Computer-implemented method of prioritizing lumen contour adjustments in angiographic image data and corresponding apparatus

Abstract

The present application relates to a computer-implemented method and corresponding device for prioritizing lumen contour adjustments in angiographic image data to improve accuracy of coronary hemodynamic index calculations. The method comprises receiving angiographic image data, determining a lumen contour by automatic or semi-automatic segmentation, and identifying candidate portions to be edited based on an uncertainty measure or user input. For each candidate portion, an impact score is determined by analyzing a plurality of contour variations to quantify the impact of the adjustment on the calculated coronary hemodynamic index value. Candidates are ranked according to their impact score, and a prioritized list is presented to the user for review and potential modification through a graphical interface. The invention enables efficient correction of diagnosis-relevant regions while minimizing unnecessary adjustments.

Description

Technical Field

[0001] This invention relates to a computer-implemented method and system for processing angiographic image data in coronary artery assessment. Background Technology

[0002] US 2025 / 0 104 228 A1 describes a technique for adjusting or editing the contour of a segment of a given lumen of a coronary artery.

[0003] There is a need to improve the processing of angiographic image data in coronary artery assessment, particularly for lumen profile adjustment in coronary hemodynamic parameter calculation systems. Summary of the Invention

[0004] This need is met by a computer-implemented method and corresponding apparatus for prioritizing luminal contour adjustments in the calculation of coronary hemodynamic parameters. Other advantageous embodiments of the invention are the subject of this application.

[0005] According to the present invention, a computer-implemented method for prioritizing lumen contour adjustments in a coronary artery hemodynamic parameters calculation system includes the following steps:

[0006] - Receive angiographic image data representing the coronary vessels;

[0007] - Determine the lumen contour in the angiography image data based on automatic or semi-automatic segmentation;

[0008] - Candidate portions of the lumen profile for editing are determined based on an uncertainty measure associated with the segmentation or based on user input;

[0009] - An impact score for each candidate segment is determined by analyzing multiple changes in the luminal contour modulation, wherein the impact score represents the estimated impact of the modulation of the corresponding candidate segment on the calculated coronary hemodynamic parameters of the relevant changes in the luminal contour modulation;

[0010] - Sort the candidate parts based on their respective influence scores to generate a priority-ranked list of edit candidates; and

[0011] - The user is provided with the priority-sorted list via a graphical user interface for viewing and potential contour adjustments.

[0012] This method helps detect potential inaccuracies, prioritizes corrections based on their impact on coronary hemodynamic parameter calculations, and provides a systematic approach to focus on the most diagnostically relevant adjustments.

[0013] This method allows users to focus their corrections on the areas of the lumen profile most relevant to the diagnosis. Instead of requiring expert reviewers to manually examine and refine all uncertain areas, this more efficient method provides a systematic prioritization of edit candidates based on their potential impact on diagnostic results, particularly on calculated coronary hemodynamic parameters.

[0014] This invention provides a system for prioritizing the impact perception of luminal contour adjustments. By ranking candidate edit regions based on their estimated impact on calculated coronary hemodynamic parameters, the proposed method allows users to focus on the most critical adjustments first, thereby improving efficiency and diagnostic accuracy.

[0015] A first aspect of the invention relates to a computer-implemented method for prioritizing luminal contour adjustments in the calculation of coronary hemodynamic parameters. The method includes receiving angiographic image data representing coronary vessels. Based on this data, an automatic or semi-automatic segmentation process is used to determine the luminal contour. This contour represents the inner boundary of the vessel and serves as the basis for further analysis. To improve the accuracy of segmentation, candidate portions of the luminal contour are identified for potential editing. This selection is based on an uncertainty measure associated with the segmentation process or on user input indicating the area requiring adjustment. For each candidate portion, an influence score is determined, representing the estimated influence of modifying that portion on the calculated coronary hemodynamic parameter value. An influence score is obtained by analyzing multiple variations in luminal contour adjustment and quantifying how these variations affect the calculated coronary hemodynamic parameter value. After assigning influence scores, the method continues to prioritize candidate portions according to their respective influence scores. This results in a priority list for editing candidates, allowing the user to focus first on the most relevant adjustments. Finally, the priority list is presented to the user via a graphical user interface (GUI). The user can then review and modify the suggested contour adjustments, ensuring that their corrections effectively target the diagnostically most relevant areas.

[0016] A second aspect of the invention relates to an apparatus for prioritizing luminal contour modifications in the calculation of coronary hemodynamic parameters. The apparatus includes an input unit configured to receive angiographic image data representing coronary vessels, which serves as the basis for segmentation and further analysis. A processing unit performs the computational steps required to identify contour modifications and prioritize them. Specifically, the processing unit uses an automatic or semi-automatic segmentation method to determine the luminal contour in the angiographic image data and identifies candidate portions of the luminal contour that may require editing, based on an uncertainty measure associated with segmentation or based on user input. For each candidate portion, the apparatus calculates an influence score by analyzing multiple variations in the luminal contour modifications, wherein the influence score quantifies the estimated impact of modifying the corresponding candidate portion on the calculated coronary hemodynamic parameter values. Based on these scores, the candidate portions are ranked to generate a priority list of editable candidates. The apparatus also includes an output unit that provides the priority list to a user via a graphical user interface. This allows the user to view and refine the suggested contour modifications, ensuring that manual modifications are focused on the diagnostically most relevant areas.

[0017] A third aspect of the invention relates to a medical system for coronary artery assessment, the medical system comprising: an imaging unit configured to acquire angiographic image data representing coronary vessels; and an apparatus according to the second aspect.

[0018] The imaging unit serves as the system's data acquisition component, providing high-resolution angiographic images that are subsequently processed by the device of this invention. By integrating the device of this invention into a medical system, automatic prioritization of lumen contour adjustments can be seamlessly incorporated into existing diagnostic workflows. This ensures that key contour modifications are identified and presented to the user in a structured and efficient manner, thereby improving diagnostic accuracy and reducing manual work.

[0019] The medical system enables real-time or near-real-time analysis of coronary vascular structures and ensures adjusted prioritization is performed directly in conjunction with the image acquisition process. By embedding the prioritization framework into a larger medical context, this invention allows for optimized interaction between automated analysis and expert-based review, thereby further improving the efficiency of clinical decision-making.

[0020] Embodiments of the present invention relate to a technique for prioritizing the contour adjustment of segmented luminal structures based on uncertainty measures and their estimated impact on fractional flow reserve (FFR) calculations.

[0021] This invention is based on the finding that imaging techniques used in coronary artery assessment can introduce uncertainties in lumen segmentation. These uncertainties arise from factors such as image noise, motion artifacts, or contrast variations, leading to inaccuracies in the depiction of vascular structures. Since the segmentation results are used as the basis for further computational analyses, particularly the calculation of coronary hemodynamic parameters, including fractional flow reserve (FFR), this inaccuracy may necessitate manual post-processing to refine the lumen profile. However, manual refinement is time-consuming, resource-intensive, and may not always be necessary for clinically insignificant biases.

