Cervical lesion prediction method and device based on image processing

Abstract

The invention relates to the technical field of image processing, and discloses a cervical lesion prediction method and device based on image processing, and the method comprises the steps: carrying out the imaging of a cervical surface through a colposcope, and obtaining a cervical surface image; analyzing the cervical surface image, and positioning a lesion area; configuring pose parameters of an ultrasonic probe according to the position of the lesion area, so that the ultrasonic probe is aligned with the lesion area for scanning, and an ultrasonic image is obtained; the ultrasonic image is analyzed, and the tissue infiltration depth and the blood flow resistance index of the lesion area are obtained; and predicting cervical lesions according to the tissue infiltration depth and the blood flow resistance index. According to the cervical lesion prediction method and device based on image processing, the colposcope and the ultrasound are combined to detect the cervical lesion in the same process, and the detection accuracy and efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a method and device for predicting cervical lesions based on image processing. Background Art

[0002] Cervical cancer is a common malignant tumor that threatens women's health worldwide. Early detection and accurate classification are of great significance for improving the cure rate of patients. Currently, the commonly used methods for cervical cancer screening and lesion detection in clinical practice mainly include cervical cytology, HPV testing, colposcopy, and tissue biopsy. Among them, colposcopy, as an intuitive image detection method, can effectively display the morphological information of the lesion area and assist doctors in making preliminary diagnoses. However, traditional colposcopy can only obtain two-dimensional planar images of the cervical surface, which makes it difficult to accurately reflect the three-dimensional infiltration depth and pathological range of the lesion tissue in the cervical stroma.

[0003] On the other hand, ultrasound imaging, especially transvaginal ultrasound, is widely used in gynecology and can be used to assess cervical structure and mass size.

[0004] Although colposcopy and ultrasound are both commonly used methods for detecting cervical lesions, ultrasound testing is currently performed separately from colposcopy (first initial screening by colposcopy, then further evaluation by ultrasound). Ultrasound lacks precise spatial registration with colposcopy, and the specific location, depth, and range of the lesions are difficult to accurately correspond to, which can easily lead to missed diagnosis or misdiagnosis.

[0005] In addition, ultrasound testing and colposcopy are performed separately, and there is a delay between the two testing processes, making it difficult to make a comprehensive decision in the first place. Summary of the Invention

[0006] To this end, the purpose of the present invention is to overcome the problem in the prior art that it is difficult to combine colposcopy and ultrasound to detect cervical lesions in the same process, resulting in false detection and missed detection, and to provide a cervical lesion prediction method and equipment based on image processing, which combines colposcopy and ultrasound to detect cervical lesions in the same process to improve detection accuracy and efficiency.

[0007] In a first aspect, in order to solve the above technical problems, the present invention provides a cervical lesion prediction method based on image processing, comprising: The cervical surface is imaged through a colposcope to obtain an image of the cervical surface; Analyzing the cervical surface image to locate the lesion area; Configuring the position parameters of the ultrasound probe according to the position of the lesion area so that the ultrasound probe is aligned with the lesion area for scanning to obtain an ultrasound image; Analyzing the ultrasound image to obtain the tissue infiltration depth and blood flow resistance index of the lesion area; Cervical lesions are predicted according to the tissue infiltration depth and blood flow resistance index.

[0008] In one embodiment of the present invention, the cervical surface is imaged by a colposcope to obtain a cervical surface image, including configuring a multispectral colposcope, imaging the cervical surface based on the multispectral colposcope, and synchronously collecting a white light reflection image, a narrow-band imaging vascular image, and a fluorescent staining image of the cervical surface.

[0009] In one embodiment of the present invention, the cervical surface image is analyzed to locate the diseased area, including performing feature extraction on the white light reflection image to obtain surface texture features and edge features; performing feature extraction on the narrow-band imaging vascular image to obtain vascular density features and vascular morphology features; performing feature extraction on the fluorescent staining image to obtain fluorescence intensity distribution features; normalizing and weightedly fusion the surface texture features, edge features, vascular density features, vascular morphology features, and fluorescence intensity distribution features to obtain a fused feature map; performing threshold segmentation on the fused feature map, and obtaining the diseased area based on the threshold segmentation result.

[0010] In one embodiment of the present invention, the multispectral colposcope is configured, including an objective lens and a dichroic prism group, wherein the objective lens converges the reflected light from the cervical surface and enters the dichroic prism group; the dichroic prism group divides the incident light into a first light path, a second light path, and a third light path according to the wavelength band; wherein an RGB image sensor is configured on the first light path, and a white light reflection image of the cervical surface is collected by the RGB image sensor; a CMOS image sensor is configured on the second light path, and a narrow-band imaging blood vessel image of the cervical surface is collected by the CMOS image sensor; and a solid-state image sensor is configured on the third light path, and a fluorescent staining image of the cervical surface is collected by the solid-state image sensor.

[0011] In one embodiment of the present invention, the posture parameters of the ultrasound probe are configured according to the position of the lesion area, including extracting the two-dimensional pixel coordinates of the lesion area in the image coordinate system; converting the two-dimensional pixel coordinates into three-dimensional spatial coordinates in a spatial rectangular coordinate system according to the imaging geometric parameters of the optical system of the colposcope; and obtaining the posture parameters of the ultrasound probe according to the end coordinates of the ultrasound probe and the three-dimensional spatial coordinates.

