A method, system, and medium for human-machine interaction measurement of major and minor axes in medical images based on orthogonal locking.

By introducing a geometric angle rigid linkage and a real-time edge re-detection mechanism, the problems of easily disrupted orthogonal relationships and cumbersome manual adjustments in the measurement of major and minor axes in medical imaging are solved, achieving efficient and accurate lesion size measurement, meeting WHO and RECIST 1.1 standards, and improving the efficiency and accuracy of clinical diagnosis.

CN123118730APending Publication Date: 2026-11-17胡晨曦 +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611256756.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-19
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing technologies, the orthogonality of major and minor axis measurements in medical images is easily disrupted, manual adjustments are cumbersome, and subjective errors exist, failing to meet the stringent compliance requirements of WHO and RECIST 1.1.

Method used

A human-machine linkage measurement method for the major and minor axes of medical images based on orthogonal locking is adopted. By introducing geometric angle rigid linkage and real-time re-probing mechanism along the normal edge, an automated and intelligent linkage closed loop for the major and minor axes is realized, ensuring the automatic maintenance of orthogonal relationship and real-time locking of the edge.

Benefits of technology

It achieves rigid orthogonal geometric linkage between the major and minor axes, improving image reading efficiency by more than 60%, eliminating subjective measurement errors, meeting WHO evaluation standards, and improving the consistency and accuracy of clinical diagnosis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123118730A_ABST
    Figure CN123118730A_ABST
Patent Text Reader

Abstract

This invention discloses a human-computer interaction measurement method, system, and medium for medical image major and minor axes based on orthogonal locking, belonging to the field of medical image processing and computer-aided diagnosis. The method includes: acquiring medical images of the target lesion and initial segmentation information; rendering and presenting the initial primary and secondary geometric axes (major and minor axes); responding to user interaction with the primary axis; calculating and updating the secondary axis normal in real time based on the updated primary axis using geometric direction constraints, maintaining the angular linkage relationship within the orthogonal constraint or preset tolerance range; and adaptively capturing the latest edge of the lesion by emitting probe rays along the secondary axis normal based on the geometric intersection point, automatically updating the secondary axis length and endpoint position. This invention solves the problems of decoupled and isolated adjustment of major and minor axes, easy destruction of orthogonal relationships, suspended endpoints, and cumbersome image reading interaction in existing technologies. It achieves closed-loop geometric linkage and edge adaptive locking of major and minor axes, effectively expanding the scope of patent protection and significantly improving the efficiency and compliance accuracy of clinical dual-axis measurement.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the fields of medical image processing, computer-aided diagnosis (CAD) and intelligent medical interaction technology. Specifically, it relates to a human-computer interaction measurement method, system and computer-readable storage medium for medical image long and short diameters based on orthogonal locking. It is widely applicable to the accurate measurement of the dual diameters of lesions in various medical imaging modalities such as CT, MRI, ultrasound (US), PET-CT, X-ray tomography and digital pathology section (WSI). Background Technology

[0002] In clinical oncology diagnosis, efficacy evaluation, and follow-up, accurately measuring the longitudinal axis (L) and transverse axis (W) of irregular lesions such as solid tumors, lymph nodes, and pulmonary nodules, according to the WHO World Health Organization Criteria and RECIST 1.1 (Response Evaluation Criteria in Solid Tumors) guidelines, is a core quantitative indicator for assessing disease progression and evaluating the effectiveness of radiotherapy, chemotherapy, and targeted therapy. Strict clinical guidelines require that the long axis must be the straight-line distance between the two farthest points on the largest cross-section of the lesion; the transverse axis must be on the same cross-section, maintaining a strict 90° perpendicular orthogonal relationship to the long axis, and spanning the widest area of ​​the lesion's contour.

[0003] With the popularization of medical artificial intelligence (AI) technology, many PACS workstations and AI-assisted diagnostic systems can automatically segment lesions and generate initial long and short axes. However, because lesion edges in medical images are often accompanied by ground-glass opacities, adhesions, inflammatory exudates, or blurred boundaries, the axes automatically generated by AI often require manual fine-tuning and intervention by radiologists.