[0022] Given these challenges, it is desirable to focus manual corrections only on those segments that have a meaningful impact on diagnostic results. This invention addresses this problem by providing an automated system that assesses the uncertainty of lumen segmentation and estimates the potential benefit of correction at each identification location. Based on this assessment, the system generates prioritized recommendations for contour adjustment, directing expert attention to the most relevant areas.

[0023] The method of this invention enables the systematic ranking of contour modification candidates based on their likelihood of affecting the calculation of coronary hemodynamic parameters, thereby ensuring the efficient allocation of clinical resources. By integrating an impact-aware prioritization mechanism, this invention reduces the workload of expert users, minimizes unnecessary refinement, and improves diagnostic accuracy.

[0024] This invention is designed for use with angiographic image data representing coronary vessels. This image data is typically obtained via X-ray coronary angiography, an imaging modality widely used in interventional cardiology. In this technique, a contrast agent is injected into the coronary arteries, and X-ray images are rapidly and continuously acquired to visualize blood flow and vessel morphology.

[0025] In many cases, angiographic data is captured as single-plane or biplane X-ray fluorescence fluoroscopy sequences, providing a dynamic view of the coronary vascular system. The resulting image sequences can be processed to reconstruct 3D coronary artery models using stereo reconstruction or deep learning-based methods. However, the present invention is not limited to 3D processing and can also be applied to 2D angiographic analysis, where the luminal contour is directly segmented from individual frames.

[0026] While this invention primarily addresses angiographic image data, it can also be applied to other imaging modalities used for coronary artery assessment, such as intravascular ultrasound (IVUS) or optical coherence tomography (OCT). These alternative imaging techniques provide cross-sectional views of the vessel lumen and can complement angiographic analysis by refining segmentation accuracy. The system can process such data independently or in combination with angiographic images to improve contour recognition and prioritization.

[0027] Input image data can be acquired in real time (online processing) during the interventional procedure in the catheter insertion laboratory or retrieved from a previously recorded dataset (offline processing). Depending on available computing resources and clinical workflows, this invention can be integrated into different stages of the diagnostic process.

[0028] In the context of this invention, the term lumen profile refers to the boundary of the blood-filled space within a coronary artery, extracted from angiographic image data. This profile represents the internal vessel wall and is a key structure in computational analysis of coronary artery physiology, including fractional flow reserve estimation.

[0029] Lumen contours are typically determined using segmentation algorithms, which can be based on traditional image processing techniques (e.g., edge detection, intensity thresholding) or deep learning models trained on angiography datasets. Depending on the segmentation method and image quality, the extracted contours may exhibit local inaccuracies or variations, leading to uncertainty that necessitates further refinement.

[0030] It is important to note that the luminal profile used in this invention does not necessarily correspond to the midline representation of the blood vessel. Instead, it follows the detected inner wall boundary, capturing potential stenosis, irregularities, or variations in vessel morphology. The accuracy of this profile is critical for downstream calculations, as even small deviations in stenotic regions can lead to significant changes in calculated coronary hemodynamic parameters (e.g., FFR values).

[0031] This invention provides a mechanism for assessing uncertainties in the segmented luminal contour and prioritizing adjustments based on their clinical relevance. This ensures that manual corrections target areas where segmentation errors have the greatest impact on diagnostic outcomes.

[0032] In the context of this invention, the term uncertainty metric refers to a measure of the confidence or reliability of a segmented luminal contour in angiographic image data. Uncertainty in segmentation can arise due to factors such as image noise, motion artifacts, or limitations in the segmentation algorithm. Uncertainty metrics serve as a basis for identifying candidate portions of the luminal contour that may require refinement.

[0033] This invention eliminates the need for users to manually examine the entire vascular structure. Instead, it utilizes this method to pre-select and adjust the regions most likely to improve the accuracy of calculated coronary hemodynamic parameters. Uncertainty metrics allow the system to distinguish between high-confidence and low-confidence regions, ensuring that calibration efforts are focused on the most diagnostically relevant areas.

[0034] The term "candidate portion of luminal profile" refers to a specific region along a segmented luminal profile that has been identified as a potential target for manual or automated refinement. These portions represent areas where segmentation may be uncertain or inaccurate, and adjustments can improve the reliability of the calculated fractional flow reserve (FLR) values.

[0035] Candidate regions can be identified based on uncertainty metrics, providing a systematic approach for detecting regions with low segmentation confidence. Alternatively, they can be selected based on user input, allowing experts to manually label regions requiring correction.

[0036] Not all uncertain regions require modification. This invention not only considers the existence of uncertainty but also the potential diagnostic impact of correcting for a given region. Therefore, candidate portions are prioritized based on their estimated impact on the calculation of coronary hemodynamic parameters, ensuring that the user's efforts are focused on the most relevant adjustments.

[0037] By intelligently selecting candidate parts, the system reduces the burden of manual review, allowing experts to focus their attention on adjusting the areas most likely to improve diagnostic accuracy, while minimizing unnecessary modifications.

[0038] The term "impact score" refers to a numerical or categorical value that quantifies the expected impact of modifying candidate portions of the luminal contour on the calculated fractional flow reserve value. This score allows the system to prioritize contour adjustments based on their potential diagnostic relevance, ensuring that the user focuses their efforts on the most meaningful corrections.

[0039] Influence scores are determined by analyzing multiple variations in the contour of candidate regions, simulating potential adjustments, and evaluating changes in the calculated coronary hemodynamic index values. The magnitude of these variations provides an estimate of how strongly adjustments in this region affect the overall diagnostic outcome. In some cases, candidate regions may exhibit high segmentation uncertainty but have little effect on the calculated coronary hemodynamic index values, making their correction less relevant. Conversely, even small modifications in critical stenosis regions can significantly alter the calculation of coronary hemodynamic indices, resulting in high influence scores.

[0040] By incorporating impact scores into a prioritization framework, this invention ensures that the most impactful adjustments are recommended first. This targeted correction process improves workflow efficiency and diagnostic accuracy, reduces unnecessary refinement, and ensures that clinically significant modifications are made when necessary.

[0041] In embodiments of the present invention, the process of determining the factors affecting the score involves analyzing multiple changes in lumen profile regulation and quantifying the resulting changes in the calculated fractional flow reserve value.

[0042] This process systematically introduces controlled changes into candidate portions of the lumen profile and re-evaluates the corresponding coronary hemodynamic parameters. By observing the fluctuations in these calculated coronary hemodynamic parameters amidst these changes, the system estimates how sensitive the final diagnosis is to modifications in the selected region. The magnitude of these changes is directly correlated with the impact score, allowing the system to rank editing candidates based on their expected diagnostic relevance.

[0043] By incorporating structured analysis of this change, this implementation ensures that the prioritization process remains data-driven and objective, rather than relying solely on static uncertainty measurements. This approach enhances the reliability of the ranking system by focusing on the actual effects of profile modifications rather than purely theoretical uncertainty measures.