[0012] In one embodiment of the present invention, the ultrasound image is analyzed to obtain the tissue infiltration depth of the lesion area, including extracting the boundary of the cervical stromal layer using an image segmentation algorithm, setting multiple measurement points evenly spaced within the projection range of the lesion area, calculating the vertical distance from each measurement point to the nearest stromal layer boundary, and taking the maximum value of the distances of all measurement points as the tissue infiltration depth.

[0013] In one embodiment of the present invention, the ultrasound image is analyzed to obtain the blood flow resistance index of the lesion area, including performing color Doppler blood flow imaging on the ultrasound image to identify all blood flow signals; calculating the peak systolic velocity PSV and the end-diastolic velocity EDV of each blood flow signal; calculating the resistance index RI of each blood flow signal according to the following method: RI=(PSV-EDV) / PSV; and taking the average of the resistance indices of all blood flow signals as the blood flow resistance index.

[0014] In one embodiment of the present invention, identifying the blood flow signal includes extracting an initial blood flow signal pixel set through color Doppler blood flow imaging processing; binarizing the initial blood flow signal pixel set to obtain an initial blood flow signal binary image; performing noise reduction processing on the initial blood flow signal binary image to obtain a noise-reduced binary image; performing region growing processing on the noise-reduced binary image and merging adjacent blood flow signals to obtain a set of blood flow connected areas; calculating the pixel area of ​​each blood flow connected area, screening the blood flow connected areas with pixel areas greater than or equal to a threshold pixel as valid blood flow areas; identifying blood flow signals corresponding to each of the valid blood flow areas to obtain all the blood flow signals.

[0015] In one embodiment of the present invention, predicting cervical lesions based on the tissue infiltration depth and the blood flow resistance index includes performing a logarithmic transformation on the tissue infiltration depth to obtain a tissue infiltration characteristic index; performing a Z-score normalization transformation on the blood flow resistance index to obtain a blood flow resistance index; and obtaining a cervical lesion risk index S based on the following method: S=0.6×ln(D+1)+0.4×(1-Z); Where D represents the depth of tissue infiltration; Z represents the normalized result of the blood flow resistance index Z-score.

[0016] In a second aspect, based on the same inventive concept, the present invention further provides a device for predicting cervical lesions based on image processing, which is used to implement the method for predicting cervical lesions based on image processing, comprising: A colposcope module, used for imaging the surface of the cervix and obtaining an image of the cervix surface; An image analysis module, configured to analyze the cervical surface image and locate the lesion area; A posture adjuster, equipped with an ultrasound probe, is used to adjust the posture parameters of the ultrasound probe so that the ultrasound probe is aligned with the lesion area; an ultrasonic probe, used for scanning the lesion area and obtaining an ultrasonic image of the lesion area; an ultrasound analysis module, configured to analyze the ultrasound image to obtain the tissue infiltration depth and blood flow resistance index of the lesion area; A lesion prediction unit is used to predict cervical lesions based on the tissue infiltration depth and blood flow resistance index.

[0017] The above technical solution of the present invention has the following beneficial effects compared with the prior art: The image processing-based cervical lesion prediction method and device described in the present invention combine colposcopy and ultrasound to detect cervical lesions in the same process, thereby improving detection accuracy and efficiency.

[0018] Among them, the lesion area on the surface of the cervix is ​​located through colposcopy imaging, the position of the lesion area is used as the scanning object to align the position parameters of the ultrasound probe, and the lesion area is scanned to obtain an ultrasound image containing the deep structure information of the lesion area. The tissue infiltration depth and blood flow resistance index are combined to evaluate cervical lesions, improve detection accuracy, and significantly reduce the false detection rate and missed detection rate. In addition, colposcopy and ultrasound detection are carried out in the same process, and the targeted scanning time from colposcopy to ultrasound is shortened to less than 2 minutes, avoiding the repetitive actions of traditional "blind scanning" and improving detection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments of the present invention in conjunction with the accompanying drawings, wherein: Figure 1 This is a flow chart of a method for predicting cervical lesions based on image processing in a preferred embodiment of the present invention; Figure 2 This is a flow chart of analyzing a cervical surface image to locate a lesion area in a preferred embodiment of the present invention; Figure 3 This is a structural block diagram of a multi-spectral colposcope in a preferred embodiment of the present invention; Figure 4 This is a flow chart of obtaining the posture parameters of the ultrasound probe in a preferred embodiment of the present invention; Figure 5 This is a structural block diagram of a cervical lesion prediction device based on image processing in a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0020] The present invention will be further described below with reference to the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it. However, the embodiments are not intended to limit the present invention.

[0021] The purpose of the embodiments of the present invention is to solve the problem in the prior art that it is difficult to combine colposcopy and ultrasound to detect cervical lesions in the same process, resulting in false detection, missed detection and low detection efficiency, and to provide a cervical lesion prediction method and device based on image processing, which combines colposcopy and ultrasound to detect cervical lesions in the same process to improve detection accuracy and efficiency.

[0022] Example 1: Reference Figure 1 As shown, the embodiment of the present invention discloses a method for predicting cervical lesions based on image processing, comprising: S100, imaging the surface of the cervix through a colposcope to obtain an image of the surface of the cervix; S200, analyzing the cervical surface image to locate the lesion area; S300, configuring the position parameters of the ultrasound probe according to the position of the lesion area, so that the ultrasound probe is aligned with the lesion area for scanning to obtain an ultrasound image; S400, analyzing the ultrasound image to obtain the tissue infiltration depth and blood flow resistance index of the lesion area; S500: Predicting cervical lesions according to the tissue infiltration depth and blood flow resistance index.