[0004] In existing technologies, the major and minor axes generated by AI or systems are usually completely decoupled (i.e., independent of each other) during manual fine-tuning. When a doctor manually drags or rotates the endpoints of the major axis, the minor axis remains in its original position and angle. This leads to two serious technical defects: First, the originally strict 90° geometric orthogonality is immediately destroyed, failing to meet the stringent compliance requirements of WHO and RECIST 1.1; second, because the position of the minor axis remains unchanged after the major axis is rotated, the two endpoints of the minor axis often detach from the actual physical boundary of the lesion, or even hang suspended in normal tissue or background.

[0005] To correct these deviations, doctors are forced to perform an extremely tedious and repetitive interactive process: first, drag and adjust the major axis; then, manually rotate the minor axis to estimate the verticality visually; and finally, stretch the two ends of the minor axis to align with the lesion edge. This process not only significantly increases the workload of image interpretation and easily leads to doctor fatigue, but also introduces unavoidable subjective human error into the visually estimated orthogonal angle, severely undermining the consistency between multicenter clinical trials and diagnoses. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of existing technologies in measuring the major and minor axes of medical images, such as easy disruption of orthogonality, cumbersome manual adjustments, easy deviation of endpoints from lesion boundaries, and large subjective human errors. This invention provides a human-machine collaborative measurement method, system, and medium for the major and minor axes of medical images based on orthogonal locking. By introducing a dual mechanism of "geometric angle rigid linkage" and "real-time re-detection along the normal edge (Rescan)," this invention achieves an automated and intelligent closed-loop linkage of the major and minor axes.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a human-machine interaction measurement method for the major and minor axes of medical images based on orthogonal locking, comprising the following steps: 1. Obtain medical images of the target lesion and corresponding initial segmentation information (such as semantic mask, polygon nodes, or depth probability map). 2. Render and present the initial primary geometric axis (major axis) and initial secondary geometric axis (minor axis) in the medical image; 3. In response to user interaction operations such as dragging, rotating, or moving the main geometric axis, obtain the updated main geometric axis; 4. Based on the updated main geometric axis, calculate and update the normal direction of the auxiliary geometric axis in real time using geometric direction constraints, so that the auxiliary geometric axis and the updated main geometric axis maintain an orthogonal constraint relationship or an angle linkage relationship within a preset angle tolerance range (such as 85°~95°); 5. Based on the geometric intersection, adaptive edge re-detection is performed along the normal direction of the auxiliary geometric axis to re-lock the lesion contour boundary and automatically update the length and endpoint position of the auxiliary geometric axis; 6. Calculate and output the physical dimension measurement results of the target lesion based on the updated primary and secondary geometric axes.

[0008] Furthermore, the geometric orthogonality calculation specifically includes: assuming the updated coordinates of the two endpoints of the principal geometric axis are respectively... and Calculate its direction vector Based on the dot product perpendicularity condition The orthogonal normal vectors of the auxiliary geometric axes are obtained by solving the problem. In embodiments with tolerances or specific angles, parameterized angle transformation relationships can be constructed by combining orthogonal rotation matrices.

[0009] Furthermore, prior to edge re-detection, a sliding optimization step for geometric intersections is included: setting a candidate intersection set. ,in Slide along the main geometric axis with a preset step size, and emit probe rays to both sides along the corresponding orthogonal normal at each candidate intersection point to calculate the profile span length. ; Select to make Candidate intersections that achieve a maximum value or meet specific clinical criteria are selected as the optimal geometric intersections. In addition, the system also supports locking the intersection point directly to the midpoint of the main geometric axis, the geometric centroid, or the historical relative scale position.

