed as being without limitation to such specifically recited examples and conditions. As a result, the disclosure is not limited to the specific embodiments or examples described below, but by the claims and their equivalents.

[0025] FIG. 1 illustrates an exemplary embodiment of a contouring support system 100 configured to provide real-time feedback during medical image delineation. In the illustrated scenario, a clinician 130 is actively generating a contour on medical image data displayed at a workstation 120. As the clinician 130 defines a user-generated contour on the displayed anatomical image, the system 100 retrieves previously stored clinician- generated contour datasets 110-1 through 110-N, generates derived spatial representations such as consensus and composite contour representations, determines spatial relationships between the user-generated contour and the derived representations, and dynamically modifies the graphical user interface to provide real-time, context-aware feedback. The figure depicts one example of how theseAttorney Docket No. 24636.796WO1 functional operations may be implemented within a distributed computing architecture; however, the architectural arrangement shown is illustrative and not limiting.

[0026] The system 100 includes the workstation 120 operated by the clinician 130. The workstation 120 includes a display device 122 configured to present medical image data representing a volumetric anatomical region. In the illustrated embodiment, the display device 122 presents at least one two-dimensional image slice from a volumetric dataset, such as a computed tomography image. The clinician 130 interacts with the display device 122 to define spatial boundaries corresponding to anatomical regions of interest. The workstation 120 may include one or more input devices such as a mouse, keyboard, stylus, touchscreen interface, haptic pen, trackball, or other user input mechanism configured to enable contour generation and modification.

[0027] The workstation 120 may include one or more processors and memory devices configured to execute instructions for presenting medical image data, receiving user-generated contour input, computing spatial relationships, and modifying graphical interface elements in response to those spatial relationships. Tn some embodiments, the workstation 120 performs at least a portion of the aggregation and comparison operations locally. In alternative embodiments, the workstation 120 functions as a client device that transmits contouring input and receives processed feedback from remote computing resources.

[0028] The system 100 further includes the network 104, which communicatively couples the workstation 120 to at least one server 102 and a data repository 106. The network 104 may comprise a local area network, wide area network, hospital intranet, secure cloud network, virtual private network, or combinations thereof. Communication over the network 104 may be wired or wireless and may utilize secure transmission protocols appropriate for medical data environments. In some embodiments, the server 102 and data repository 106 may be colocated with the workstation 120, and the network 104 may be internal to a single computing environment.

[0029] The server 102 is configured to perform computational operations associated with generating derived contour representations and evaluating spatial relationships. The serverAttorney Docket No. 24636.796WO1 102 may include one or more processors, memory devices, and storage media configured to execute software instructions for aggregating clinician-generated contour datasets 110, computing consensus contour representations indicative of spatial agreement, generating composite contour representations indicative of spatial variability, and calculating deviation metrics between user-generated contours and derived representations. In certain embodiments, the functionality attributed to the server 102 may be implemented within the workstation 120 or distributed across multiple computing nodes.

[0030] The data repository 106 stores a plurality of clinician-generated contour datasets 110-1, 110-2, through 110-N. The data repository 106 may store contour datasets 110 derived from multiple institutions, clinical specialties, or imaging modalities. The server 102 may retrieve subsets of contour datasets 110 based on contextual parameters associated with a current contouring session. Each contour dataset 110 may comprise spatial boundary information corresponding to delineations of a common anatomical region by respective clinicians. The contour datasets 110 may be stored as pixel masks, voxel segmentations, vector boundary coordinates, spline-based representations, polygonal meshes, or other spatial encodings. In certain embodiments, each contour dataset 110 is stored in association with structured case attributes such as anatomical region identifiers, imaging modality identifiers, clinical protocol indicators, institutional identifiers, or other contextual parameters enabling selective retrieval and aggregation of subsets of datasets.

[0031] In certain embodiments, each clinician-generated contour dataset 110 is stored as a structured data record comprising a contour identifier, a clinician identifier, an anatomical region identifier, an imaging modality identifier, a timestamp indicating when the contour was generated, and spatial boundary data. For volumetric contour datasets, the structured data record may include spatial boundary data for each of a plurality of image slices, enabling reconstruction of a three-dimensional contour surface.

[0032] In some embodiments, the structured data record further includes one or more quality indicators, such as a self-assessed confidence score provided by the contributing clinician, a peer review status, a time-to-completion metric, or a flag indicating whether the contour was generated under supervised or unsupervised conditions. These quality indicatorsAttorney Docket No. 24636.796WO1 may be used as weighting factors during generation of the consensus contour representation, such that contour datasets with higher quality indicators contribute more strongly to the consensus. In further embodiments, quality indicators may be combined with the structured case attributes described above to enable generation of consensus contour representations that are both context-specific and quality-weighted.

[0033] During operation, when the clinician 130 defines a user-generated contour on the display device 122, the workstation 120 transmits contour data to the server 102 via the network 104, or alternatively processes the contour locally. The server 102 retrieves relevant contour datasets 110 from the data repository 106 and generates derived spatial representations. A consensus contour representation may be computed as a region corresponding to at least a threshold level of overlap among selected clinician- generated contour datasets 110. A composite contour representation may be computed as a spatial union or envelope encompassing variability across the plurality of contour datasets 110.

[0034] The system 100 determines spatial relationships between the user-generated contour and the derived contour representations. Such spatial relationships may include boundary distances, overlap percentages, inclusion within defined regions, exclusion outside defined regions, or quantitative deviation measures. Based on the determined spatial relationships, the workstation 120 dynamically modifies one or more visual characteristics of the contouring interface presented on the display device 122. These visual characteristics may include color changes, line thickness adjustments, opacity variations, pattern overlays, shading effects, cursor modifications, or other graphical properties. Alternatively or additionally, the system 100 may provide haptic feedback through a stylus or other input device, or may generate auditory indicators corresponding to spatial deviation thresholds.

[0035] In certain embodiments, the system 100 further computes performance metrics representing deviation of the user-generated contour from the consensus contour representation. Such metrics may be stored in association with a user identifier corresponding to the clinician 130. The system 100 may compute additional metrics quantifying differences between deviation values obtained from multiple distinct contouring sessions associated with the clinician 130, thereby enabling temporal performance tracking across sessions. These computed metrics mayAttorney Docket No. 24636.796WO1 be stored within the data repository 106 or within a separate analytics database accessible via the network 104.

[0036] In certain embodiments, the system 100 maintains a longitudinal performance record for each clinician 130, comprising performance metrics computed across a plurality of contouring sessions conducted over an extended time period. The performance metrics may include spatial similarity measures such as a Dice similarity coefficient, a Hausdorff distance, or a mean surface distance between the user-generated contour and the consensus contour representation. The system 100 may compute trend indicators from the longitudinal performance record, such as a determination of whether deviation metrics have decreased below a predefined proficiency threshold over successive sessions. In some embodiments, the system 100 may identify specific anatomical subregions in which the clinician's performance consistently deviates from the consensus contour representation, enabling targeted training or review for those subregions.

[0037] Although FIG. 1 illustrates a single clinician 130 and workstation 120, the system 100 may support multiple concurrent users operating on multiple workstations. For example, multiple users may simultaneously contour the same anatomical region on separate workstations, each receiving independent real-time feedback based on the consensus contour representation and the composite contour representation. The system may optionally incorporate newly completed user-generated contours into the pool of clinician- generated contour datasets, dynamically updating the consensus contour representation and the composite contour representation as additional contours are submitted. This dynamic updating may cause the derived contour representations to evolve during a multi-user contouring session, such that users who complete their contours later receive feedback based on a consensus that reflects contributions from users who completed their contours earlier.