[0023] In specific application scenarios, narrow-band imaging (NBI) or fluorescence colposcopy is used to enhance the contrast of lesions (such as acetowhite epithelium and atypical blood vessels). Videos of the dynamic changes of the cervix in the presence of acetic acid / iodine solution are recorded, and abnormal areas are extracted using inter-frame differencing. A U-Net network is trained to segment the cervical transformation zone, and a thresholding method in HSV color space is used to identify suspicious lesions (such as bright / dark spots). A lesion probability heatmap is generated by combining texture features, vascular morphology, and boundary irregularity. Manual annotation and correction are supported, and the center coordinates and boundaries of the lesion area are output. A colposcopy-ultrasound coordinate system conversion model is established, and the spatial relationship between the two is calibrated using a calibration plate. The robotic arm adjusts the ultrasound probe based on the coordinates of the lesion area to ensure that the sound beam enters the lesion perpendicularly. In ultrasound images, a dynamic programming algorithm is used to track the basement membrane rupture point, calculate the deepest lesion penetration distance, and obtain the tissue invasion depth. Peak systolic velocity and end-diastolic velocity are extracted from Doppler spectra to calculate the resistance index. Cervical lesions are assessed by combining tissue invasion depth and resistance index.

[0024] The image processing-based cervical lesion prediction method of the present invention combines colposcopy and ultrasound to detect cervical lesions in the same process, thereby improving detection accuracy and efficiency.

[0025] Among them, the lesion area on the cervical surface is located through colposcopy imaging, and the position parameters of the ultrasound probe are aligned with the location of the lesion area as the scanning object. The lesion area is scanned to obtain the deep structural information of the lesion area. The cervical lesions are evaluated by combining the tissue infiltration depth and blood flow resistance index, improving the detection accuracy and significantly reducing the false detection rate and missed detection rate. In addition, colposcopy and ultrasound testing are performed in the same process, and the targeted scanning time from colposcopy to ultrasound is shortened to less than 2 minutes, avoiding the repetitive actions of traditional "blind scanning" and improving detection efficiency.

[0026] Through the multimodal collaboration of "optical manifestation + deep function", accurate stratification and prediction of cervical lesions can be achieved, which is particularly suitable for the dual needs of rapid screening in primary hospitals and detailed evaluation in tertiary hospitals.

[0027] On the basis of the above embodiment scheme, the cervical surface is imaged by a colposcope to obtain a cervical surface image, including configuring a multispectral colposcope, imaging the cervical surface based on the multispectral colposcope, and synchronously collecting a white light reflection image, a narrow-band imaging vascular image, and a fluorescent staining image of the cervical surface.

[0028] In specific application scenarios, the hardware configuration of the multispectral colposcope includes: Three-channel optical system, integrating white light LED (400-700nm), narrow-band blue light (415nm) and green light (540nm) light sources, as well as fluorescence excitation module (such as 405nm excitation light + fluorescent dye).

[0029] Time-sharing / light-sharing synchronous acquisition: Time-sharing mode: Sequentially capture white light reflectance (RGB) images, narrow band imaging (NBI), and fluorescence images by switching the filter wheel at high speed (<100ms interval).

[0030] Spectroscopic mode: Using a three-in-one prism beam splitter, light of different wavelengths is simultaneously projected onto three sets of image sensors, achieving true real-time synchronous imaging. Synchronous imaging means acquiring trimodal data with a single exposure (traditional layered inspection takes 3-5 minutes, now reduced to within 30 seconds).

[0031] Fluorescent staining assistance: Spray fluorescent agents such as 5-aminolevulinic acid (5-ALA) before detection. Cancerous areas (such as high-metabolism cells) will show red fluorescence under blue light excitation.

[0032] The hardware trigger signal ensures the time alignment of the three images to avoid misalignment due to cervical movement.

[0033] White light imaging uses broad-spectrum white light to illuminate the surface of the cervix. The captured reflected images provide natural color anatomical images for observing surface features such as mucosal color, boundaries, and erosions.

[0034] Narrow-band imaging uses specific narrow-band spectra (415nm blue light and 540nm green light) to enhance the contrast of the microvascular and glandular structures on the surface of the mucosa. The 415nm blue light shows superficial capillaries, and the 540nm green light shows slightly deeper vascular networks.

[0035] Fluorescence staining imaging utilizes the selective binding properties of specific fluorescent dyes to cervical lesion tissue. Under the excitation of blue light or near-infrared light, the lesion area emits fluorescent signals, thereby providing metabolic activity information.

[0036] White light is sensitive to surface structures (such as erosions), narrow-band imaging (NBI) enhances the detection of vascular abnormalities, and fluorescence highlights metabolically active areas. Combined, these three methods enable a detection rate of greater than 95% for cervical intraepithelial neoplasia grade 2 or higher (CIN2+). Negative fluorescence can exclude false positives for inflammation.

[0037] Through the "structure-vascular-metabolism" trinity imaging, the problem of traditional colposcopy relying on subjective experience and missing tiny invasive lesions is solved. It is especially suitable for rapid screening and precision medicine scenarios in primary hospitals.