[0010] Furthermore, the termination conditions for the adaptive edge re-probing include: semantic mask mutation condition, grayscale gradient mutation condition, deep learning edge probability below a preset threshold condition, and distance safety radius limitation condition. Probe sampling supports sub-pixel precision calculations such as bilinear interpolation, bicubic interpolation, Lanczos interpolation, or nearest neighbor interpolation.

[0011] Furthermore, this method also includes a bidirectional, equivalent, dual-axis linkage mechanism: when the user drags the auxiliary geometric axis (short axis), the system automatically reverses and locks the direction of the main geometric axis (long axis) and adaptively re-probes its endpoint along the normal of the long axis; it also includes a three-dimensional tomographic cross-layer following and abnormal over-limit protection mechanism, when the slice is rolled to a lesion-free area, the auxiliary geometric axis is automatically hidden or a prompt is given that the measurement range is exceeded.

[0012] Secondly, the present invention provides a medical image long and short axis human-machine interaction measurement system, including an image and segmentation information acquisition module, an initial graphics rendering module, an interactive capture and orthogonal calculation module, an edge re-detection and endpoint update module, and a measurement result output module.

[0013] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0014] The beneficial effects of this invention are as follows: 1. Achieved rigid orthogonal geometric linkage between major and minor axes: When the user drags and adjusts the main axis (major axis), the auxiliary axis (minor axis) automatically maintains orthogonal deflection in real time, eliminating the need for manual secondary alignment. A single operation can complete the dual-axis update, improving image reading efficiency by more than 60%. At the same time, through the higher-level angular constraint expression, it effectively prevents competitors from bypassing patent protection through minute angular deviations.

[0015] 2. A dynamic rescan mechanism and edge adaptive locking are introduced: after the minor axis is deflected, rays are automatically emitted along the orthogonal normal to re-detect the latest real edge of the lesion, which completely solves the problem of the endpoint being suspended and detached from the physical boundary in the traditional decoupling mode, and greatly eliminates subjective measurement errors.

[0016] 3. Possesses extremely high clinical compliance and algorithm robustness: The sliding optimization mechanism perfectly meets the maximum vertical span requirement in the WHO assessment criteria; sub-pixel interpolation and multimodal termination conditions ensure measurement stability in low-contrast, blurred boundary, and concave lesion scenarios. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the overall algorithm of the method of the present invention; Figure 2 This is a schematic diagram illustrating the principle of orthogonal geometric linkage between major and minor axes and edge re-detection. Figure 3 A schematic diagram of local magnification and subpixel step sampling for X-ray redetection; Figure 4 This is a hardware topology diagram of the measurement system of the present invention. Detailed Implementation

[0018] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0019] like Figure 1 As shown, the present invention provides a human-machine interaction measurement method for the major and minor axes of medical images based on orthogonal locking, which includes the following core processing steps: Step S101: The system loads and acquires the DICOM medical images (such as CT, MRI, PET-CT, ultrasound, etc.) of the target lesion and the corresponding initial segmentation information (such as UNet, SAM predicted mask, contour polygon or depth probability map).

[0020] Step S102: Render and present the initial long axis in the viewing interface. With the initial minor axis The geometric cross that is formed.

[0021] Step S103: Capture user input to the long axis endpoint or Drag, rotate, or move interactive events.

[0022] Step S104: Perform geometric orthogonal calculation. Calculate the direction vector based on the updated major axis endpoints. And through the dot product equation calculate In embodiments with angular tolerance (e.g., 85°~95°), the normal can also be calculated by combining the rotation matrix with the tolerance parameter.

[0023] Step S105: Geometric Intersection Optimization. In embodiments that strictly conform to WHO standards, candidate intersection points are set. Slide the search along the major axis and find the point with the largest vertical span. The corresponding point is taken as the optimal intersection point. In a simplified embodiment, the intersection point can also be directly specified as the midpoint of the major axis or the geometric centroid.