[0038] Accordingly, FIG. 1 depicts one representative embodiment of the system 100 for providing real-time, context-aware feedback during medical image delineation. The specific allocation of computational tasks among the workstation 120, server 102, and data repository 106, as well as the specific network configuration 104, may vary without departing from the scope of the disclosed system.Attorney Docket No. 24636.796WO1

[0039] FIG. 2 illustrates an exemplary process 200 for providing real-time, context-aware feedback during medical image delineation. The process 200 represents one implementation of operational actions performed by the system described in connection with FIG. 1. Although the process elements of the process 200 are illustrated in a particular order, the sequence shown is exemplary, and variations in ordering, subdivision, combination, repetition, or parallel execution of the process elements may be implemented while remaining within the scope of the disclosed subject matter.

[0040] The process 200 begins at the process element 202, in which a consensus contour representation and a composite contour representation are generated from a plurality of cliniciangenerated contour datasets. In the process element 202, the system accesses stored contour datasets corresponding to a common anatomical region. Each clinician-generated contour dataset defines spatial boundaries previously identified by a respective clinician on medical image data of the anatomical region. These boundaries may be represented as voxel masks, pixel masks, coordinate sets, spline curves, polygonal outlines, region-growing outputs, or other spatial encodings.

[0041] To generate the consensus contour representation at the process element 202, the system evaluates spatial overlap among the plurality of clinician- generated contour datasets. In some embodiments, the consensus contour representation corresponds to a region in which at least a predetermined threshold proportion of the clinician- generated contour datasets overlap. For example, for each spatial location within the anatomical region, the system may determine how many or what percentage of clinician- generated contour datasets include that location within their respective boundaries. Spatial locations satisfying or exceeding a defined overlap threshold may be incorporated into the consensus contour representation. In alternative embodiments, the consensus contour representation may be computed using weighted aggregation, wherein certain contour datasets are assigned greater influence based on predefined qualification criteria or contextual attributes. In further embodiments, the consensus contour representation may be derived from probabilistic agreement maps, voxel-wise agreement scores, boundary density analyses, or other spatial aggregation techniques.Attorney Docket No. 24636.796WO1

[0042] For example, in one embodiment the consensus contour representation may correspond to regions where at least 70% of the clinician-generated contour datasets overlap, though the threshold may be set at any other value. In certain embodiments, the threshold may be configurable by an administrator or may be automatically adjusted based on the number of clinician-generated contour datasets available for a given anatomical region. For instance, when fewer than five clinician-generated contour datasets are available, the system may apply a lower threshold to generate a meaningful consensus contour representation, whereas when twenty or more contour datasets are available, a higher threshold may be applied to produce a more selective consensus region. In further embodiments, different threshold values may be applied to different anatomical subregions within the same image slice or volumetric dataset, reflecting that certain anatomical boundaries may exhibit greater inter-clinician variability than others.

[0043] In one embodiment, the consensus contour representation may be generated using a boundary-averaging approach. For example, for each image slice, the system may parameterize each clinician-generated contour boundary as a function of angular position relative to a centroid of the contour and computes, for each angular position, a mean or median radial distance across the plurality of clinician-generated contour boundaries. The resulting mean or median boundary defines the consensus contour representation for that slice. The system may further compute a standard deviation or interquartile range of radial distances at each angular position and use these variability measures to define the composite contour representation as a band extending outward from the consensus contour boundary.

[0044] In a further embodiment, the system generates the consensus contour representation using a STAPLE (Simultaneous Truth and Performance Level Estimation) algorithm or a variant thereof. The STAPLE algorithm iteratively estimates a probabilistic ground truth segmentation and individual clinician performance parameters from the plurality of clinician-generated contour datasets. The resulting probabilistic ground truth, thresholded at a predefined confidence level, defines the consensus contour representation. Individual clinician performance parameters estimated by the STAPLE algorithm may additionally be stored in association with respective clinician identifiers and used to weight contributions of those clinicians in subsequent consensus generation operations.Attorney Docket No. 24636.796WO1

[0045] In addition to the consensus contour representation, the system generates the composite contour representation at the process element 202. The composite contour representation may define a spatial extent encompassing variability among the cliniciangenerated contour datasets. In one embodiment, the composite contour representation corresponds to a spatial union of all included contour datasets, such that any spatial location included within at least one clinician-generated contour dataset is incorporated within the composite contour representation. In alternative embodiments, the composite contour representation may correspond to an envelope boundary enclosing all clinician-generated contours, a region defined by a lower overlap threshold than that used for the consensus contour representation, or a variability band representing outer limits of clinician delineation patterns. The composite contour representation therefore captures the range of clinically observed delineation practices for the anatomical region.

[0046] In certain embodiments, the system may determine that an insufficient number of clinician-generated contour datasets are available for a given anatomical region or image slice to generate a reliable consensus contour representation. For example, if fewer than a predefined minimum number of datasets are available, the system may suppress real-time feedback for that region or slice, display a notification indicating that insufficient reference data is available, or fall back to comparison against a single reference contour rather than a consensus representation. In some embodiments, the system may detect outlier contour datasets within the plurality of clinician-generated contour datasets by identifying datasets whose spatial boundaries deviate from the aggregate of remaining datasets beyond a predefined outlier threshold. Identified outlier contour datasets may be excluded from generation of the consensus contour representation and the composite contour representation, or may be assigned a reduced weighting factor, to prevent anomalous delineations from distorting the derived representations.

[0047] Following generation of the consensus contour representation and the composite contour representation at the process element 202, the process 200 proceeds to the process element 204, in which user input defining a user-generated contour is received via an interactive graphical user interface. In the process element 204, a user views medical image data representing a volumetric anatomical region. The user defines spatial boundaries directly on the displayed image by interacting with a graphical contouring tool. The user-generated contour mayAttorney Docket No. 24636.796WO1 be defined incrementally, such as through sequential placement of control points, freehand drawing motions, brush-based segmentation actions, polygon tracing, or other interactive delineation mechanisms.

[0048] As the user defines the contour at the process element 204, the user-generated contour evolves dynamically. The contour may consist of connected boundary segments, interpolated spline curves, voxel selections, or region-based masks that collectively define a closed spatial boundary. The graphical user interface displays the evolving boundary in real time as it is created, extended, or modified.

[0049] At the process element 206, as the user-generated contour is being created, the system determines spatial relationships between the user-generated contour and the previously generated consensus contour representation and composite contour representation. This determination occurs contemporaneously with receipt of user input, rather than after the contour is finalized. The spatial relationships may include whether portions of the user-generated contour fall within the consensus contour representation, within the composite contour representation but outside the consensus contour representation, or outside the composite contour representation entirely. The system may further compute quantitative deviation measures, including distances between corresponding boundary segments, percentage overlap values, volumetric differences, inclusion ratios, or other geometric comparisons.

[0050] The determination at the process element 206 may be performed incrementally as each new segment of the user-generated contour is defined. Upon placement of each control point, completion of a boundary segment, or modification of a region, the system may evaluate the spatial location of that portion relative to the consensus contour representation and the composite contour representation. In some embodiments, deviation values may be computed locally for discrete portions of the contour, enabling localized feedback rather than only global feedback across the entire contour. The evaluation may occur continuously, periodically, or in response to specific user interactions.