[0044] In embodiments of the invention, candidate portions of the lumen profile are identified based on an uncertainty metric, which serves as an indicator of segmentation reliability. This uncertainty metric can be derived using various methods. One possible approach involves multi-frame segmentation aggregation, where the lumen profile is extracted from multiple angiographic images, and variations between segmentation results are analyzed to estimate uncertainty. Another approach utilizes machine learning-based prediction, where a trained model estimates segmentation uncertainty based on pre-computed training data by identifying patterns indicating low-confidence regions. Furthermore, uncertainty can be inferred through local geometric feature analysis, where abrupt changes in lumen curvature or discontinuities in segmentation are detected as indicators of potential inaccuracy.

[0045] By employing one or more of these techniques, the present invention ensures that candidate portions are selected systematically rather than arbitrarily. This targeted selection process enables finer prioritization of editing candidates, allowing corrective efforts to be effectively allocated to adjusting the areas most likely to improve diagnostic accuracy.

[0046] In embodiments of the present invention, the determination of the influence score considers not only individual candidate segments in isolation, but also the interdependencies between multiple contour adjustments. In some cases, modifying a segment of the luminal contour can alter the hemodynamic properties of the entire vessel, thereby affecting the calculated flow reserve score more than the effect of a single local change. To address this issue, the system jointly evaluates multiple candidate segments when determining the influence score.

[0047] By combining this joint assessment, the system can capture the combined effects when multiple modifications interact within the computational model. For example, proximal narrowing may initially appear to have a limited effect on coronary hemodynamic parameters due to an underestimated distal stenosis. However, once the proximal segment is corrected, the hemodynamic impact of distal stenosis can increase significantly. Therefore, the impact score reflects the total effect of contour adjustments from the joint assessment, rather than treating each modification as an independent event.

[0048] This approach improves the accuracy of the prioritization framework by ensuring that critical regions are not overlooked due to isolated assessments. By taking into account interdependencies, the invention allows for more precise ranking of candidate components, directing correction efforts towards regions where cumulative modifications have the most significant impact on diagnostic results.

[0049] In embodiments of the present invention, the determination of the influence score is applied to various coronary hemodynamic parameters. These parameters provide different perspectives on coronary physiology and can be used to assess vascular function under different physiological conditions. Therefore, the present invention allows for the determination of an influence score for at least one coronary hemodynamic parameter selected from fractional flow reserve (FFR), instantaneous flow-free ratio (iFR), resting distal pressure to aortic pressure ratio (resting Pd / Pa), coronary flow reserve (CFR), and microvascular resistance index (IMR).

[0050] By incorporating multiple hemodynamic parameters, this invention extends its applicability beyond FFR-based assessments, where FFR is the most widely used coronary hemodynamic parameter. While FFR is widely used in clinical practice to assess the functional significance of coronary artery stenosis under congestive conditions, other parameters such as iFR and resting Pd / Pa provide similar insights without requiring pharmacologically induced vasodilation. On the other hand, CFR measures the ratio of maximum blood flow to resting blood flow, reflecting the total volume of the coronary circulation, while IMR provides an estimate of microvascular resistance. Depending on the clinical context, the system can be configured to prioritize luminal profile modulation based on the most relevant indices for a given patient case.

[0051] This flexibility ensures that prioritization remains applicable to different diagnostic methods, thereby allowing for optimized assessment of coronary physiology regardless of the specific hemodynamic parameters used.

[0052] In embodiments of the invention, the determination of the influence score is performed dynamically by applying multiple variations in the lumen profile and calculating the corresponding coronary hemodynamic parameters in real time. These variations represent potential profile adjustments and are used to assess the sensitivity of the coronary hemodynamic parameter calculations to modifications in the segmented lumen profile.

[0053] Variations can be generated in different ways. One approach involves random perturbations, where contour modifications are introduced within a range defined by an uncertainty metric. This allows for probabilistic estimation of the effects of small local adjustments. Alternatively, predefined systematic modifications, including controlled shifts, scaling, or smoothing of contour segments, can be applied to explore the effects of structured changes. Furthermore, variations can be algorithmically optimized, meaning they are generated using an optimization process that seeks to maximize the variance of the calculated coronary hemodynamic parameters, thus ensuring the capture of the most influential modifications.

[0054] For each variation, coronary hemodynamic parameters were calculated, and an impact score was derived based on the variance of the calculated coronary hemodynamic parameters across all generated contour modifiers. High variance indicates that modifications in a given region significantly affect the diagnostic outcome, while low variance indicates that adjustments in that region have minimal impact.

[0055] By continuously evaluating and updating impact scores in real time, this online approach enables adaptive and data-driven prioritization of contour correction. This ensures that the most diagnostically relevant areas are dynamically identified and prioritized, thereby improving efficiency and accuracy in the review process.

[0056] In embodiments of the invention, a pre-trained machine learning model is used to determine the impact score, which allows estimation of the effect of contour modifications without requiring real-time calculation of coronary hemodynamic parameters. This approach significantly reduces computational requirements while still providing a reliable ranking of contour adjustments based on their expected impact on diagnostic outcomes.

[0057] Unlike online variants that dynamically calculate the impact on scores by performing repeated coronary hemodynamic parameters for different contour variations, offline variants utilize machine learning models trained on large datasets of previously analyzed cases. These models can predict the expected variance in coronary hemodynamic parameter values ​​based on learned patterns, thus achieving an efficient and computationally inexpensive assessment of the impact on scores.

[0058] By using a pre-trained model, the system can provide instantaneous prioritization of contour adjustments, making it particularly suitable for applications where real-time computation would be too resource-intensive. Offline variants can operate independently or in combination with online variants, depending on available computational resources and the desired accuracy of the impact assessment. For example, an offline model can be used to generate an initial prioritization, which can then be refined using real-time computation in selected high-priority regions. This combination of offline and online processing ensures the system remains adaptable to diverse clinical settings, balancing computational efficiency with diagnostic accuracy.

[0059] In embodiments of the invention, the method further includes a training step, wherein a machine learning model is trained to predict the impact score of contour modifications. The training process ensures that the model can reliably estimate the impact of modulation on the calculated blood flow reserve score without requiring direct real-time computation.

[0060] To train the model, multiple training samples were provided, each containing luminal contours and relevant data associated with the prediction of influence scores. These training samples were drawn from real patient data, synthetic angiography data, or a combination of both, allowing the model to learn from diverse datasets. For each training sample, the influence score was calculated via direct analysis or assigned based on predefined criteria. The machine learning model was then trained using these labeled samples, enabling it to generalize and predict influence scores for novel, previously unseen luminal contours.

[0061] By incorporating this training process, the system ensures that the impact score estimates in offline variants remain robust and accurate. A well-trained model can provide instantaneous prioritization of contour adjustments, significantly reducing computational costs while maintaining high diagnostic reliability. Furthermore, the training method allows for continuous improvement, as new data can be integrated into the model to enhance its predictive accuracy over time.

[0062] In embodiments of the invention, the determination of influence scores is further refined by incorporating information from previously performed lumen profile adjustments. This adaptive process enables the system to learn from past corrections, thereby ensuring that future prioritization is more consistent with expert preferences and real-world correction patterns.