[0038] Based on the above embodiment scheme, refer to Figure 2 As shown, the cervical surface image is analyzed to locate the diseased area, including feature extraction of the white light reflection image to obtain surface texture features and edge features; feature extraction of the narrow-band imaging vascular image to obtain vascular density features and vascular morphology features; feature extraction of the fluorescent staining image to obtain fluorescence intensity distribution features; normalization and weighted fusion of the surface texture features, edge features, vascular density features, vascular morphology features and fluorescence intensity distribution features to obtain a fused feature map; threshold segmentation is performed on the fused feature map, and the diseased area is obtained according to the threshold segmentation result.

[0039] In specific application scenarios, the white light reflection image is preprocessed to adjust the brightness and contrast, and denoising is performed to reduce image noise. Subsequently, image analysis tools are used to identify the subtle texture of the cervical surface. For example, the grayscale co-occurrence matrix method is used to statistically analyze the relationship between different brightness levels in the image, thereby analyzing the detailed structure of the cervical surface tissue and obtaining texture features. Edge detection algorithms, such as the Canny algorithm or the Sobel algorithm, are then used to find boundaries in the image where brightness changes significantly and obtain edge features.

[0040] The NBI mode uses light of a specific wavelength to highlight superficial blood vessels, making the vascular network on the surface of the cervix more clearly visible. The acquired NBI image is first processed for vascular enhancement. Specific methods include using filtering algorithms (such as the Frangi vascular enhancement filter) to highlight the vascular region, suppress background tissue, and make the vascular structure more prominent. Image contrast and brightness are adjusted to further highlight small blood vessels. Next, the enhanced image is processed using an image segmentation algorithm to separate the blood vessels from the surrounding tissue. This method includes setting a pixel grayscale threshold, separating the blood vessels from the background based on threshold segmentation, and outlining the blood vessels using edge detection and region growing. After segmenting the vascular region, the total length, number of blood vessels per unit area, or the proportion of area occupied by blood vessels are calculated to obtain vascular density characteristics. Morphological analysis is then performed on the segmented vascular region, calculating morphological features such as the number of branches, branch angles, and curvature of the vessels. Abnormal morphologies such as uneven thickness and disordered orientation of the vessels are determined, and vascular morphological characteristics are derived based on the morphological analysis results.

[0041] Fluorescence images of the cervical surface are collected under specific excitation illumination, and the fluorescence images are denoised and brightness normalized to eliminate interference caused by uneven illumination or staining, making the image clearer and more uniform. A brightness threshold is set, and pixels with brightness higher than the threshold are classified as "abnormal areas", and the rest are normal areas. Areas with strong (bright) fluorescence signals are separated from the background. Within the segmented abnormal area, the following features are extracted to obtain fluorescence intensity distribution characteristics: Fluorescence average intensity: the average brightness of the entire lesion area; Maximum / minimum intensity: find the fluorescence intensity of the brightest and darkest points; Intensity distribution uniformity: count the amplitude of brightness changes in the area to see whether the fluorescence signal is evenly distributed; Area ratio: calculate the proportion of high-intensity areas to the entire cervical surface.

[0042] The value ranges and scales of different features (such as texture, edges, vascular density, vascular morphology, and fluorescence intensity) may be completely different. In order to enable unified comparison, normalization is required first: each feature value is converted to a value between 0 and 1 according to its maximum and minimum values. For example, if the original value range of a feature is 10 to 100, after normalization, 10 becomes 0, 100 becomes 1, and the remaining values ​​are proportionally mapped between 0 and 1.

[0043] Based on empirical data, each feature is assigned a weight (i.e., importance coefficient), with the sum of the weights being 1. For example, the weight of surface texture is 0.2, edge characteristics is 0.2, vascular density is 0.2, vascular morphology is 0.2, and fluorescence intensity is 0.2 (each accounting for 20%). Alternatively, based on actual needs, one feature can be weighted higher, such as 0.3 for fluorescence intensity and 0.175 for the others. Weights can be optimized through statistical analysis of a large number of cases or adjusted by the physician based on experience.

[0044] The normalized features are weighted and superimposed according to the corresponding weights. After the calculation is completed, a "fused feature map" of the same size as the original image is obtained. In this map, the higher the color or grayscale value, the more likely it is to be a diseased area after combining multiple features.

[0045] By normalizing and fusing the multimodal features of white light, narrowband imaging, and fluorescent staining images, the complementary information reflected by different imaging modalities on lesions can be fully utilized, significantly improving the accuracy of identifying and localizing lesions. The complementary features of different modalities effectively suppress interference and errors in a single modality, reducing missed detections and false detections caused by factors such as image noise and imaging conditions, thereby improving the reliability of cervical lesion detection.

[0046] Based on the above embodiment, the multispectral colposcope is configured, including an objective lens and a beam splitter prism group, wherein the objective lens converges the reflected light from the cervical surface and enters the beam splitter prism group; the beam splitter prism group divides the incident light into a first light path, a second light path, and a third light path according to wavelength bands; wherein an RGB image sensor is configured on the first light path, and a white light reflection image of the cervical surface is captured by the RGB image sensor; a CMOS image sensor is configured on the second light path, and a narrow-band imaging blood vessel image of the cervical surface is captured by the CMOS image sensor; and a solid-state image sensor is configured on the third light path, and a fluorescent staining image of the cervical surface is captured by the solid-state image sensor.