[0051] The process 200 proceeds to the process element 208, in which at least one visual characteristic of a contouring tool displayed in the graphical user interface is dynamicallyAttorney Docket No. 24636.796WO1 modified based on the determined spatial relationships. The modification at the process element 208 is responsive to the real-time spatial evaluation performed at the process element 206. For example, the color of the contour boundary may change when the user-generated contour extends outside the consensus contour representation. In another embodiment, line thickness, opacity, pattern, shading, glow effects, cursor appearance, or overlay indicators may be altered to indicate whether a region of the contour falls within an area of high clinician agreement, within an area of known variability, or outside the composite contour representation.

[0052] The dynamic modification at the process element 208 may occur continuously as the contour is being drawn. Different segments of the contour may display different visual characteristics simultaneously, reflecting localized spatial relationships relative to the consensus contour representation and the composite contour representation. In some embodiments, alternative or additional non- visual cues may be provided, including haptic signals delivered through an input device or auditory indicators corresponding to predefined spatial deviation thresholds.

[0053] Although FIG. 2 illustrates a single pass through the process element 202, the process element 204, the process element 206, and the process element 208, the process 200 may be iterative. The system may repeatedly evaluate spatial relationships and dynamically modify visual characteristics as the user edits, refines, deletes, or redefines portions of the contour. The process elements may be subdivided into finer-grained operations, combined into integrated operations, reordered, or executed in parallel in alternative embodiments.

[0054] Accordingly, FIG. 2 depicts a representative process 200 in which derived spatial representations based on clinician-generated contour datasets are used to provide real-time, interface-integrated feedback during user contouring of medical image data. The illustrated process elements are exemplary and may be adapted to different image modalities, anatomical regions, contouring tools, and clinical workflows while maintaining the core functionality of real-time, context-aware spatial evaluation and dynamic interface modification.

[0055] FIG. 3 illustrates an exemplary decision process 208 associated with dynamically modifying at least one visual characteristic of a contouring tool based on determined spatialAttorney Docket No. 24636.796WO1 relationships. The decision process 208 represents a more detailed implementation of the process element 208 described in connection with FIG. 2. While the decision logic is illustrated as a sequence of discrete decision elements and outcome elements, the structure shown is exemplary. In alternative embodiments, the determinations may be performed in parallel, reordered, consolidated, or implemented using threshold comparisons, rule-based logic, lookup tables, state machines, or continuous evaluation functions rather than discrete branching.

[0056] The decision process 208 begins with an evaluation at the decision element 210. At the decision element 210, the system determines whether the user-generated contour is within the consensus contour representation. In certain embodiments, the determination at the decision element 210 may involve assessing whether a currently evaluated segment of the user-generated contour lies entirely within a spatial region defined by the consensus contour representation. In other embodiments, the determination may be based on whether a predefined percentage of the evaluated contour segment overlaps the consensus contour representation, whether the boundary distance between the user-generated contour and the consensus contour boundary is within a tolerance threshold, or whether deviation metrics fall below a predefined agreement threshold.

[0057] If the determination at the decision element 210 is affirmative, indicating that the contour or the evaluated portion thereof is within the consensus contour representation, the process proceeds to the outcome element 212. At the outcome element 212, the system changes the contouring tool to indicate consensus approval. The change at the outcome element 212 may include modifying a color of the contour boundary to a first predefined color, increasing opacity, applying a visual highlight, displaying a confirmation indicator, or otherwise altering a graphical characteristic of the contouring tool to convey that the evaluated portion of the contour aligns with regions of high clinician agreement. Alternatively or additionally, the system may modify a haptic characteristic of the contouring tool. For example, when the contouring tool comprises a stylus or haptic-enabled input device, the system may reduce resistance, provide a smooth tactile sensation, or generate a subtle confirmation vibration when the contour lies within the consensus contour representation. The visual and haptic changes at the outcome element 212 may be applied globally to the entire contour or locally to only the portion evaluated as being within the consensus contour representation.Attorney Docket No. 24636.796WO1

[0058] In certain embodiments, the contouring tool comprises a brush tool having a circular marker that follows the user's pointer. The brush tool defines contour boundaries based on the perimeter of the circular marker, such that the shape of the user-generated contour is determined by the outer edge of the brush as it is dragged across the medical image. However, the visual characteristic of the brush tool — such as the color of the circular marker — is determined based on the spatial location of the center of the circular marker relative to the consensus contour representation and the composite contour representation, rather than based on the perimeter. As a result, the feedback color displayed by the brush tool reflects the spatial zone in which the center point lies, while the contour boundary being drawn corresponds to the perimeter of the brush, which may lie in a different spatial zone than the center.

[0059] This separation between the point of feedback evaluation (the center) and the point of contour generation (the perimeter) is intentional. Because the feedback color corresponds to the center of the brush rather than to the boundary being drawn, the user cannot precisely determine the location of the consensus contour boundary or the composite contour boundary by observing the color transition. A user attempting to trace the exact boundary of the consensus contour representation would need to account for the offset between the brush center and the brush perimeter, which varies with brush size and prevents direct reverse-engineering of the underlying reference contour locations. This design thereby provides non-revealing feedback that guides the user's awareness toward regions of agreement and variability without disclosing the precise spatial boundaries of the derived contour representations, supporting the system's objective of preserving user autonomy and encouraging learning of the underlying anatomy rather than rote tracing of a reference contour.

[0060] If the determination at the decision element 210 is negative, the decision process 208 proceeds to the decision element 214. At the decision element 214, the system determines whether the user-generated contour is outside the consensus contour representation and within the composite contour representation. In this context, the determination at the decision element 214 evaluates whether the contour falls within the broader region capturing observed variability among clinician-generated contour datasets but does not satisfy the stricter agreement criteria of the consensus contour representation. The determination at the decision element 214 may involve computing spatial inclusion within the composite contour representation whileAttorney Docket No. 24636.796WO1 simultaneously confirming non-inclusion within the consensus contour representation.Quantitative thresholds, boundary distance calculations, or overlap percentages may be used to support this determination.

[0061] If the determination at the decision element 214 is affirmative, indicating that the contour lies within the composite contour representation but outside the consensus contour representation, the process proceeds to the outcome element 216. At the outcome element 216, the system changes the contouring tool to indicate composite approval. The change at the outcome element 216 may involve applying a second predefined visual characteristic distinct from that used at the outcome element 212. For example, the contour boundary may be displayed using a different color, dashed line pattern, reduced opacity, or alternate shading effect to indicate that the contour falls within an acceptable but variable region. Alternatively or additionally, the system may generate a corresponding haptic response distinct from that provided at the outcome element 212. For instance, the stylus may provide a mild tactile pulse, a different vibration pattern, or a slightly increased resistance compared to the consensus approval state, thereby conveying that the contour is acceptable but exhibits deviation from areas of high agreement.

[0062] If the determination at the decision element 214 is negative, the decision process 208 proceeds to the decision element 218. At the decision element 218, the system determines whether the user-generated contour is outside the composite contour representation. In certain embodiments, the determination at the decision element 218 may be satisfied when the evaluated contour segment lies beyond the spatial envelope defined by the composite contour representation, or when deviation metrics exceed a predefined variability threshold. This determination reflects that the contour falls outside the range of spatial delineations observed across the stored clinician-generated contour datasets.