[0024] Step S106: X-ray re-detection. Starting from, along Probe rays are emitted to both sides in orthogonal normal directions with sub-pixel steps (e.g., 0.25 pixels). When sampling non-integer coordinates, bilinear interpolation, bicubic interpolation, Lanczos interpolation, or nearest neighbor interpolation algorithms are used to calculate the pixel grayscale value, gradient derivative, or semantic mask probability value.

[0025] Step S107: Edge Termination Detection. Proceed along the ray, and when the semantic mask changes from 1 to 0 and a first-order grayscale gradient is detected... When the depth probability falls below the confidence threshold or reaches the upper limit of the safety radius, detection stops and a new short axis endpoint is locked. and For irregularly shaped lesions with depressions, the first edge mutation point closest to the geometric intersection should be identified first.

[0026] Step S108: Calculate the actual physical size by combining the Pixel Spacing (mm / pixel) property in the DICOM header file, and dynamically refresh and present it on the UI interface in real time.

[0027] Furthermore, this embodiment also supports a bidirectional, peer-to-peer, dual-axis linkage mechanism: when the doctor actively drags the short axis endpoint, the system state machine is triggered in reverse, using the short axis as the main driving axis, locking the long axis direction with geometric orthogonal constraints, and re-probing the long axis endpoint along the long axis direction. When continuously scrolling through a three-dimensional tomographic image (such as a CT slice), the system maintains the orthogonal deflection angle of the short and long axes unchanged, automatically following the centroid and contour of the lesion in the new slice layer to re-emit X-rays for detection; if scrolling to a lesion-free area or an area where the edges cannot be identified, the system automatically triggers an over-limit protection mechanism, hiding the short axis or providing a UI over-limit prompt.

Claims

1. A medical image long-short diameter man-machine linkage measurement method based on orthogonal locking, characterized in that, Includes the following steps: Acquire medical images of the target lesion and the corresponding initial segmentation information; The initial primary geometric axis and the initial secondary geometric axis are rendered and presented in the medical image; In response to user interaction operations such as dragging, rotating, or moving the main geometric axis, obtain the updated main geometric axis; Based on the updated main geometric axis, the normal direction of the auxiliary geometric axis is calculated and updated in real time using geometric direction constraints, so that the auxiliary geometric axis and the updated main geometric axis maintain an orthogonal constraint relationship or an angle linkage relationship within a preset angle tolerance range. Based on the geometric intersection, adaptive edge re-detection is performed along the normal direction of the auxiliary geometric axis to re-lock the lesion contour boundary and automatically update the length and endpoint position of the auxiliary geometric axis. The physical dimensions of the target lesion are calculated and output based on the updated primary and secondary geometric axes.

2. The medical image long-short diameter man-machine linkage measurement method according to claim 1, characterized in that, The initial segmentation information includes at least one of semantic segmentation mask, contour polygon nodes, probabilistic graph and deep learning prediction boundary; the main geometric axis is the long axis of the lesion and the auxiliary geometric axis is the short axis of the lesion; or the main geometric axis is the short axis of the lesion and the auxiliary geometric axis is the long axis of the lesion.

3. The medical image long-short diameter man-machine linkage measurement method according to claim 1 or 2, characterized in that, The preset angle tolerance range is 85° to 95°; the step of calculating and updating the normal direction of the auxiliary geometric axis in real time based on the updated main geometric axis using geometric direction constraints specifically includes: Let the coordinates of the two end points of the updated master geometric axis be and , and calculate the direction vector ; Based on the vector perpendicular dot product constraint condition Or based on the preset inclination angle α The normal vector of the auxiliary geometric axis is calculated and solved .

4. The medical image long-short diameter man-machine linkage measurement method according to claim 1, characterized in that, Before performing adaptive edge re-detection based on geometric intersections along the normal direction of the auxiliary geometric axis, a sliding optimization step for geometric intersections is also included, specifically: Let the candidate intersection set be set. ,in t ∈ 0 1 , The endpoints of the main geometric axis The direction vector of the principal geometric axis; Sliding search is performed along the main geometric axis with a preset step size, and a detection ray is emitted to both sides along a corresponding normal direction at each candidate intersection point, and the contour crossing length at the corresponding position is calculated and obtained ; Selecting candidate intersection points that maximize the length of the contour crossing Taking the candidate intersection point that maximizes or meets a certain clinical specification maximum as the optimal geometric intersection point Alternatively, the midpoint, geometric center, or historically fixed relative proportion point of the main geometric axis can be directly locked as the geometric intersection point.