[0047] In specific application scenarios, refer to Figure 3 As shown, a high-quality objective lens assembly is installed at the front end of the colposcope to focus on the surface of the cervix, efficiently converging various reflected or emitted light, ensuring that the subsequent optical system can obtain a clear and bright image. After the objective lens, a multi-layer beam splitter prism group is configured. The beam splitter prism group can accurately separate multiple light paths according to the different wavelengths (colors) of light: The first light path: mainly used for the passage of visible light (RGB, red, green, and blue); Second optical path: Targets specific narrowband wavelength regions, such as the blue-green band used to highlight vascular structures; The third optical path is specifically used to separate the fluorescence signal based on the special excitation and emission bands used for fluorescent staining imaging.

[0048] A high-sensitivity RGB image sensor is deployed in the first optical path. This RGB sensor captures ordinary visible light images of the cervical surface, clearly recording the tissue structure and surface condition. A CMOS image sensor optimized for a specific wavelength is deployed in the second optical path. This captures images of enhanced blood vessels under narrowband imaging, highlighting the distribution and morphology of tiny blood vessels. A dedicated solid-state image sensor is deployed in the third optical path. This sensor is highly sensitive to fluorescent wavelengths (such as green or red fluorescence) and can efficiently capture fluorescent signals from diseased areas. Three sets of image sensors simultaneously capture three different types of image data from the same cervical region, transmitting them to the back-end analysis system via a high-speed data interface, enabling the simultaneous acquisition and processing of multimodal images of the same area.

[0049] A beam splitter prism separates light from different wavelengths at the same location, sending them to separate sensors. This allows for simultaneous acquisition of white light, vascular, and fluorescence images, eliminating the need for repeated switching of light sources or lenses. This significantly improves image acquisition speed and detection efficiency. The three sets of sensors are simultaneously focused on the same area, allowing the resulting multimodal images to be naturally registered (aligned). This ensures precise information overlay during subsequent image fusion analysis, avoiding misjudgments due to image misalignment and improving the accuracy of lesion localization and diagnosis.

[0050] On the basis of the above embodiment, the position parameters of the ultrasound probe are configured according to the position of the lesion area. Figure 4 As shown, it includes extracting the two-dimensional pixel coordinates of the lesion area in the image coordinate system; converting the two-dimensional pixel coordinates into three-dimensional spatial coordinates in a spatial rectangular coordinate system according to the imaging geometric parameters of the optical system of the colposcope; and obtaining the posture parameters of the ultrasound probe according to the end coordinates of the ultrasound probe and the three-dimensional spatial coordinates.

[0051] In specific application scenarios, an optical tracking system and a robotic arm system are deployed. The optical tracking system uses an infrared optical positioning device (such as NDI Polaris) and installs reflective marker balls on the colposcope body and ultrasound probe to capture their spatial position and posture in real time. The robotic arm system uses a six-degree-of-freedom collaborative robot (such as the UR5e) with an ultrasound probe at the end. The colposcope imaging system is equipped with a high-resolution digital camera. The lens must be pre-calibrated through a checkerboard grid to obtain the intrinsic parameter matrix, including focal length, principal point coordinates, and distortion coefficients. The ultrasound probe uses a high-frequency linear array probe with an optical marker ball fixed to the probe handle to ensure that the geometric relationship between the marker point and the probe scanning plane is known.

[0052] Convert 2D pixel coordinates to 3D space coordinates: Image coordinate system: The upper left corner of the image is the origin, the u axis points to the right, and the v axis points downward; Camera coordinate system: The optical center of the colposcope lens is the origin, and the Z axis is along the optical axis; World coordinate system: The global coordinate system defined by the optical tracking system.

[0053] The largest connected domain is extracted from the binary mask generated by the fused feature map, and the two-dimensional pixel coordinates (u, v) of the geometric center of this connected domain are calculated. Using the colposcope's internal parameters (including focal length, principal point coordinates, and distortion coefficients), the pixel coordinates (u, v) are converted into three-dimensional ray directions in the camera coordinate system. Combined with the colposcope's external parameters (rotation matrix and translation vector) provided by the optical tracking system, the ray from the camera coordinate system is transformed into the world coordinate system. Based on a preset initial lesion depth (typically set to 15 mm), the three-dimensional coordinates (X, Y, Z) of the lesion center in the world coordinate system are calculated. Based on the ultrasound probe's scanning plane geometry (given the fixed relationship between the probe's markers and the scanning plane), the probe's desired position and attitude are calculated so that its scanning plane precisely passes through the lesion's center. The attitude angles (θx, θy, θz) are calculated using spatial geometric relationships to ensure alignment between the probe's scanning plane and the lesion's normal vector.

[0054] By precisely locating the lesion area on the vaginal image in real space, the ultrasound probe can be precisely aligned and scanned, significantly improving the accuracy of detecting lesion infiltration depth and blood flow parameters. The deep integration of colposcopy and ultrasound imaging enables comprehensive, one-stop testing from surface images to internal structures, providing clinicians with more comprehensive and objective information on cervical lesions and optimizing the screening and diagnosis process.

[0055] Based on the above embodiment, the ultrasound image is analyzed to obtain the tissue infiltration depth of the lesion area, including extracting the boundary of the cervical stromal layer using an image segmentation algorithm, setting multiple measurement points evenly spaced within the projection range of the lesion area, calculating the vertical distance from each measurement point to the nearest stromal layer boundary, and taking the maximum value of the distances of all measurement points as the tissue infiltration depth.