[0063] If the determination at the decision element 218 is affirmative, the process proceeds to the outcome element 219. At the outcome element 219, the system changes the contouring tool to indicate disapproval. The change at the outcome element 219 may include modifying the contour boundary to a third predefined color, increasing line thickness, applying a flashing or blinking effect, introducing a warning icon, or otherwise altering a visualAttorney Docket No. 24636.796WO1 characteristic to signal that the evaluated contour portion lies outside the composite contour representation. Alternatively or additionally, the system may generate a more pronounced haptic response, such as increased resistance in the stylus movement, a stronger vibration, a repeated pulse pattern, or a brief tactile interruption intended to alert the user to significant deviation. The visual and haptic indications at the outcome element 219 may be more pronounced than those associated with the outcome element 212 and the outcome element 216 in order to convey a higher degree of deviation.

[0064] In certain embodiments, if the determination at the decision element 218 is negative, the system may default to maintaining a prior visual and haptic state or may apply an intermediate characteristic representing partial overlap conditions. Although not explicitly illustrated in FIG. 3, the system may support additional decision branches addressing partial segment overlap, volumetric thresholds, slice-by- slice evaluation, weighted clinician subsets, or graded deviation categories.

[0065] The decision process 208 may operate continuously as the user-generated contour is being defined. Rather than applying the logic only after completion of a full contour, the determinations at the decision element 210, the decision element 214, and the decision element 218 may be executed incrementally for newly added contour segments or modified regions. As a result, different portions of a single contour may simultaneously exhibit different visual and haptic characteristics corresponding to the outcome element 212, the outcome element 216, or the outcome element 219. The dynamic modification may be updated in real time as spatial relationships change during editing, deletion, refinement, or extension of the contour.

[0066] Although FIG. 3 depicts three discrete approval states — consensus approval, composite approval, and disapproval — the decision process 208 may be generalized to support additional graded levels of agreement. For example, multiple consensus thresholds may be defined, resulting in graduated visual and haptic states corresponding to high, moderate, or low clinician agreement. Alternatively, continuous deviation values may be mapped to a gradient of visual characteristics and tactile responses without discrete branching.Attorney Docket No. 24636.796WO1

[0067] Accordingly, FIG. 3 illustrates one embodiment of the decision process 208 for translating determined spatial relationships into dynamic interface modifications. The specific logical structure and number of decision elements and outcome elements may vary in alternative embodiments while preserving the core functionality of conditionally modifying contouring tool characteristics based on real-time spatial evaluation relative to the consensus contour representation and the composite contour representation.

[0068] In some embodiments, in addition to or as an alternative to dynamically modifying the visual characteristics of the contouring tool during active drawing, the system may generate color-coded indicator dots placed on a completed or partially completed user-generated contour. Each indicator dot may be positioned at a selected location along the contour boundary and colored according to whether the underlying region falls within the consensus contour representation, within the composite contour representation but outside the consensus contour representation, or outside the composite contour representation. To determine the number and placement of indicator dots, the system may calculate the number of distinct segments within the user-generated contour, multiply the number of segments by a scaling factor such as 0.7, and upper-bound the resulting value by a maximum number of dots such as ten. The system may then randomly select from among the segments to determine which segments receive indicator dots. In this manner, regions of the contour that have been more densely refined by the user — resulting in more segments — are likely to receive more indicator dots, meaning the user receives proportionally more feedback on more complex and clinically significant regions of the contour. The indicator dots may provide a post-placement summary of contour accuracy that complements the real-time brush-based feedback provided during active drawing.

[0069] FIG. 4 illustrates an exemplary graphical user interface 300 during active contouring of a medical image slice 304. The graphical user interface 300 presents a cross-sectional anatomical image corresponding to a selected slice within a volumetric dataset. In the illustrated embodiment, the medical image slice 304 represents an axial image of an anatomical region, such as a supraclavicular region, displayed for purposes of delineating a clinical target volume.Attorney Docket No. 24636.796WO1

[0070] Within the graphical user interface 300, the medical image slice 304 is centrally displayed to permit detailed visualization of anatomical structures. The user is prompted, via an instruction panel 302, to contour a specified anatomical region. The instruction panel 302 includes textual guidance identifying the anatomical target and specifying a slice range over which contouring is to occur. In the illustrated embodiment, the instruction panel 302 indicates that contouring is to be performed from a first slice to a second slice, thereby guiding the user to apply consistent delineation across multiple image slices within the volumetric dataset.

[0071] The instruction panel 302 further includes selectable contouring tools, such as a brush tool, a pen tool, a delete tool, and submission controls. These tools allow the user to define and refine a user-generated contour directly on the medical image slice 304. When the brush tool or pen tool is active, the user may draw a boundary along perceived anatomical borders. The boundary may be defined incrementally, such as by tracing along visible tissue interfaces, vascular structures, or bony landmarks visible within the medical image slice 304.

[0072] As the user interacts with the graphical user interface 300, a user-generated contour is created on the medical image slice 304. The contour may appear as a closed boundary surrounding a region of interest corresponding to the clinical target. The contour may be represented as a continuous line, a series of connected points, or a filled mask region depending on the selected contouring mode. The graphical user interface 300 may display the evolving contour in real time as the user draws or adjusts boundary segments.

[0073] In certain embodiments, while the user-generated contour is being defined on the medical image slice 304, the system simultaneously evaluates the spatial relationship between the user-generated contour and previously generated consensus and composite contour representations associated with the anatomical region. Although those derived representations are not explicitly shown in FIG. 4, the graphical user interface 300 may dynamically modify the visual characteristics of the contour being drawn to reflect the evaluated spatial relationship. For example, portions of the contour that fall within the consensus contour representation may be displayed using a first visual characteristic, while portions that fall outside the consensus contour representation but within the composite contour representation may be displayed using a secondAttorney Docket No. 24636.796WO1 visual characteristic. Portions extending outside the composite contour representation may be displayed using a third visual characteristic indicative of greater deviation.

[0074] The graphical user interface 300 may support slice navigation controls, such as a scroll bar or slider, enabling the user to move between adjacent image slices within the defined slice range. As the user navigates from one slice to another, the user-generated contour may be updated or newly defined for each slice. The system may evaluate spatial relationships independently on a slice-by- slice basis or may consider volumetric continuity across multiple slices when determining deviation metrics.

[0075] The instruction panel 302 may further display contextual information, such as the current slice number, tool status indicators, or progress metrics. In some embodiments, the instruction panel 302 may display deviation-related feedback, such as an approval indicator, a warning message, or a visual cue corresponding to the real-time evaluation logic described in connection with FIG. 3. Alternatively or additionally, the contouring tool selected within the instruction panel 302 may exhibit modified visual characteristics directly within the drawing region of the medical image slice 304.

[0076] The graphical user interface 300 may also incorporate alternative or additional non-visual feedback mechanisms during contouring. For example, if the contouring tool comprises a haptic -enabled stylus, the system may adjust tactile feedback based on whether the user-generated contour aligns with the consensus contour representation, falls within the composite contour representation, or extends outside the composite contour representation. Increased resistance, vibration patterns, or other tactile signals may be provided to convey deviation without requiring additional on-screen text.