[0063] This adjustment is based on analysis of historical user corrections stored in a database and used to refine uncertainty metrics and impact score predictions. By evaluating how experts adjusted the lumen profile in previous cases, the system can identify recurring patterns during manual refinement and adjust their priorities accordingly. If certain types of adjustments consistently lead to significant changes in calculated coronary hemodynamic parameters, the system can assign higher impact scores to similar cases in the future. Conversely, if historical data indicates that certain uncertainties frequently cause small or negligible changes in coronary hemodynamic parameters, the model can downgrade those areas to avoid unnecessary corrections.

[0064] By integrating historical correction data, this invention enhances the adaptability of the prioritization system, allowing it to evolve over time based on real-world user interactions. This ensures that the sequencing of contour adjustments remains dynamic, thereby improving the efficiency and accuracy of clinical workflows.

[0065] In embodiments of the invention, the method further includes a classification step, wherein regions of the coronary artery lumen are assigned to categories based on coronary hemodynamic parameters. This classification is performed using one or more predefined thresholds, thereby allowing for an objective assessment of whether a given region is considered functionally significant.

[0066] The determination of the impact score for classification is based on whether the modification of the luminal profile causes a shift between classification categories. If the profile adjustment results in a shift from one diagnostic category to another (e.g., from non-significant stenosis to significant stenosis), this shift will be reflected in the adjusted impact score. In contrast, modifications that do not affect classification can be assigned a lower impact score because they are less likely to alter clinical decision-making.

[0067] A particularly relevant example is the use of predefined thresholds for coronary hemodynamic parameters, defining whether stenosis is classified as functionally significant. If a modification to the luminal profile results in a shift across such a threshold, the corresponding impact score is weighted more strongly than for cases where the coronary hemodynamic parameters remain within the same diagnostic category before and after the modification. This ensures that modifications with the potential to alter clinical conclusions take precedence over changes that, while numerically relevant, do not affect the classification outcome.

[0068] By incorporating this classification-based adjustment, the present invention ensures that prioritization depends not only on the absolute changes in hemodynamic values, but also on whether those changes affect diagnostic outcomes. This refinement further improves the efficiency of the prioritization framework by directing the correction work toward modifying areas with the greatest clinical relevance.

[0069] In an embodiment of the invention, the device further includes a storage unit and a processing unit, the storage unit and the processing unit being used to facilitate the training of a machine learning model for predicting the impact score of lumen profile adjustment.

[0070] The storage units are configured to store and provide training data, ensuring that the relevant datasets are accessible during the training process. These training datasets can include real patient data, synthetic angiography data, or a combination thereof, enabling the machine learning model to effectively generalize to different types of input data.

[0071] The processing unit is responsible for generating training samples, which include luminal contours and associated impact scores. It also uses direct coronary hemodynamic parameters or predefined criteria to calculate or assign impact scores to these training samples. Once the labeled training data is ready, the processing unit is used to train a machine learning model, allowing it to predict impact scores for new cases based on the learned patterns.

[0072] By combining dedicated storage units and processing units for model training, this invention ensures that the machine learning framework can be continuously improved and adapted to new data, thereby increasing the accuracy and efficiency of score prediction over time.

[0073] Extracting and segmenting the coronary artery lumen from angiographic images is a necessary step in computational assessments such as minimum lumen diameter measurement, stenosis severity estimation, and fractional flow reserve (FFR) calculation. However, medical imaging techniques inherently introduce uncertainties due to factors such as image noise, motion artifacts, and variations in vessel contrast. As a result, automatically or semi-automatically generated lumen profiles often exhibit inaccuracies that require further refinement.

[0074] In current clinical workflows, these refinements are typically performed manually by specialists who review and adjust the segmented luminal contours to ensure diagnostic accuracy. This process is highly time-consuming and labor-intensive, requiring a detailed examination of the entire luminal structure to identify and correct potential inaccuracies. The need for manual adjustments significantly prolongs the overall examination time and can delay the generation of the final clinical report, which is particularly critical in time-sensitive procedures such as coronary artery assessment.

[0075] Beyond the direct impact on diagnostic efficiency, extended segmentation time increases operational costs and limits the availability of imaging resources. The extended time required for expert review reduces the number of patients that can be examined within a given timeframe and may delay the scheduling of subsequent examinations. In busy clinical settings, this creates a bottleneck in the diagnostic workflow, ultimately impacting patient volume and resource allocation.

[0076] Existing methods for segmentation refinement offer users only limited guidance on which areas of the lumen profile require correction. Some methods visualize segmentation uncertainty through confidence maps or color-coded overlays, but these techniques merely highlight uncertain areas without prioritizing them based on clinical relevance. Users still must manually evaluate all marked areas without knowing which corrections will have the most impact on the final diagnostic outcome.

[0077] The proposed system identifies areas of uncertainty that have a clinically significant impact on the calculation of coronary hemodynamic parameters, allowing users to allocate their time more efficiently. By directing attention to the most critical areas, the time required for manual examination and adjustment can be minimized, resulting in a reduction in overall examination duration and optimization of resource utilization in the clinical workflow. This, in turn, enables higher patient throughput while ensuring that corrections contribute meaningfully to diagnostic accuracy, rather than refining contours with negligible impact on the final assessment.

[0078] Existing methods often struggle to distinguish between uncertainties relevant to clinical decision-making and uncertainties that have the least impact on diagnostic outcomes. Therefore, users may expend considerable effort correcting for contour regions that substantially do not affect the calculated coronary hemodynamic parameters, while potentially ignoring areas where adjustments could significantly alter diagnostic results.

[0079] Where appropriate, the above configurations, developments, and implementations can be combined with each other as needed, provided that this is reasonable. Other possible configurations, developments, and implementations of the invention include combinations of features of the invention not explicitly mentioned previously described or described below with reference to embodiments. In particular, those skilled in the art will also add aspects in this context as improvements or supplements to the basic form of the invention. Attached Figure Description

[0080] The invention is described in more detail below with reference to the embodiments shown in the accompanying drawings, wherein:

[0081] Figure 1 A schematic overview of the system architecture is shown, illustrating the key components involved in acquiring, processing, and analyzing angiographic image data;

[0082] Figure 2 An example of a vascular structure with candidate portions identified for luminal contour adjustment is shown, including multiple variations used to determine the impact on the score;

[0083] Figure 3 A graphical user interface is shown that displays angiography image data, identified candidate segments, and a priority list for user interaction; and

[0084] Figure 4 A process flowchart outlining a method for prioritizing lumen profile adjustments is shown, including key processing steps performed by the system.

[0085] The accompanying drawings are intended to provide a further understanding of embodiments of the invention. They illustrate embodiments and, in conjunction with the description, help to explain the principles and concepts of the invention. Other embodiments and numerous advantages mentioned will become apparent from the drawings. Elements in the drawings are not necessarily shown to scale.

