Search

CN-122004940-A - UBM image quantitative analysis method and system for silicone oil emulsification evaluation

CN122004940ACN 122004940 ACN122004940 ACN 122004940ACN-122004940-A

Abstract

The invention discloses a UBM image quantitative analysis method and system for silicone oil emulsification evaluation, belongs to the field of medical image processing, and aims to objectively quantify emulsification-related acoustic signs in UBM images after silicone oil filling operation. The method comprises the steps of communicating with UBM equipment, monitoring the perpendicularity of a gain and a probe, outputting a control/prompt instruction to perform standardized acquisition, desensitizing an image, masking a non-scanning area, extracting characteristics such as floating particle density, artifact intensity, high echo particle aggregation area occupation ratio, tissue infiltration signal-to-noise ratio, cornea endothelial post-deposition density and the like, fusing the characteristics of a center and each azimuth, introducing a space weight matrix to calculate weighted acoustic distribution characteristic values ADCV, and outputting ADCV and a structured quantification result for clinical evaluation and follow-up comparison.

Inventors

  • ZHAO HONGMEI
  • JIANG CHUNHUI
  • WEI JIAOJIAO

Assignees

  • 复旦大学附属眼耳鼻喉科医院

Dates

Publication Date
20260512
Application Date
20260130

Claims (10)