[0077] Accordingly, FIG. 4 depicts an active contouring session in which a user is delineating a target region on the medical image slice 304 using tools provided within the graphical user interface 300. The graphical user interface 300 integrates instructional guidance, tool selection, and dynamic contour visualization to support real-time evaluation of the usergenerated contour relative to previously aggregated clinician-generated contour datasets, thereby enabling context-aware feedback during the contouring process.Attorney Docket No. 24636.796WO1

[0078] FIG. 5 illustrates an exemplary contour refinement interface 320 and associated contour manipulation features within the graphical user interface. The figure demonstrates how a user-generated contour may be created, adjusted, and locally modified on a medical image slice, and how discrete control points and boundary segments are manipulated during the contouring process.

[0079] In the upper portion of FIG. 5, a plurality of magnified views 320, 324, and 328 are shown, each corresponding to localized regions of a medical image in which contour editing operations are being performed. These magnified views illustrate how the user may interact with anatomical boundaries to define a contour around a target region.

[0080] The magnified view 320 shows a first anatomical region 322 in which a contour point labeled “a” is positioned relative to surrounding anatomical structures. In this view, the user may place, move, or adjust a control point along a perceived anatomical boundary. The control point may represent a vertex of a polygonal contour, a control handle of a spline-based curve, or another manipulable boundary element. The placement of the control point influences the shape of adjacent boundary segments and determines how the contour conforms to underlying anatomical features.

[0081] The magnified view 324 shows a second anatomical region 326 in which a control point labeled “b” is positioned near a structural interface. In this region, the user may refine the contour to more closely track subtle anatomical contours. For example, the user may drag the control point to adjust curvature, modify the slope of a boundary segment, or reposition the boundary to better align with visible tissue gradients. The system may interpolate between adjacent control points to maintain a smooth boundary transition.

[0082] The magnified view 328 shows a third anatomical region 330 in which a control point labeled “c” is being manipulated, for example using a stylus or cursor input. In this view, the contouring tool may be used to reposition the control point relative to a high-contrast anatomical feature. The system may provide visual feedback as the control point is moved, such as dynamically redrawing the connected boundary segments or highlighting affected portions of the contour.1Attorney Docket No. 24636.796WO1

[0083] In the lower portion of FIG. 5, a larger medical image view 332 displays a complete user-generated contour 334 surrounding a region of interest. The contour 334 is illustrated as a boundary composed of a plurality of discrete control points 336 connected by boundary segments. The control points 336 may be represented as selectable nodes distributed along the contour 334. Each control point 336 may be independently adjustable, allowing the user to locally refine the contour shape without redefining the entire boundary.

[0084] The boundary segments connecting adjacent control points 336 may be straight-line segments, spline-interpolated curves, Bezier segments, or other curve representations depending on the contouring mode. When a control point 336 is repositioned, the adjacent boundary segments are recalculated to maintain continuity and smoothness. In some embodiments, the system may apply smoothing constraints, curvature limits, or snapping behavior to encourage anatomically plausible contours.

[0085] The figure illustrates that the contouring process is not limited to initial boundary tracing but includes iterative refinement. The user may insert additional control points, delete existing control points, or adjust spacing between control points 336 to increase or decrease boundary resolution in selected regions. In areas with complex anatomical geometry, the user may increase the density of control points to achieve finer contour conformity. In smoother regions, fewer control points may be used to maintain efficiency.

[0086] In certain embodiments, while the user manipulates the control points 336 and boundary segments of the contour 334, the system may simultaneously evaluate spatial relationships between the contour 334 and derived consensus and composite contour representations. Although not explicitly depicted in FIG. 5, portions of the contour 334 may dynamically change visual characteristics to indicate levels of agreement or deviation relative to stored clinician-generated contour datasets. For example, specific boundary segments or control points 336 may be highlighted if they fall outside a composite contour representation, thereby guiding targeted refinement.

[0087] The contour 334 may also support slice-to- slice propagation within a volumetric dataset. Adjustments made in the displayed slice 332 may be interpolated to adjacent slices orAttorney Docket No. 24636.796WO1 may inform initialization of contours in neighboring image planes. In certain embodiments, the system may maintain volumetric continuity constraints such that movement of a control point 336 in one slice influences corresponding control points in adjacent slices.

[0088] In some embodiments, the user-generated contour may be initialized using an automated segmentation algorithm, such as a deep learning-based contouring model, prior to manual refinement by the user. The system may present an automatically generated initial contour on the medical image data, and the user may edit, adjust, or override portions of the initial contour using the contouring tools described herein. The system may evaluate spatial relationships between the edited contour and the consensus contour representation and the composite contour representation in the same manner as for a fully manually generated contour, providing real-time feedback as the user modifies the auto-generated initial contour. In some embodiments, the clinician-generated contour datasets stored within the data repository may include contours generated by automated algorithms in addition to contours generated by human clinicians, and the system may store metadata indicating whether each contour dataset was generated by a human clinician, by an automated algorithm, or by a human clinician who refined an automatically generated initial contour. This metadata may be used as a structured case attribute to selectively include or exclude algorithmically generated contours from the consensus and composite contour generation process.

[0089] Accordingly, FIG. 5 depicts detailed aspects of contour creation and refinement, including placement and manipulation of control points 336. dynamic updating of boundary segments, and localized anatomical alignment within magnified regions 320, 324, and 328. The illustrated interface supports precise, iterative contour shaping while enabling real-time evaluation and feedback relative to aggregated clinician-derived contour representations.

[0090] FIG. 6 illustrates an exemplary contextual feedback interface 340 presented in association with a medical image 342 during contouring of an anatomical region. The figure demonstrates how the system may provide structured explanatory feedback and clinical context in response to determined spatial relationships between a user-generated contour and previously generated consensus and composite contour representations.Attorney Docket No. 24636.796WO1

[0091] In the illustrated embodiment, the medical image 342 corresponds to a magnified view of an anatomical region, such as a supraclavicular area, displayed within the graphical user interface. A user-generated contour is shown overlaid on the medical image 342, surrounding a region of interest. An arrow indicator within the medical image 342 visually highlights a portion of the anatomical region that may be under-contoured or excluded from the current usergenerated contour. The arrow may be dynamically positioned by the system to direct the user’s attention toward a region that falls outside the consensus contour representation or otherwise deviates from stored clinician- generated contour datasets.

[0092] The contextual feedback interface 340 is displayed adjacent to the medical image 342 and includes structured textual guidance. In the illustrated embodiment, the contextual feedback interface 340 presents a message indicating that the clinical target volume should include the supraclavicular area. The message may be generated in response to a determination that a portion of the user-generated contour lies outside the consensus contour representation while falling within the composite contour representation or outside both representations. The system may determine that the user-generated contour does not adequately encompass a region commonly delineated by other clinicians and may therefore generate targeted instructional feedback.

[0093] The contextual feedback interface 340 may include selectable controls, such as a dismissal control and an informational control. The dismissal control allows the user to close or suppress the contextual feedback interface 340 without modifying the current contour. The informational control may provide access to additional explanatory material, such as anatomical references, contouring guidelines, consensus statistics, or supporting documentation. In certain embodiments, activation of the informational control may display additional images, textual explanations, or interactive overlays illustrating how clinician-generated contour datasets define the relevant anatomical region.

[0094] Below the contextual feedback interface 340, the figure includes a clinical consequence description 344. The clinical consequence description 344 provides explanatory text describing potential outcomes associated with under-contouring of the anatomical region. In the illustrated embodiment, the clinical consequence description 344 explains that failure toAttorney Docket No. 24636.796WO1 adequately cover supraclavicular lymph nodes may lead to long-term recurrence and adverse outcomes. The clinical consequence description 344 may be generated based on stored explanatory content associated with the identified anatomical region and may be conditionally presented when the system determines that the user-generated contour deviates beyond a predefined threshold.