[0086] In the accompanying drawings, unless otherwise specified, identical, functionally equivalent, and identically operating elements, features, and parts have the same reference numerals in every case. Detailed Implementation

[0087] For better illustration, and unless otherwise stated, embodiments of the invention are described with respect to the calculation of fractional flow reserve (FFR), and the invention is not limited to the calculation of this particular coronary hemodynamic parameter. FFR is used as an example of a coronary hemodynamic parameter, and the described embodiments can be used to calculate other coronary hemodynamic parameters, particularly those selected from instantaneous flow-free ratio (iFR), the ratio of resting distal pressure to aortic pressure (resting Pd / Pa), coronary flow reserve (CFR), and microvascular resistance (IMR).

[0088] Figure 1 An exemplary system architecture for prioritizing luminal contour adjustments in fractional flow reserve (FFR) calculation is schematically illustrated. The system includes an imaging unit 10 for acquiring angiographic image data, a preprocessing unit 20 for enhancing and preparing the image data, and a device 30 configured to analyze the luminal contour and generate a priority list of editing candidates.

[0089] Imaging unit 10 includes an angiography table 11 on which the patient 12 is positioned during the procedure. A radiographic system 13 with a C-arm is used to acquire X-ray images of the coronary arteries, which is essential for assessing the presence and severity of coronary artery disease (CAD). These arteries supply the heart with oxygenated blood, and any narrowing or stenosis can significantly affect cardiac function.

[0090] The acquired image data is transmitted to preprocessing unit 20, which performs various enhancement operations to optimize the images for further analysis. These preprocessing steps may include techniques such as noise reduction, contrast enhancement, and segmentation refinement. Depending on the imaging protocol and clinical requirements, the image data may include individual X-ray images or entire sequences. Analyzing multiple frames in the sequence can improve accuracy by compensating for motion artifacts and providing a more comprehensive view of the coronary artery anatomy.

[0091] Accurate assessment of the luminal contour is crucial for determining the severity of stenosis and calculating the fractional flow reserve (FFR), which provides a functional evaluation of flow restriction. However, several factors can introduce uncertainty into luminal contour determination. Variations in image quality, such as noise and contrast limitations, can affect segmentation accuracy. In some cases, portions of a vessel may be partially obscured due to overlapping arteries or calcifications, making it difficult to extract a clear luminal contour. The appearance of a vessel segment may also be distorted depending on the viewpoint from which the image was captured, leading to inaccurate contour estimation. Furthermore, motion artifacts caused by patient movement or cardiac motion can further complicate the segmentation process.

[0092] The input unit 31 of device 30 receives preprocessed angiography image data and provides it to the graphical user interface 33, where the images are displayed to the user. At this stage, the images primarily serve as a visual representation of the coronary anatomy, allowing the user to review the acquired data before applying any further computational processing. The initial display does not yet include any segmentation overlay or uncertainty indicators, ensuring that the raw images remain usable for review in their unchanged form.

[0093] While angiography provides a powerful tool for assessing coronary artery anatomy, the presence of uncertainties in the extracted luminal contours can pose a challenge to accurate diagnosis. These uncertainties are caused by a variety of factors, including image noise, vessel overlap, and perspective-related distortion, as previously described. In clinical practice, resolving these uncertainties is typically performed manually, requiring expert users to review the images and make corrections as needed.

[0094] To assist this process, existing software solutions can incorporate automated or semi-automated tools that attempt to highlight areas of potential segmentation uncertainty. These tools analyze variations in contour detection and use predefined criteria to mark areas where inconsistencies may exist. Alternatively, some systems allow users to manually select areas they deem uncertain based on their own clinical judgment. Whether uncertainty is automatically or manually identified, the process of examining and adjusting contour segments remains time-consuming and can significantly delay the final diagnosis.

[0095] Given the complexity and effort involved in manual correction, this invention aims to optimize uncertainty handling by identifying the most critical uncertainties. Instead of presenting the user with all potential areas of uncertainty, the system assesses which uncertainties are most relevant to the final diagnostic outcome and prioritizes these uncertainties for correction. This approach ensures that users can focus on modifications that offer the greatest potential for improvement while minimizing unnecessary effort.

[0096] Once the angiography image data has been received and displayed, the processing unit 32 performs an automatic analysis of the luminal contours within the displayed image. The contours are initially determined based on an automatic or semi-automatic segmentation process that extracts the vessel boundaries from the image. This segmentation serves as the basis for further refinement and uncertainty assessment.

[0097] To ensure that only the most relevant regions are considered for potential corrections, the system identifies candidate portions of the lumen profile that may require editing. These candidates are determined either based on a pre-calculated uncertainty metric (which assesses the confidence level of segmentation accuracy) or through manual selection by the user.

[0098] For each identified candidate, the system calculates an influence score, which estimates the extent to which modifying that particular region will affect the final calculated fractional flow reserve (FFR) value. The influence score is expressed as a numerical value within a predefined range, such as between 0 and 1, 0 and 255, or another suitable scale. A higher influence score indicates that adjustments in the corresponding region are expected to have a greater impact on the final diagnostic result, while a lower influence score indicates that modifications in that region may have only a smaller impact.

[0099] Scores are calculated by generating multiple variations of the contour adjustment and analyzing how these variations affect the calculated FFR results. Based on these calculations, the system generates a priority list of edit candidates and sorts the candidates according to their respective influence scores.

[0100] Then, a priority list is presented to the user via a graphical user interface 33, allowing the user to focus on the modifications that are expected to have the most significant impact on the diagnostic results.

[0101] In online variant analysis, the system directly utilizes fractional flow reserve (FFR) calculations to assess the effect of modifying specific candidate portions of the luminal profile. For each candidate, multiple profile variations are generated, and their impact on the calculated FFR value is analyzed. This approach allows for direct assessment of the strength of profile adjustments' influence on diagnostic results, ensuring that only the most relevant modifications are prioritized.

[0102] The process begins with the selection of candidate segments, identified through an uncertainty metric or manual user selection. Once a candidate is selected, the system generates multiple variations of its contour. These variations can be generated using various methods, including random perturbations, where slight contour adjustments are applied within a predefined range of uncertainty, or system modifications such as shifting, scaling, or smoothing of contour segments. Additionally, algorithmic optimization modifications can be used to maximize the variance of the calculated FFR value. The number of variations applied can be dynamically adjusted based on image attributes, available computational resources, or user-defined parameters.

[0103] For each generated variation, the system calculates the FFR value and records the resulting variation. By analyzing the variance across all calculated values, an influence score is determined, quantifying the extent to which the moderating effect of that candidate component is expected to influence the overall FFR outcome. This process is repeated for all identified candidates, generating a priority list of editable candidates ranked according to their respective influence scores.

[0104] Users can interact with the system via a graphical user interface 33 to refine their analysis. The system can be configured to evaluate all identified candidates, focusing on areas of high uncertainty, or allow users to manually select a subset of candidates for evaluation. Once a priority list is generated, the system presents it to the user, enabling them to focus on modifications that have the highest expected impact on the final diagnostic outcome.