[0056] In specific application scenarios, before analyzing tissue invasion depth, ultrasound images are preprocessed. This includes using methods such as Gaussian filtering and mean filtering to remove noise from the image to prevent it from interfering with boundary extraction. Histogram equalization or CLAHE (contrast-limited adaptive histogram equalization) is then used to enhance image contrast, particularly between the stroma and the lesion. A U-Net network is then used to segment the preprocessed ultrasound images, generating segmentation maps for the epithelium, stroma, and lesion. The lesion area was determined to be within the projection of the stromal boundary. A minimum circumscribed rectangular ROI (with the long side oriented along the cervical axis) was generated. Measurement lines were set every 0.5 mm along the long side of the ROI (corresponding to a 10-pixel spacing in the ultrasound image), and a measurement point was taken every 0.3 mm on each measurement line (6-pixel spacing). For example, a 5 mm × 3 mm lesion area → 11 measurement lines × 11 points = 121 measurement points. For each measurement point, a search was performed along the normal direction for the stromal boundary (the location where the grayscale value suddenly dropped by >30 HU). The Bresenham algorithm was used for fast linear traversal. The vertical distances {d1, d2, ..., dn} of all measurement points were recorded. The final invasion depth D = max(di) × calibration factor (probe frequency-dependent, = 0.98 at 10 MHz).

[0057] On the basis of the above embodiment scheme, the ultrasound image is analyzed to obtain the blood flow resistance index of the lesion area, including performing color Doppler blood flow imaging processing on the ultrasound image to identify all blood flow signals; calculating the peak systolic velocity PSV and the end-diastolic velocity EDV of each blood flow signal; calculating the resistance index RI of each blood flow signal according to the following method: RI=(PSV-EDV) / PSV; and taking the average value of the resistance index of all blood flow signals as the blood flow resistance index.

[0058] In specific application scenarios, ultrasound images are processed using Color Doppler Flow Mapping (CFM). CFM generates a color representation of blood flow within the image based on the speed and direction of blood flow. During this process, the speed and direction of blood flow are encoded as different colors (for example, red indicates blood flow toward the probe, blue indicates blood flow away from the probe), making the blood flow signal more distinct in the image. Furthermore, the colored regions within the image are identified and the blood flow signal is separated. Blood flow signals typically occur within blood vessels, distinguishing them from surrounding static tissue. For each identified blood flow signal, its time-velocity curve is further analyzed, and two key points of the flow velocity signal are calculated: Peak Systolic Velocity (PSV) and End Diastolic Velocity (EDV). Peak Systolic Velocity (PSV) represents the highest blood flow velocity during cardiac systole. By detecting the blood flow velocity waveform, the highest point is automatically identified. End Diastolic Velocity (EDV) represents the blood flow velocity at the end of diastole, identifying the end of diastole and obtaining the velocity value at that moment. For each identified blood flow signal, the resistance index (RI) is calculated based on RI = (PSV - EDV) / PSV. This index reflects the resistance characteristics of blood flow. Areas with high blood flow resistance are often associated with diseased areas (such as tumors, vascular abnormalities, etc.). The resistance index (RI) is calculated for all detected blood flow signals, and the average resistance index of all signals is taken. This average is the blood flow resistance index of the diseased area and is used to evaluate the hemodynamic characteristics of that area. By accurately calculating PSV, EDV, and RI, the resistance characteristics of blood flow can be quantitatively analyzed. Color Doppler images are generated in real time, and changes in blood flow velocity and resistance index can be dynamically observed, allowing doctors to obtain blood flow information immediately during actual operations and quickly determine the progression of the disease.

[0059] Based on the above embodiment scheme, identifying the blood flow signal includes extracting an initial blood flow signal pixel set through color Doppler blood flow imaging processing; binarizing the initial blood flow signal pixel set to obtain an initial blood flow signal binary image; performing noise reduction processing on the initial blood flow signal binary image to obtain a noise-reduced binary image; performing region growing processing on the noise-reduced binary image, and merging adjacent blood flow signals to obtain a set of blood flow connected areas; calculating the pixel area of ​​each blood flow connected area, and screening the blood flow connected areas with pixel areas greater than or equal to a threshold pixel as valid blood flow areas; identifying blood flow signals corresponding to each of the valid blood flow areas to obtain all the blood flow signals.

[0060] In specific application scenarios, color Doppler technology is used to process ultrasound images to generate color-coded images of blood flow velocity and direction. The color of each pixel represents the velocity and direction of blood flow, with red and blue usually used to represent blood flow toward and away from the probe, respectively. By setting an appropriate threshold, pixels in all blood flow signal areas are extracted. Blood flow signals are usually bright areas that represent the dynamic changes of blood flow. The set of these pixels is the initial blood flow signal pixel set. Setting an appropriate threshold, pixels below this threshold are classified as background, and pixels above this threshold are classified as foreground (blood flow signals). This can effectively reduce interference from low-signal areas, making the blood flow signal more prominent, and obtaining a preliminary blood flow signal binary image. Since there may be interference such as noise and false signals in the image, the binarized blood flow signal image needs to be denoised to remove background noise and isolated noise points. Region growing is performed on the denoised binary image to further extract blood flow signal regions and merge adjacent blood flow signal regions to form complete blood flow connected regions. Region growing here involves gradually expanding similar pixels (such as pixels with similar color or grayscale) from one or more seed points to the surrounding area until a set similarity threshold or other stopping condition is met. Through region growing, connected regions of blood flow signals can be effectively identified and scattered small regions can be merged into a large region to ensure the integrity of the blood flow signal. For each blood flow connected region, the number of pixels it contains (i.e., the pixel area of ​​the region) is calculated. This area represents the size of the blood flow region and further helps to screen valid blood flow signal regions. Based on the pixel area of ​​the blood flow connected region, a threshold is set, and only blood flow connected regions with an area greater than or equal to the threshold are screened as valid blood flow regions. For each screened valid blood flow region, its blood flow signal characteristics are further calculated, and these blood flow signals are identified as the final valid blood flow signals.