[0095] The contextual feedback interface 340 may be triggered dynamically during contour creation or modification. For example, when the system detects that a deviation exceeds a predefined threshold associated with clinically significant under-coverage, the contextual feedback interface 340 may automatically appear. In alternative embodiments, the contextual feedback interface 340 may be displayed only after completion of a contour, upon submission of a contour, or in response to selection of a review mode.

[0096] In certain embodiments, the contextual feedback interface 340 may be integrated with the visual modification logic described in connection with FIG. 3. For example, if a portion of the user-generated contour transitions from a composite approval state to a disapproval state, the system may both modify the contouring tool characteristics and present the contextual feedback interface 340. The medical image 342 and the contextual feedback interface 340 may therefore operate cooperatively to provide layered feedback, combining graphical, textual, and contextual guidance.

[0097] In some embodiments, the contextual feedback interface 340 may further incorporate alternative or additional non- visual cues. For example, when the contextual feedback interface 340 is displayed, the system may generate an auditory notification or a haptic signal through a contouring input device to draw the user’s attention to the identified deviation. The intensity or pattern of the non-visual cue may correspond to the severity of the detected deviation.

[0098] The medical image 342 may remain interactive while the contextual feedback interface 340 is displayed. The user may continue adjusting the contour in real time, and the arrow indicator within the medical image 342 may reposition dynamically as the contour is modified. When the system determines that the user-generated contour has been adjusted to fallAttorney Docket No. 24636.796WO1 within the consensus contour representation or within an acceptable threshold of the composite contour representation, the contextual feedback interface 340 and the clinical consequence description 344 may automatically dismiss or transition to a confirmation state.

[0099] Accordingly, FIG. 6 illustrates an embodiment in which real-time spatial evaluation of a user-generated contour is supplemented with structured contextual guidance and clinical consequence information. The integration of the medical image 342, the contextual feedback interface 340, and the clinical consequence description 344 enables the system to provide not only spatial deviation feedback but also explanatory and clinically relevant information associated with identified contouring deviations.

[0100] FIG. 7 illustrates a conceptual representation 400 of relationships among a consensus contour, a composite contour, and a learner contour, as well as associated contextual information derived from clinician-generated contour datasets. The figure is intended to visually demonstrate how agreement and variability among multiple clinician-generated contour datasets may be represented and compared to a user-generated contour during evaluation.

[0101] In the upper portion of the conceptual representation 400, multiple overlapping contour outlines are shown superimposed over a generalized anatomical shape. These overlapping outlines represent individual clinician- generated contour datasets corresponding to the same anatomical region. Each clinician-generated contour may define a slightly different spatial boundary, reflecting differences in interpretation, technique, or clinical judgment. The overlapping region common to a threshold number of clinician-generated contour datasets corresponds to the consensus contour. The consensus contour is depicted as a centrally positioned dashed region representing areas of high spatial agreement among clinicians.Surrounding the consensus contour is a broader boundary representing the composite contour. The composite contour corresponds to the spatial union or envelope encompassing the full extent of variability observed across the clinician-generated contour datasets.

[0102] The middle portion of the conceptual representation 400 shows a learner contour superimposed relative to both the consensus contour and the composite contour. The learner contour represents a user-generated contour produced during a contouring session. In the illustrated example, portions of the learner contour overlap with the consensus contour, portions fall within the composite contour but outside the consensus contour, and portions extend outsideAttorney Docket No. 24636.796WO1 the composite contour entirely. These relative positions demonstrate how different segments of the learner contour may correspond to different levels of agreement or deviation.

[0103] The composite contour is illustrated as a boundary encompassing all cliniciangenerated contours, while the consensus contour is illustrated as a more restricted region of concentrated overlap. The learner contour is shown using a distinct boundary style to differentiate it from the consensus and composite contours. In this way, the figure demonstrates that the learner contour may be evaluated relative to two distinct spatial benchmarks: a high-agreement region represented by the consensus contour and a broader variability region represented by the composite contour.

[0104] In the lower portion of the conceptual representation 400, a simplified anatomical shape is again shown with the consensus contour and composite contour superimposed. An arrow is depicted pointing toward a region within the consensus contour, illustrating how the system may direct attention to a region of agreement or to a region where deviation has occurred. Adjacent to this depiction is an expert notes panel labeled as an expert notes region. The expert notes region may store explanatory annotations, guidance, or contextual commentary associated with specific anatomical subregions. In certain embodiments, when a learner contour deviates from the consensus contour beyond a predefined threshold, the system may retrieve and display relevant expert notes corresponding to the affected subregion.

[0105] A legend positioned at the bottom of the figure indicates the visual differentiation among the consensus contour, the composite contour, and the learner contour. The consensus contour may be represented using a first line style, such as a dashed boundary. The composite contour may be represented using a second line style, such as a solid or heavier boundary. The learner contour may be represented using a third line style distinct from the consensus and composite contours. These visual distinctions enable simultaneous display and comparison of the different contour representations within a single graphical context.

[0106] The conceptual representation 400 illustrates that the consensus contour may be entirely contained within the composite contour, while the composite contour encompasses the full range of clinician-derived spatial boundaries. The learner contour may intersect, overlap, or extend beyond these reference contours in varying ways. The figure therefore demonstrates the underlying spatial logic used by the system to classify contour segments as consensus-aligned, variability-aligned, or outside the composite range.Attorney Docket No. 24636.796WO1

[0107] In certain embodiments, the conceptual relationships shown in the conceptual representation 400 may be implemented volumetrically rather than in a single two-dimensional slice. The consensus contour and composite contour may be generated across multiple slices and evaluated in three-dimensional space. The learner contour may similarly be evaluated volumetrically. with deviation metrics computed across the entire anatomical volume.

[0108] Accordingly, FIG. 7 provides a high-level conceptual visualization of how clinician- generated contour datasets are aggregated into a consensus contour and a composite contour, how a learner contour is compared against those aggregated representations, and how associated contextual information, such as expert notes, may be integrated into the contour evaluation framework.

[0109] The following example illustrates one representative use case of the system 100. A radiation oncology resident is tasked with contouring a supraclavicular clinical target volume on a planning CT scan. The system retrieves fifteen clinician- generated contour datasets for the supraclavicular region from the data repository 106 and generates a consensus contour representation corresponding to the spatial region where at least a threshold proportion of the contour datasets overlap, and a composite contour representation corresponding to the spatial union of all fifteen contour datasets. The resident begins contouring on a first image slice using the pen tool. As the resident traces the medial boundary of the target volume, the contour boundary is displayed using a first visual characteristic indicating that the traced boundary falls within the consensus contour representation. The resident continues tracing laterally, and the contour boundary transitions to a second visual characteristic indicating that the boundary has moved outside the consensus contour representation but remains within the composite contour representation. On a subsequent image slice, the resident omits a posterior extension of the supraclavicular region commonly included by the contributing clinicians. The system detects that a portion of the consensus contour representation is not encompassed by the resident's contour and presents the contextual feedback interface 340 with a message indicating that the target volume should include the supraclavicular area. Upon completion of contouring across the specified slice range, the system computes a Dice similarity coefficient of 0.87 between the resident's contour and the consensus contour representation and a mean surface distance of 2.3 mm, and stores these performance metrics in association with the resident's user identifier. The system further retrieves the resident's performance metrics from two prior contouring sessionsAttorney Docket No. 24636.796WO1 for the same anatomical region, which yielded Dice coefficients of 0.79 and 0.83, and computes a positive trend indicating improvement across the three sessions.