[0105] Figure 2 The evaluation process for candidate portions of the lumen profile based on the above method is schematically illustrated. The coronary vessel 100 includes two identified candidate portions 110 and 120, representing regions where segmentation uncertainty has been detected or where manual correction has been considered.

[0106] For each candidate segment, multiple variations are generated and evaluated to determine their impact on the calculated FFR value. In the first candidate 110, three variations 111, 112, and 113 are applied, each representing a possible adjustment to the profile. The system then calculates the corresponding FFR value, analyzes the variations, and determines the impact score SC1 for the first candidate. A similar process is applied to the second candidate 120, where multiple variations are introduced, and the impact score SC2 is calculated.

[0107] Although Figure 2 Only two candidate parts, 110 and 120, and three variations for each candidate are shown; however, this representation is merely an example. The actual number of candidates and variations can vary depending on the specific application and computational constraints.

[0108] Compared to online variants that compute FFR values ​​on an instantaneous basis for each contour change, offline variants can rely on pre-trained machine learning models to directly predict impact scores. This approach significantly reduces the computational burden and enables faster prioritization of edit candidates, making it particularly advantageous in time-sensitive clinical settings.

[0109] To train the model, a large dataset of annotated luminal contours is collected, which may include real patient data, synthetic angiography data, or a combination of both. Editing candidates are identified based on uncertainty measures, manual annotation, or randomized sampling along the contour. For each candidate, an influence score is calculated using a method similar to online variants, where variations are applied and their impact on the calculated FFR value is analyzed. The training process then involves learning the relationship between contour features, uncertainty measures, and influence scores, allowing the model to predict the expected impact of contour adjustments without explicit FFR calculations.

[0110] Once trained, the model can be deployed in real-time applications, where it predicts the influence score for each candidate region based on the patterns it has learned. This allows the system to generate a priority list of edit candidates much faster than in online variants, while still maintaining high accuracy. Offline methods can be used independently or in combination with online variants. In hybrid systems, the machine learning model provides an initial estimate of the influence score, which can then be refined by selectively calculating the FFR for the most relevant candidates, thus balancing speed and accuracy.

[0111] To implement offline variants, a machine learning model must first be trained using a dataset representing a wide range of possible luminal contour variations and their associated impact scores. The training data consists of luminal contours extracted from real patient angiography images, synthetic datasets, or a combination of both. Each dataset sample includes one or more candidate portions of the luminal contour, identified based on uncertainty metrics, manual annotation, or random selection along the contour.

[0112] For each candidate, the system generates multiple contour variations, similar to the approach used in online variants, and evaluates their impact on the calculated FFR value. These evaluations provide ground-based impact scores, which are stored along with the contour data. A machine learning model is then trained to learn the relationship between the contour's geometric features, segmentation uncertainty, and the resulting impact scores.

[0113] During training, the model improves its ability to predict influence scores based on novel, unseen contour data without requiring direct FFR calculations. This allows the system to approximate the expected influence of contour adjustments purely based on learned patterns, significantly reducing computational overhead compared to online methods.

[0114] Once trained, the model is deployed in an offline variant to predict impact scores in real time. When processing new angiographic images, the model receives lumen contour data and candidate portion information, then outputs a predicted impact score, which is used to generate a priority list of edit candidates.

[0115] The effectiveness of a trained model depends on the quality and diversity of the training data. Therefore, the training dataset can be periodically updated with new real-world cases to improve performance. Furthermore, the model can incorporate feedback from previous user corrections, allowing it to continuously refine its prioritization strategy.

[0116] While the online and offline variants of the system offer different advantages, they are not mutually exclusive. Instead, they can be strategically combined to balance computational efficiency with diagnostic accuracy. Combining the two approaches allows the system to benefit from the speed of the offline variant while retaining the high accuracy of the online variant in critical cases.

[0117] In one possible implementation, the offline model is used as the first step to generate an initial priority list of edit candidates based on the predicted impact scores. This step can be performed quickly because the offline variant does not require direct FFR calculation. Once the list is generated, the highest-ranked candidates can be further analyzed using the online variant, where direct FFR calculation validates and refines the estimated impact scores. This ensures that high-priority corrections are evaluated with maximum accuracy, while lower-priority candidates are still prioritized based on pre-computed offline predictions.

[0118] Another strategy involves dynamically adjusting the balance between online and offline processing based on system constraints. In settings with limited computational resources, the system may rely more heavily on offline variants, while in situations where real-time accuracy is critical, online methods can be used to evaluate a larger proportion of candidates.

[0119] In addition to prioritizing candidates, the system can continuously improve its offline model by incorporating feedback from online evaluations. If a difference is observed between the predicted impact score (offline) and the actual calculated value (online), this data can be used to improve the training set for future iterations, resulting in a more accurate offline model over time.

[0120] This adaptive hybrid approach offers flexibility, ensuring an optimal balance between computational constraints, diagnostic requirements, and system efficiency. In addition to prioritizing lumen profile adjustments based on uncertainty and impact scores, the system can further enhance its performance by learning from previously performed corrections. This approach allows the system to adjust over time, thereby improving the offline impact assessment model and the overall prioritization strategy.

[0121] Whenever a user modifies a candidate portion of the lumen profile, the system stores the original profile, the adjustments made, and their impact on the calculated FFR value. By aggregating such data across multiple cases, a historical correction database is built, capturing patterns in common segmentation errors, user preferences, and typical influence score distributions. The stored data can be used to refine uncertainty measurements, enhance influence score predictions, and adjust prioritization rules. If certain regions are corrected frequently in a similar manner, the system can adjust its uncertainty estimates to reflect these recurring trends, thereby improving candidate selection for future scenarios. Machine learning models used in offline variants can be incorporated into historical user corrections to better estimate the actual impact of profile modifications on FFR calculations. The system can also identify whether a user consistently focuses on a particular anatomical region or the type of segmentation problem, allowing for dynamic refinement of candidate ranking.

[0122] By leveraging past corrections, the system effectively moves towards an adaptive improvement cycle, where each user interaction contributes to a prioritized process of continuous optimization. This ensures that the accuracy and efficiency of lumen profile adjustment improves over time, thereby reducing the need for manual intervention and streamlining the clinician's workflow.

[0123] In some implementations, the calculated FFR value is assigned to a diagnostic category or class based on predefined thresholds. These classification thresholds may be determined based on clinical guidelines or specific diagnostic criteria.

[0124] The impact of contour modification on score calculation can be considered in determining whether it causes the FFR value to shift from one category to another. For example, if the FFR value is initially below a critical threshold, such as 0.8, and contour modification results in a value above that threshold, the classification of coronary artery lesions can change from clinically significant stenosis to insignificant stenosis. Conversely, modifications that shift the FFR value from above the threshold to below the threshold can also have significant diagnostic implications.

[0125] To reflect this clinical relevance, the system can assign higher impact scores to modifications that might lead to such a classification shift. A threshold of 0.8 is commonly used in clinical practice, but the system can be configured to handle different predefined thresholds based on clinical settings. Furthermore, unlike FFR values ​​which remain within the same category before and after adjustment, impact scores can be more strongly weighted when modifications affect diagnostic decisions.