[0061] Color Doppler blood flow imaging can extract all preliminary blood flow signal pixels from ultrasound images. This process ensures that all blood flow signal regions can be identified and provides basic data for subsequent processing steps. The region growing method processes the de-noised binary image, merging adjacent blood flow signal regions to obtain a set of connected blood flow regions. This step ensures complete identification of blood flow signal regions, avoiding missed or false detections due to irregular blood flow signal distribution. In particular, it effectively addresses the discontinuity or dispersion of blood flow signals in the image.

[0062] Based on the above embodiment, cervical lesions are predicted based on the tissue infiltration depth and blood flow resistance index, including performing logarithmic transformation on the tissue infiltration depth to obtain a tissue infiltration characteristic index; performing Z-score normalization transformation on the blood flow resistance index to obtain a blood flow resistance index; and obtaining a cervical lesion risk index S based on the following method: S=0.6×ln(D+1)+0.4×(1-Z); Where D represents the depth of tissue infiltration; Z represents the normalized result of the blood flow resistance index Z-score.

[0063] The tissue invasion depth D is usually a continuous value that represents the depth of the lesion area. Since the tissue invasion depth has a skewed distribution, logarithmic transformation can make the data more consistent with the normal distribution and more suitable for subsequent analysis. Adding 1 to the invasion depth D and performing a natural logarithmic transformation can avoid problems caused by zero and negative values. This can effectively compress large values ​​and amplify small values, facilitating unified analysis of characteristics across different depth ranges. The blood flow resistance index RI reflects the degree of blood flow abnormality in the lesion area. Considering that the blood flow resistance index may vary significantly between different patients, the blood flow resistance index is standardized using the Z-score to ensure its effectiveness in the prediction model.

[0064] Z-score normalization formula: ; represents the mean; represents the standard deviation; through standardization, the blood flow resistance index will be converted to a distribution with a mean of 0 and a standard deviation of 1, making it comparable between different cases.

[0065] Example 2: Based on the same inventive concept, the embodiment of the present invention provides a cervical lesion prediction device based on image processing, which is used to implement the cervical lesion prediction method based on image processing. Figure 5 As shown, the device includes, A colposcope module, used for imaging the surface of the cervix and obtaining an image of the cervix surface; An image analysis module, configured to analyze the cervical surface image and locate the lesion area; A posture adjuster, equipped with an ultrasound probe, is used to adjust the posture parameters of the ultrasound probe so that the ultrasound probe is aligned with the lesion area; an ultrasonic probe, used for scanning the lesion area and obtaining an ultrasonic image of the lesion area; an ultrasound analysis module, configured to analyze the ultrasound image to obtain the tissue infiltration depth and blood flow resistance index of the lesion area; A lesion prediction unit is used to predict cervical lesions based on the tissue infiltration depth and blood flow resistance index.

[0066] The embodiment of the present invention and the first embodiment are based on the same inventive concept and have the same technical effects, which will not be described in detail here.

[0067] The image processing-based cervical lesion prediction method and device described in the present invention combine colposcopy and ultrasound to detect cervical lesions in the same process, thereby improving detection accuracy and efficiency.

[0068] Among them, the lesion area on the surface of the cervix is ​​located through colposcopy imaging, the position of the lesion area is used as the scanning object to align the position parameters of the ultrasound probe, and the lesion area is scanned to obtain an ultrasound image containing the deep structure information of the lesion area. The tissue infiltration depth and blood flow resistance index are combined to evaluate cervical lesions, improve detection accuracy, and significantly reduce the false detection rate and missed detection rate. In addition, colposcopy and ultrasound detection are carried out in the same process, and the targeted scanning time from colposcopy to ultrasound is shortened to less than 2 minutes, avoiding the repetitive actions of traditional "blind scanning" and improving detection efficiency.

[0069] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0070] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a box or multiple boxes.

[0071] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0072] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0073] Obviously, the above embodiments are merely examples for clarity of explanation and are not intended to limit the implementation methods. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all implementation methods here. Obvious variations or modifications arising therefrom remain within the scope of protection of the present invention.

Claims

1. A method for predicting cervical lesions based on image processing, characterized by: include, The cervical surface is imaged through a colposcope to obtain an image of the cervical surface; Analyzing the cervical surface image to locate the lesion area; Configuring the position parameters of the ultrasound probe according to the position of the lesion area so that the ultrasound probe is aligned with the lesion area for scanning to obtain an ultrasound image; Analyzing the ultrasound image to obtain the tissue infiltration depth and blood flow resistance index of the lesion area; Cervical lesions are predicted according to the tissue infiltration depth and blood flow resistance index.