[0110] FIG. 8 illustrates an exemplary computing environment 500 operable to perform one or more operations associated with the systems and methods disclosed herein. The computing environment 500 may include one or more computing systems 502-1 through 502-N, which may be implemented as local clinical workstations, institutional servers, remote server systems, or virtualized instances within a cloud-based infrastructure. Each computing system 502 may execute program instructions tangibly embodied on a computer-readable storage medium 514 to perform all or portions of the processes described with respect to the figures above.

[0111] Each computing system 502 may comprise at least one processor 504, program and data memory 506. one or more input / output (VO) devices 508, a display device interface 512, a network interface 510, and a computer-readable storage medium 514, all coupled by a system bus 516. The processor 504 may represent any suitable processing circuitry, including a central processing unit (CPU), graphics processing unit (GPU), digital signal processor (DSP), field-programmable gate array (FPGA), application- specific integrated circuit (ASIC), or combinations thereof. In some embodiments, multiple heterogeneous processors may cooperate to execute parallelized contour aggregation, spatial comparison, and graphical rendering tasks associated with medical image contouring operations.

[0112] The program and data memory 506 may store executable instructions and runtime data used by the processor 504. Such memory may include local cache, random-access memory (RAM), and bulk storage devices (e.g., solid-state drives or magnetic disks). The processor 504 may access portions of the memory 506 for temporary buffering of medical image slices, clinician- generated contour datasets, derived consensus and composite contour representations, user-generated contour data, and computed deviation metrics generated during real-time contouring operations.

[0113] The VO devices 508 may include peripherals such as keyboards, pointing devices, a stylus, a haptic-enabled pen, touchscreen interfaces, trackballs, microphones, or other input mechanisms configured to enable contour generation, modification, and refinement on displayed medical image data. The VO devices 508 may also include devices capable of generating nonvisual feedback, such as haptic-enabled input devices configured to provide tactile responses corresponding to evaluated spatial relationships between a user-generated contour and derivedAttorney Docket No. 24636.796WO1 contour representations. The display device interface 512 may drive integrated or remote display devices used to present medical image data, graphical user interfaces, contouring tools, dynamic visual feedback indicators, contextual guidance panels, and performance metrics. The network interface 510 provides wired or wireless connectivity — such as Ethernet, Wi-Fi, cellular, or optical links — to communicate with other computing systems 502 or with networked data repositories storing clinician-generated contour datasets.

[0114] The computer-readable storage medium 514 may comprise any non-transitory medium capable of storing program code and data, including semiconductor memory, optical media, or magnetic storage. Examples include ROM, RAM, flash memory, CD-ROM, DVD, and Blu-ray Disc. Instructions stored on the storage medium 514, when executed by the processor 504, may cause the computing system 502 to carry out one or more of the operations described herein — such as generating consensus and composite contour representations from stored clinician-generated contour datasets, receiving user input defining a user-generated contour via an interactive graphical user interface, determining spatial relationships between the usergenerated contour and the derived contour representations in real time, dynamically modifying visual or haptic characteristics of a contouring tool based on determined spatial relationships, computing deviation metrics, generating contextual feedback, and storing performance metrics in association with user identifiers.

[0115] In some embodiments, all or portions of these operations may be implemented within a data repository 520, which may include one or more storage components 522 and associated clinician-generated contour datasets 524-1 through 524-N. The data repository 520 may provide scalable storage infrastructure for large-scale aggregation of clinician- generated contour datasets across multiple institutions, clinical specialties, and imaging modalities.Contour datasets from local computing systems 502 may be uploaded to the data repository 520 for aggregation, consensus generation, and long-term archiving, and derived contour representations may be returned to client computing systems for real-time evaluation and feedback presentation. The data repository 520 may support on-demand scaling, resource pooling, and measured service typical of modem cloud architectures.

[0116] Any combination of the computing systems 502-1 through 502-N and the data repository 520 may cooperate to execute different stages of the disclosed methods. For instance, a local computing system 502-1 operating as a clinical workstation may perform initialAttorney Docket No. 24636.796WO1 acquisition of user-generated contour data and real-time feedback presentation, while the data repository 520 stores and aggregates clinician- generated contour datasets and generates consensus and composite contour representations for retrieval by the local computing system 502-1. In another configuration, a remote computing system 502-N may perform contour aggregation and spatial comparison operations, returning evaluation results to the local computing system 502-1 for rendering within the graphical user interface. In this manner, FIG. 8 represents a flexible computing framework in which hardware and software resources, whether local or distributed, operate collectively to perform the medical image contouring support functions disclosed herein.

[0117] Any of the embodiments disclosed herein may be rearranged, combined, or otherwise modified without departing from the scope of the invention. Accordingly, the inventive concepts described are not limited to any particular embodiment illustrated or discussed.

[0118] The computing or control elements shown in the figures and the storage devices, may be realized in dedicated hardware, programmable hardware executing software or firmware, or any suitable combination thereof. For example, a functional module may be implemented as a processor, controller, engine, or logic circuit executing instructions stored in memory, or as a dedicated hardware component such as a digital signal processor (DSP), application- specific integrated circuit (ASIC), or field-programmable gate array (FPGA).

[0119] When provided by a processor executing instructions, the described operations may be carried out by a single processor, a shared processor, or a plurality of distributed processors, some of which may be shared among subsystems. Explicit use of the terms processor, controller, or engine should not be construed as referring exclusively to hardware capable of executing software. Rather these terms encompass any physical computing element, whether dedicated or shared, that performs the described operations.

[0120] Software and firmware implementations may include firmware, microcode, resident software, virtualized components, or containerized services operating in orchestrated environments such as Kubernetes or equivalent systems. The instructions, when executed, cause the underlying hardware to perform the disclosed operations.

[0121] In one exemplary embodiment, instructions stored on a non-transitory computer-readable medium direct a computing system, such as a local server, cloud instance, or edge node,Attorney Docket No. 24636.796WO1 to perform the operations disclosed herein. The non-transitory computer-readable medium may include any magnetic, optical, or solid-state storage device or other physical medium capable of storing executable code.

[0122] Although the invention has been described with reference to certain exemplary embodiments, it will be understood that the scope of the invention is not limited to the specific implementations disclosed. Various modifications, substitutions, and alterations may be made by those skilled in the art without departing from the spirit and scope of the invention as defined by the appended claims.

[0123] All such variations, modifications, and alternative embodiments are intended to fall within the scope of the appended claims. Furthermore, the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting. Rather, the scope of the present invention should be defined solely by the claims and their equivalents.