[0126] By integrating classification-dependent weighting into the impact score calculation, the system ensures that the most diagnostically relevant modifications are prioritized, thereby improving clinical efficiency and accuracy.

[0127] In addition to assessing the effects of individual contour modifications in isolation, the system can also consider the interdependencies between different candidate segments of the luminal contour. In many cases, adjusting one candidate segment may not fully realize its effect until another segment is modified. After previous modifications have altered the overall flow conditions within the vessel, candidate segments that initially appeared to have low influence scores may become more relevant.

[0128] For example, if upstream stenosis is underestimated due to an initially incorrect profile, the corresponding flow resistance in the vessel can be artificially increased. This may result in more distal stenosis appearing to have only a minor impact on the calculated FFR value. However, once the upstream stenosis is corrected, distal stenosis may become a more significant factor because the adjusted flow distribution leads to different pressure conditions downstream. To account for such effects, the system can be configured to analyze modifications not only individually but also in combination with other candidate adjustments, allowing for a more comprehensive assessment of the impact.

[0129] In some implementations, the priority of editing candidates can be dynamically adjusted as modifications are applied. Instead of treating each candidate as an isolated entity, the system can continuously re-evaluate the ranking of remaining candidates to reflect changes in flow conditions, vascular resistance, or profile geometry. This adaptive prioritization strategy helps ensure that clinically significant modifications are not overlooked simply because they initially seem less relevant. By incorporating this approach, the system can improve the accuracy of the prioritization process and enable users to focus their attention on the most impactful adjustments at each stage of the review process.

[0130] As described above, the methods described are not limited to fractional flow reserve, but can also be applied to other coronary hemodynamic parameters that provide additional physiological insights into coronary artery disease. Depending on the diagnostic purpose, the system can determine the impact score based on different indices, including instantaneous flow-free ratio (iFR), the ratio of resting distal pressure to aortic pressure (resting Pd / Pa), coronary flow reserve (CFR), and microvascular resistance index (IMR).

[0131] Each of these indices has specific characteristics that influence how contour modifications affect the final diagnostic outcome. While FFR is calculated under congestion conditions and relies on flow resistance modeling, iFR and resting Pd / Pa operate under resting conditions and may exhibit different sensitivities to contour changes. CFR provides an assessment of coronary blood flow volume and can be influenced by both epicardial and microvascular resistance, while IMR is primarily used to assess microvascular function.

[0132] Because of these differences, the system can adjust its impact score determination based on the selected hemodynamic parameters. In some implementations, different prioritization strategies can be applied depending on whether the assessment focuses on the congestion index or the resting state index. Furthermore, the interdependence between contour modifications can vary based on the physiological characteristics of the selected indices. For example, autoregulation mechanisms in the resting state index can compensate for some contour adjustments, thereby reducing their overall impact on the diagnostic outcome.

[0133] By supporting multiple coronary hemodynamic parameters, this system enables a flexible and clinically relevant approach to prioritize luminal contour adjustment. This adaptability ensures the method remains applicable across different diagnostic protocols, providing clinicians with optimized guidance based on the most relevant physiological parameters.

[0134] Figure 3 An exemplary graphical user interface 200 is schematically illustrated, designed to facilitate viewing and adjusting the luminal contour within angiographic images. A central display area presents the vascular structure 210, where certain candidate portions 201-205 have been identified as uncertain based on predefined uncertainty measures or user input. These candidates may represent regions where contour adjustment could potentially improve the accuracy of FFR calculations.

[0135] This system allows users to interact with identified candidates using input devices such as a mouse or touchscreen. Users can manually select one or more candidate subsets for analysis, or alternatively, the system can automatically determine a subset of candidates based on predefined selection criteria. Analysis can be performed on individual user-selected candidates, all identified candidates, or automatically determined groups that the system prioritizes.

[0136] Once the impact assessment is complete, the system provides visual feedback to indicate the priority of each candidate segment. This may include color-coded highlighting, numerical labels, or other graphical markers reflecting the estimated impact score of each candidate. Alternatively or additionally, the priority ranking results can be presented in a structured list 220 displayed alongside the vascular structures. This list provides an ordered ranking of candidate segments, allowing users to quickly navigate to the most relevant areas for review and potential modifications.

[0137] The graphical user interface 200 is designed to simplify the lumen profile adjustment process by providing intuitive visualization and an efficient selection mechanism. This ensures that users can focus their attention on the most critical profile modifications, thereby improving the overall efficiency and accuracy of the diagnostic workflow.

[0138] Figure 4The illustration shows a schematic process flow of a computer-implemented method for prioritizing lumen profile adjustments as described in the independent claims. The steps of the method are labeled S1 to S6, corresponding to the main operations performed by the system. It should be understood that the method can be implemented using the previously described apparatus, and the features of the apparatus and method can be combined in any suitable manner. Additionally, the method may include further steps that facilitate the execution of functions provided by the apparatus, and vice versa.

[0139] In step S1, angiographic image data representing the coronary vessels is received. This data serves as the basis for subsequent analysis and is typically acquired using an imaging system as described in the preceding sections.

[0140] In step S2, the lumen contour is determined based on automatic or semi-automatic segmentation. The segmentation process can utilize image processing algorithms, machine learning models, or a combination thereof to extract the vessel boundaries.

[0141] In step S3, candidate portions of the lumen profile for editing are identified. This can be based on an uncertainty metric that quantifies segmentation reliability, or on manual user input where the user directly selects the region of interest. As previously mentioned, candidate determination can involve multi-frame aggregation, uncertainty prediction based on machine learning, or analysis of geometric features.

[0142] In step S4, an impact score for each candidate section is determined, for example, by analyzing multiple changes in the lumen profile adjustment. The impact score represents the estimated effect of modifying the corresponding candidate section on the calculated FFR value. Depending on the implementation, different methods may be used, including online calculation methods that simulate multiple changes in real time or offline models that rely on pre-trained machine learning predictions.

[0143] In step S5, the candidate parts are ranked based on their respective influence scores to generate a priority list of edit candidates. This ranking allows the system to direct the user's attention to the most relevant modifications, thereby ensuring effective and clinically meaningful adjustments.

[0144] In step S6, a priority list is presented to the user via a graphical user interface. The visualization may include a structured list of candidates, a color-coded display of an angiography image, or a combination of both. The user can then interact with the priority list, selecting areas to view and modify.

[0145] The methods described herein can be implemented as a computer program product stored on a non-transitory computer-readable storage medium. The computer program includes instructions that, when executed by one or more processors, cause the processor(s) to perform the steps of the methods described in the preceding sections. The computer-readable storage medium may include, but is not limited to, flash memory, optical discs, solid-state drives, or magnetic storage devices.

[0146] This method can be executed on a single processing unit or distributed across multiple processors, depending on computational requirements and system architecture. In some implementations, computational tasks related to uncertainty estimation, influence score determination, and ranking can be performed on a dedicated processing unit, while the graphical user interface and user interaction are handled by a separate processing system.