2. The method for predicting cervical lesions based on image processing according to claim 1, characterized in that: The cervical surface is imaged by a colposcope to obtain a cervical surface image, including configuring a multispectral colposcope, imaging the cervical surface based on the multispectral colposcope, and synchronously collecting a white light reflection image, a narrow-band imaging blood vessel image, and a fluorescent staining image of the cervical surface.

3. The method for predicting cervical lesions based on image processing according to claim 2, characterized in that: Analyzing the cervical surface image to locate the lesion area includes: performing feature extraction on the white light reflection image to obtain surface texture features and edge features; performing feature extraction on the narrow-band imaging blood vessel image to obtain blood vessel density features and blood vessel morphology features; performing feature extraction on the fluorescent staining image to obtain fluorescence intensity distribution characteristics; Normalizing the surface texture features, edge features, blood vessel density features, blood vessel morphology features, and fluorescence intensity distribution features and performing weighted fusion to obtain a fusion feature map; Perform threshold segmentation on the fused feature map, and obtain the lesion area according to the threshold segmentation result.

4. The method for predicting cervical lesions based on image processing according to claim 2 or 3, characterized in that: Configuring the multispectral colposcope includes: An objective lens and a beam splitter prism group are configured. The objective lens converges the reflected light from the cervical surface into the beam splitter prism group; the beam splitter prism group divides the incident light into a first light path, a second light path, and a third light path according to wavelength bands; wherein, An RGB image sensor is configured on the first light path, and a white light reflection image of the cervical surface is collected by the RGB image sensor; A CMOS image sensor is configured on the second optical path to capture a narrow-band imaging blood vessel image of the cervical surface through the CMOS image sensor; A solid-state image sensor is disposed on the third optical path, and a fluorescent staining image of the cervical surface is collected by the solid-state image sensor.

5. The method for predicting cervical lesions based on image processing according to claim 1, characterized in that: The ultrasound probe posture parameters are configured according to the position of the lesion area, including: Extracting the two-dimensional pixel coordinates of the lesion area in the image coordinate system; converting the two-dimensional pixel coordinates into three-dimensional spatial coordinates in a rectangular coordinate system according to the imaging geometric parameters of the optical system of the colposcope; The position parameters of the ultrasonic probe are obtained according to the end coordinates of the ultrasonic probe and the three-dimensional space coordinates.

6. The method for predicting cervical lesions based on image processing according to claim 1, characterized in that: Analyze the ultrasound image to obtain the tissue invasion depth of the lesion area, including extracting the boundary of the cervical stromal layer using an image segmentation algorithm, setting multiple measurement points evenly spaced within the projection range of the lesion area, calculating the vertical distance from each measurement point to the nearest stromal layer boundary, and taking the maximum value of the distance of all measurement points as the tissue invasion depth.

7. The method for predicting cervical lesions based on image processing according to claim 1 or 6, characterized in that: Analyzing the ultrasound image to obtain the blood flow resistance index of the lesion area includes: performing color Doppler blood flow imaging processing on the ultrasound image to identify all blood flow signals; Calculating the peak systolic velocity PSV and the end-diastolic velocity EDV of each of the blood flow signals; The resistance index RI of each blood flow signal was calculated as follows: RI=(PSV-EDV) / PSV The average value of the resistance indexes of all blood flow signals is taken as the blood flow resistance index.

8. The method for predicting cervical lesions based on image processing according to claim 7, characterized in that: Identifying the blood flow signal includes, Extracting an initial blood flow signal pixel set by color Doppler blood flow imaging processing; performing binarization processing on the initial blood flow signal pixel set to obtain an initial blood flow signal binary image; Performing noise reduction processing on the initial blood flow signal binary image to obtain a noise-reduced binary image; performing region growing processing on the binary image after noise reduction and merging adjacent blood flow signals to obtain a set of blood flow connected regions; Calculating the pixel area of ​​each blood flow connected region, and screening the blood flow connected regions whose pixel area is greater than or equal to a threshold pixel as valid blood flow regions; A blood flow signal is identified corresponding to each effective blood flow area to obtain the entire blood flow signal.

9. The method for predicting cervical lesions based on image processing according to claim 1, characterized in that: Predicting cervical lesions according to the tissue infiltration depth and blood flow resistance index includes: performing logarithmic transformation on the tissue infiltration depth to obtain a tissue infiltration characteristic index; Performing Z-score standardization conversion on the blood flow resistance index to obtain a blood flow resistance index; The cervical lesion risk index S is obtained based on the following method: S=0.6×ln(D+1)+0.4×(1-Z); Where D represents the depth of tissue infiltration; Z represents the normalized result of the blood flow resistance index Z-score.

10. An image processing-based cervical lesion prediction device, configured to implement the image processing-based cervical lesion prediction method according to any one of claims 1 to 9, characterized in that: include, A colposcope module, used for imaging the surface of the cervix and obtaining an image of the cervix surface; An image analysis module, configured to analyze the cervical surface image and locate the lesion area; A posture adjuster, equipped with an ultrasound probe, is used to adjust the posture parameters of the ultrasound probe so that the ultrasound probe is aligned with the lesion area; an ultrasonic probe, used for scanning the lesion area and obtaining an ultrasonic image of the lesion area; an ultrasound analysis module, configured to analyze the ultrasound image to obtain the tissue infiltration depth and blood flow resistance index of the lesion area; A lesion prediction unit is used to predict cervical lesions based on the tissue infiltration depth and blood flow resistance index.

Images