Claims

Attorney Docket No. 24636.796WO1 CLAIMSWhat is claimed is:

1. A computer-implemented system for providing real-time feedback during medical image contouring, comprising:a processor;a database configured to store medical image data representing a volumetric anatomical region and a plurality of clinician-generated contour datasets corresponding to the anatomical region, each clinician- generated contour dataset defining spatial boundaries delineated by a respective clinician; anda memory storing instructions that, when executed by the processor, cause the system to: generate, from the plurality of clinician- generated contour datasets, (i) a consensus contour representation indicating regions of spatial agreement among the clinician-generated contour datasets, and (ii) a composite contour representation indicating a spatial extent encompassing variability among the clinician-generated contour datasets;receive user input defining a user-generated contour on the medical image data via an interactive graphical user interface;determine, as the user-generated contour is being created, spatial relationships between the user-generated contour and the consensus contour representation and the composite contour representation;and dynamically modify at least one characteristic of a contouring tool displayed in the graphical user interface based on the determined spatial relationships.

2. The system of claim 1, wherein the dynamic modification provides feedback indicative of whether portions of the user-generated contour fall:within the consensus contour representation;outside the consensus contour representation and within the composite contour representation; oroutside the composite contour representation.Attorney Docket No. 24636.796WO1 3. The system of claim 1, wherein the consensus contour representation is generated by determining a spatial region defined by at least a threshold level of overlap among the plurality of clinician-generated contour datasets.

4. The system of claim 1, wherein the composite contour representation comprises a spatial union of the plurality of clinician-generated contour datasets.

5. The system of claim 1, wherein dynamically modifying the at least one characteristic comprises altering at least one of: color; line thickness; opacity; pattern; orhaptic feedback associated with the contouring tool.

6. The system of claim 1, wherein determining the spatial relationships comprises calculating, in real time, at least one of: a distance between a boundary of the user-generated contour and a boundary of the consensus contour representation; a percentage overlap between the user-generated contour and the consensus contour representation; or a deviation of the usergenerated contour from the composite contour representation.

7. The system of claim 1, wherein the system updates a dynamic visual modification continuously as additional segments of the user-generated contour are received.

8. The system of claim 1, wherein the system is configured to suppress display of the clinician-generated contour datasets while providing the dynamic modification, thereby avoiding presentation of a reference contour to the user.

9. The system of claim 1, wherein each clinician-generated contour dataset is stored in association with structured case attributes, and wherein generation of the consensus contour representation is based on a subset of the plurality selected according to one or more of the structured case attributes.

10. The system of claim 1, wherein the medical image data comprises a plurality of two-dimensional image slices corresponding to a volumetric anatomical region, and wherein theAttorney Docket No. 24636.796WO1 spatial relationships are determined across successive slices as the user navigates through the volumetric anatomical region.

11. The system of claim 1, further comprising generating, based on the determined spatial relationships, a performance metric indicative of deviation of the user-generated contour from the consensus contour representation, and storing the performance metric in association with a user identifier.

12. The system of claim 11, wherein the system computes at least one performance metric quantifying a difference between deviation values obtained from at least two distinct contouring sessions associated with the user identifier.

13. The system of claim 1, wherein generation of the consensus contour representation is based on clinician-generated contour datasets associated with clinicians satisfying at least one predetermined qualification criterion.

14. A computer-implemented method for providing real-time feedback during medical image contouring, comprising:generating, by a processor, from a plurality of clinician- generated contour datasets stored in a database, each clinician-generated contour dataset defining spatial boundaries of a common anatomical region delineated by a respective clinician, (i) a consensus contour representation indicating regions of spatial agreement among the clinician-generated contour datasets, and (ii) a composite contour representation indicating a spatial extent encompassing variability among the clinician- generated contour datasets;receiving user input defining a user-generated contour on medical image data representing a volumetric anatomical region via an interactive graphical user interface;determining, as the user-generated contour is being created, spatial relationships between the user-generated contour and the consensus contour representation and the composite contour representation; anddynamically modifying at least one characteristic of a contouring tool displayed in the graphical user interface based on the determined spatial relationships.Attorney Docket No. 24636.796WO115. The method of claim 14, wherein the dynamic modification provides feedback indicative of whether portions of the user-generated contour fall:within the consensus contour representation;outside the consensus contour representation and within the composite contour representation; oroutside the composite contour representation.

16. The method of claim 14, wherein generating the consensus contour representation comprises determining a spatial region defined by at least a threshold level of overlap among the plurality of clinician-generated contour datasets.

17. The method of claim 14, wherein the composite contour representation comprises a spatial union of the plurality of clinician-generated contour datasets.

18. The method of claim 14, wherein dynamically modifying the at least one characteristic comprises altering at least one of: color; line thickness; opacity; pattern; or haptic feedback associated with the contouring tool.

19. The method of claim 14, wherein determining the spatial relationships comprises calculating, in real time, at least one of: a distance between a boundary of the user-generated contour and a boundary of the consensus contour representation; a percentage overlap between the user-generated contour and the consensus contour representation; or a deviation of the usergenerated contour from the composite contour representation.

20. The method of claim 14, further comprising updating the dynamic modification continuously as additional segments of the user-generated contour are received.

21. The method of claim 14, further comprising suppressing display of the cliniciangenerated contour datasets while providing the dynamic modification, thereby avoiding presentation of a reference contour to a user.Attorney Docket No. 24636.796WO122. The method of claim 14, wherein each clinician- generated contour dataset is stored in association with structured case attributes, and wherein generating the consensus contour representation is based on a subset of the plurality selected according to one or more of the structured case attributes.

23. The method of claim 14, wherein the medical image data comprises a plurality of two-dimensional image slices corresponding to the volumetric anatomical region, and wherein the spatial relationships are determined across successive slices as a user navigates through the volumetric anatomical region.

24. The method of claim 14, further comprising:generating, based on the determined spatial relationships, a performance metric indicative of deviation of the user-generated contour from the consensus contour representation; and storing the performance metric in association with a user identifier.

25. The method of claim 24, further comprising computing at least one performance metric quantifying a difference between deviation values obtained from at least two distinct contouring sessions associated with the user identifier.

26. The method of claim 14, wherein generating the consensus contour representation is based on clinician-generated contour datasets associated with clinicians satisfying at least one predetermined qualification criterion.

27. A non-transitory computer-readable storage medium storing instructions that, when executed by a processor, cause a computing system to:generate, from a plurality of clinician- generated contour datasets stored in a database, each clinician-generated contour dataset defining spatial boundaries of a common anatomical region delineated by a respective clinician, (i) a consensus contour representation indicating regions of spatial agreement among the clinician-generated contour datasets, and (ii) a compositeAttorney Docket No. 24636.796WO1 contour representation indicating a spatial extent encompassing variability among the cliniciangenerated contour datasets;receive user input defining a user-generated contour on medical image data representing a volumetric anatomical region via an interactive graphical user interface;determine, as the user-generated contour is being created, spatial relationships between the user-generated contour and the consensus contour representation and the composite contour representation; anddynamically modify at least one characteristic of a contouring tool displayed in the graphical user interface based on the determined spatial relationships.

28. The non-transitory computer-readable storage medium of claim 27. wherein the instructions further cause the computing system to provide feedback indicative of whether portions of the user-generated contour fall:within the consensus contour representation;outside the consensus contour representation and within the composite contour representation; oroutside the composite contour representation.

29. The non-transitory computer-readable storage medium of claim 27, wherein the instructions further cause the computing system to suppress display of the clinician-generated contour datasets while providing the dynamic modification, thereby avoiding presentation of a reference contour to a user.

30. The non-transitory computer-readable storage medium of claim 27, wherein the instructions further cause the computing system to generate, based on the determined spatial relationships, a performance metric indicative of deviation of the user-generated contour from the consensus contour representation, and store the performance metric in association with a user identifier.

Images