[0147] This ensures that the method can be effectively deployed on different hardware configurations, including standalone workstations, cloud-based processing environments, and integrated medical imaging systems.

[0148] Although the present invention has been described above by way of embodiments, the invention is not limited thereto, and can be modified in various ways. In particular, the invention can be changed or modified in various ways without departing from its core principles.

[0149] In summary, this invention provides a computer-implemented method and corresponding apparatus for prioritizing luminal contour adjustments in angiographic image data to improve the accuracy of coronary hemodynamic parameter calculations. The method includes receiving angiographic image data, determining the luminal contour through automatic or semi-automatic segmentation, and identifying candidate portions for editing based on uncertainty metrics or user input. For each candidate portion, an influence score is determined by analyzing multiple contour variations, quantifying the impact of adjustments on the calculated coronary hemodynamic parameter values. Candidates are ranked according to their influence scores, and the priority list is presented to the user via a graphical interface for review and potential modification. This invention enables effective correction of diagnostically relevant regions while minimizing unnecessary adjustments.

Claims

1. A computer-implemented method for prioritizing lumen contour adjustment in a coronary artery hemodynamic parameters calculation system, characterized in that, The method includes: Receive (S1) angiographic image data representing coronary vessels; The lumen contour in the angiography image data is determined based on automatic or semi-automatic segmentation (S2); The candidate portions of the lumen profile to be edited are determined based on an uncertainty measure associated with the segmentation or based on user input (S3). An impact score for each candidate segment is determined by analyzing multiple changes in the lumen profile adjustment (S4), wherein the impact score represents the estimated impact of the adjustment of the corresponding candidate segment on the calculated coronary hemodynamic parameters of the relevant changes in the lumen profile adjustment. The candidate parts are sorted based on their respective influence scores (S5) to generate a priority ranking list of edit candidates; and The user is provided with the priority-sorted list described in (S6) via the graphical user interface (200) for viewing and potential contour adjustments.

2. The method according to claim 1, characterized in that, The determination (S4) of the impact score includes analyzing multiple changes in the luminal profile adjustment and quantifying the changes in the calculated coronary hemodynamic parameters.

3. The method according to any one of the preceding claims, characterized in that, Candidate portions of the lumen profile are identified based on an uncertainty metric, wherein the uncertainty metric is derived from one or more of the following: - Multi-frame segmentation aggregation, in which the lumen contour is extracted from multiple angiography images and the changes between segments are analyzed. - Machine learning-based prediction of segmented uncertainty, where a trained model estimates uncertainty based on pre-computed training data, and - Analysis of local geometric features, in which sudden changes in lumen curvature or discontinuities in segmentation are detected as indicators of uncertainty.

4. The method according to any one of the preceding claims, characterized in that, The influence score (S4) is determined by jointly evaluating multiple candidate portions of the lumen profile to account for the interdependence between profile modulations, wherein the influence score is determined based on the combined effect of the jointly evaluated profile modulations on the calculated coronary hemodynamic parameters.

5. The method according to any one of the preceding claims, characterized in that, Coronary hemodynamic indices are selected from fractional flow reserve (FFR), instantaneous flow-free ratio (iFR), ratio of resting distal pressure to aortic pressure, resting pressure (Pd / Pa), coronary flow reserve (CFR), and microvascular resistance index (IMR).

6. The method according to any one of the preceding claims, characterized in that, The impact score, as determined in (S4), includes: The lumen profile is modified by applying multiple variations, which are selected from one or more of the following: - Random perturbations within the range determined by the uncertainty measure, - Predefined system modifications, including shifting, scaling, or smoothing of contour segments, and - Modifications to the algorithm optimization, where changes are generated based on an optimization process aimed at maximizing the expected variance of coronary hemodynamic parameters. Calculate the coronary hemodynamic parameters for each variation, and The influence score is determined based on the variance of the calculated coronary hemodynamic parameters.

7. The method according to any one of the preceding claims, characterized in that, Determining (S4) the impact score includes using a previously trained machine learning model to predict the impact score.

8. The method according to claim 7, characterized in that, The method further includes training the machine learning model to predict influence scores, wherein the training further includes: Multiple training samples are provided, each including lumen profile and correlation data related to the impact score prediction. Calculate or assign influence scores to training samples, and The machine learning model is trained based on the training samples to predict the influence score of the new lumen profile. The training samples are obtained from one or more of real patient data, synthetic angiography data, or a combination thereof.

9. The method according to any one of the preceding claims, characterized in that, Determining (S4) the impact score also includes adjusting the impact score calculation based on previously performed lumen profile adjustments, specifically, wherein the adjustments are based on an analysis of historical user corrections stored in a database and are used to refine uncertainty measures or impact score predictions.

10. The method according to any one of the preceding claims, characterized in that, The method further includes: The coronary artery lumen is classified based on coronary hemodynamic parameters by categorizing regions according to one or more predefined thresholds. The determination of the impact score includes determining or adjusting the impact score based on whether modifying the lumen profile causes a change in the classification of the coronary artery region.

11. An apparatus (30) for prioritizing lumen contour adjustments in a coronary hemodynamic parameters calculation system, particularly by employing a computer-implemented method according to any one of claims 1-10, characterized in that, The device includes: The input unit (31) is configured to receive angiographic image data representing coronary vessels; Processing unit (32), which is configured as follows: - Determine the lumen contour in the angiography image data based on automatic or semi-automatic segmentation; - Candidate portions of the lumen profile to be edited are determined based on an uncertainty measure associated with the segmentation or based on user input; - An impact score for each candidate segment is determined by analyzing multiple changes in the luminal contour modulation, wherein the impact score represents the estimated impact of the modulation of the corresponding candidate segment on the calculated coronary hemodynamic parameters of the relevant changes in the luminal contour modulation; - Sort the candidate parts based on their respective influence scores to generate a priority-ranked list of edit candidates; and The output unit (33) is configured to provide the user with the prioritized list for viewing and potential contour adjustments via a graphical user interface (200).

12. The device according to claim 11, characterized in that, The device also includes: A display unit is configured to present angiographic image data representing coronary vessels to a user, wherein the display unit is further configured to indicate candidate portions of the lumen profile for editing based on the priority-sorted list of editing candidates. The candidate portions are provided as a structured list of editable candidates, which are sorted based on their respective impact scores and / or by adjusting the visual representation of the candidate portions within the angiographic image data based on the relevant impact scores of their respective candidate portions.

13. The device according to claim 11 or 12, characterized in that, The device also includes: A storage unit configured to store and provide training data for a machine learning model used to predict the impact score of luminal contour adjustment; and The processing unit is configured to generate training samples, calculate or assign influence scores to the training samples, and train a machine learning model based on the training samples.

14. A medical system for coronary artery assessment, characterized in that, The system includes: Imaging unit (10), which is configured to acquire angiographic image data representing coronary vessels; The device (30) according to any one of claims 11 to 13.

15. A computer program product comprising instructions, characterized in that, When executed by a processor, the instructions cause the processor to perform the method according to any one of claims 1 to 10.

Images