  1. 1. A UBM image quantitative analysis method for silicone oil emulsion evaluation, comprising the steps of: S0, image acquisition standardized closed-loop control, wherein the image acquisition standardized closed-loop control is communicated with Ultrasonic Biological Microscope (UBM) equipment to acquire acquisition parameters and/or real-time image quality indexes in real time, the acquisition parameters at least comprise gain values G, the quality indexes at least comprise cornea echo line widths W, and when the acquisition parameters and/or the quality indexes do not meet preset standards, control instructions and/or interaction prompts are generated to restrict an acquisition process, so that a target image meeting standardized requirements is obtained; s1, image preprocessing, namely performing data desensitization on the target image, and performing mask processing on a non-effective scanning area to obtain an intraocular scanning area limited by an effective scanning area mask, wherein the effective scanning area mask is generated and applied by performing threshold segmentation on the image and selecting a maximum connected area; S2, extracting multi-dimensional acoustic features, namely performing morphological operation, spot detection, threshold segmentation, texture analysis and/or signal-to-noise ratio calculation on the preprocessed image, and extracting continuous physical quantization features for quantizing medium distribution; S3, space weighted quantization evaluation, namely normalizing and fusing the continuous physical quantization characteristics to obtain a central region characteristic value P_center and each azimuth characteristic value P_i, introducing a space weight matrix to distribute weight coefficients w_i for different azimuths and calculating weighted acoustic distribution characteristic values ADCV, wherein the P_center and the P_i are characteristic values obtained by fusing the normalized continuous physical quantization characteristics; And S4, outputting a result, namely outputting ADCV and/or a structural quantization result associated with the ADCV.
  2. 2. The method according to claim 1, wherein in step S0, the determining whether the image quality meets the preset criteria based on the acquisition parameters specifically includes: (1) When G <79dB or G >81dB, generating a locking instruction to suspend image freezing and saving, and generating an interaction prompt to guide G to be adjusted to a preset interval; And/or the number of the groups of groups, (2) Performing edge detection on each frame of image of the acquired video stream in the central area of the image to extract a cornea echo line, acquiring a gray level section along the normal direction of the cornea echo line and fitting the gray level section to calculate full width at half maximum (FWHM) to obtain the width W (unit pixel) of the cornea echo line, judging whether the W is smaller than or equal to a 3-pixel threshold value, and generating a probe adjustment prompt when the W average value of the continuous preset frame number exceeds the threshold value.
  3. 3. The method according to claim 1, wherein the control instructions in step S0 at least comprise prohibiting freezing/saving of a current frame, sending a lock instruction to a device to suspend freezing and saving, unlocking and allowing saving when the acquisition parameters and/or the quality index resume meeting a preset criterion.
  4. 4. The method of claim 1, wherein the target image comprises 1 wide view image and 8 narrow view images of different orientations, the orientations comprising at least 1:30, 3:00, 4:30, 6:00, 7:30, 9:00, 10:30, 12:00.
  5. 5. The method of claim 1, wherein the continuous physical quantification feature comprises at least one of anterior chamber planktonic particle density, intensity of multiple reflection artifacts after corneal endothelium, area duty cycle of high echogenic particle accumulation region, tissue infiltration signal to noise ratio, and density of deposit after corneal endothelium, wherein the area duty cycle of high echogenic particle accumulation region is the area duty cycle of high echogenic foreground connected region in the intraocular scan region, which is the intraocular effective scan region defined by the effective scan region mask in step S1.
  6. 6. The method of claim 1 or 5, wherein the extraction of anterior chamber planktonic particle density comprises: (1) Determining an anterior chamber region of interest (ROI) in the UBM image, and acquiring a gray matrix I (x, y) in the ROI, wherein an upper boundary of the ROI comprises a corneal endothelial boundary curve, a lower boundary of the ROI preferably comprises an iris anterior surface boundary curve, and the lower boundary can be further complemented or corrected by using the lens anterior capsule upper edge boundary curve when the lens anterior capsule upper edge can be reliably identified; (2) Performing morphological top hat transformation on I (x, y) to obtain a particle-enhanced image I_th (x, y), wherein I_th (x, y) =I (x, y) -Open (I (x, y), and Open (·, B) represents morphological Open operation performed by a structural element B, and B is a circular structural element; (3) Performing laplace-gaussian LoG convolution on i_th (x, y) to obtain a response graph r_σ (x, y) = ∇ 2 (g_σ×i_th) (x, y), and calculating the response graph for at least two different scale parameters σ by adopting a multi-scale detection strategy; (4) Selecting a local extremum point meeting a preset gray threshold value from the response graph as a candidate particle center, and generating a candidate particle connected domain set on I_th (x, y) by taking the candidate particle center as a seed; (5) Calculating the area A_k and the perimeter P_k of each candidate particle connected domain, and calculating the shape roundness according to the roundness C_k=4pi A_k/P_k2; (6) Counting the number N of connected domains corresponding to the screened target particle set, converting the pixel area of the ROI into a real area A_mm2 according to a pixel interval label (0028,0030) of the DICOM header file, and calculating the Density of floating particles=N/A_mm2, wherein the Density is expressed in units of one per square millimeter.
  7. 7. The method of claim 6, wherein the scale parameter σ of the multi-scale detection in step (3) includes at least 1.5 pixels, 2.5 pixels, and 3.5 pixels, and a maximum value of each scale response is taken as a final spot intensity of each pixel point.
  8. 8. The method of claim 1, wherein quantifying the intensity of the post-corneal endothelial multiple reflection artifact comprises: (1) The artifact analysis area is determined, namely, a cornea endothelial boundary extracted by cornea echo rays in the UBM image is taken as a reference, a preset physical distance is shifted along the intraocular direction, and pixel displacement is converted according to pixel spacing, so that a banded area is obtained as the artifact analysis area; (2) Texture enhancement, namely performing vertical Sobel gradient operator convolution on the gray level image in the artifact analysis area to obtain a texture enhancement image; (3) Texture quantization, namely constructing a gray level co-occurrence matrix GLCM in an artifact analysis area of a texture enhancement map, setting a pixel pair distance d and a direction angle theta, and calculating a Contrast characteristic value according to contrast=Σ (i-j) 2.P (i, j); (4) And outputting the contrast characteristic value as an Artifact intensity quantization index artifact_strength.
  9. 9. The method of claim 1, wherein the spatially weighted quantization evaluation comprises: (1) Normalizing and fusing the continuous physical quantization characteristics to obtain a central region characteristic value P_center and each azimuth characteristic value P_i; (2) Constructing a space weight matrix, dividing the eyeball position into 8 positions, wherein the 8 positions respectively correspond to the 8 narrow-view images, and distributing weight coefficients higher than those of other positions for a lower position set S_inf, wherein the S_inf at least comprises 6:00 positions, and preferably further comprises 4:30 and/or 7:30 positions adjacent to the S_inf; (3) The weighted acoustic distribution eigenvalues ADCV are calculated as follows: ADCV = 100×( α·P_center + β·( Σ(w_i·P_i) / Σw_i ) ) wherein w_i is the ith azimuth weight coefficient, α and β are preset harmonic coefficients and α+β=1.
  10. 10. A UBM image quantitative analysis system for silicone oil emulsion evaluation, comprising: the acquisition control and monitoring module is used for executing the standardized closed-loop control of the image acquisition in the step S0 in the claim 1; An image preprocessing module for performing the image desensitization and masking process of step S1 of claim 1; a feature extraction module for performing the multi-dimensional acoustic feature extraction of step S2 of claim 1; A spatial weighted quantization module for performing the spatial weighted quantization evaluation of step S3 of claim 1 and calculating ADCV; An output module, configured to perform the output of the result of step S4 in claim 1.