5. The medical image long-short diameter man-machine linkage measurement method according to claim 1, characterized in that, The termination condition for the adaptive edge re-detection includes at least one of the following: Semantic Mask mutation condition: current sample point coordinates along the probing ray Satisfies initial segmentation Mask state converted from lesion internal value to background external value; Gray-level gradient abrupt change condition: The first or second order gray-level gradient derivative along the ray direction of the current coordinate is calculated in real time based on the interpolation operator. When the gradient magnitude reaches the adaptive threshold Th, it is determined that the edge has been reached. Model prediction probability condition: The predicted probability value of the lesion edge depth of the current pixel is lower than the preset confidence threshold; Distance safety boundary limit condition: The travel distance of the detection ray exceeds the safety detection radius set based on the length of the main geometric axis.

6. The medical image long-short diameter man-machine linkage measurement method according to claim 1, characterized in that, The sampling process in the edge re-detection adopts sub-pixel level precision calculation, sets the sampling step size to less than 1 pixel, and uses bilinear interpolation, bicubic interpolation, Lanczos interpolation or nearest neighbor interpolation algorithm to calculate its pixel gray value, gradient derivative or semantic mask probability value for non-integer coordinate sampling points.

7. The medical image long-short diameter man-machine linkage measurement method according to any one of claims 1 to 6, characterized in that, It also supports a two-way peer-to-peer dual-axis linkage mechanism: In response to the user's dragging or rotating operation on the endpoints or axis of the auxiliary geometric axis, the system automatically switches the auxiliary geometric axis to the main drive axis, locks the direction of the main geometric axis in the opposite direction based on geometric orthogonal constraints, and adaptively re-detects its endpoint position along the direction of the main geometric axis.

8. The medical image long-short diameter man-machine linkage measurement method according to any one of claims 1 to 6, characterized in that, It also includes a three-dimensional fault cross-layer following and anomaly edge fault tolerance mechanism: When continuously scrolling or switching slice layers in three-dimensional tomographic images, the relative angle constraint between the main geometric axis and the auxiliary geometric axis remains unchanged, and the X-ray re-detection is automatically performed based on the changes in the centroid and contour of the lesion in the new slice layer. When scrolling to the lesion disappearance layer, the edge undetectable layer, or the fault missing layer, the over-limit protection mechanism is triggered, automatically hiding the auxiliary geometric axis, presenting the dashed line prediction status, or giving a UI prompt that the measurement is out of the effective range. For concave or irregular multi-connected lesions, during re-probing, prioritize locking the first mutation edge point closest to the geometric intersection, or select the edge point pair with the largest span of the connected domain.

9. A medical image long-short diameter man-machine linkage measurement system, characterized in that, include: The image and segmentation information acquisition module is used to acquire medical images of the target lesion and the corresponding initial segmentation information; An initial graphics rendering module is used to render and present the initial primary geometric axis and the initial secondary geometric axis in the medical image; The interactive capture and orthogonal calculation module is used to respond to the user's interactive dragging of the main geometric axis and to calculate and update the normal direction of the auxiliary geometric axis in real time based on the geometric direction constraints. The edge re-detection and endpoint update module is used to emit detection rays along the normal of the auxiliary geometric axis starting from the geometric intersection, adaptively capture the latest edge of the lesion and automatically update the endpoint position and axis length. The measurement result output module is used to calculate and output the final size measurement result of the target lesion based on the physical pixel spacing.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the program is executed by the processor, it implements the steps of the human-machine interaction measurement method for the major and minor diameters of medical images as described in any one of claims 1 to 8.