Description

UBM image quantitative analysis method and system for silicone oil emulsification evaluation Technical Field The invention relates to the technical field of medical image processing and computer-aided diagnosis, in particular to a UBM image quantitative analysis method and system for silicone oil emulsification evaluation. Background Ultrasonic Biological Microscopy (UBM) is a key imaging tool for observing anterior ocular segment fine structures, and has important application value in evaluation after intraocular silicone oil filling. As a long-acting filler commonly used in vitreoretinal surgery, silicone oil may emulsify to form micron-sized droplets in the eye for a long period of time and may induce complications such as secondary glaucoma, corneal endothelial decompensation, etc., so that a repeatable and comparable evaluation of the degree of emulsification of silicone oil is required to assist in clinically determining the timing of intervention and follow-up strategies. The prior UBM evaluation method for silicone oil emulsification is roughly subjected to two paths, namely subjective qualitative evaluation, wherein the subjective qualitative evaluation mainly depends on a doctor to visually observe UBM images and judges according to experience descriptions such as punctiform echo, pine needle-like artifact and the like, and has obvious subjectivity and is difficult to compare across operators and time. Secondly, a semi-quantitative scoring system is provided, in order to reduce subjectivity, 0/1 discrete scoring based on various emulsifying signs is provided and added up to obtain a total score (for example, the total score is formed by adding up a plurality of sign items of a horizontal wide view and infiltration sign items of an 8-azimuth narrow view), and a double-blind interpretation and bifurcation arbitration flow is often adopted to obtain a final score. Although the semi-quantitative system described above is improved over pure subjective assessment, the following technical bottlenecks still exist: (1) The quantification precision is limited, the existing scores are mostly 'on/off' or discrete scores in a grading manner, and small changes of continuous physical quantities such as particle density, artifact strength, area ratio of a hyperechoic particle aggregation area and the like are difficult to reflect, so that sensitivity to disease progression or curative effect monitoring is insufficient. (2) The acquisition process lacks automation constraint, namely the real-time monitoring and closed-loop control on key acquisition conditions such as UBM equipment gain, probe angle/verticality and the like are lacking, so that the image quality is greatly influenced by operators, and the comparability of data from different sources is poor. (3) The clinical relevance is insufficient, the existing scoring model usually processes weights in different directions, and spatial distribution differences of emulsified silicone oil under the influence of gravity are not explicitly considered, so that the relevance of results and real clinical risks is insufficient. (4) The existing flow generally does not integrate privacy protection preprocessing links such as medical image data desensitization and the like, and the compliance requirements of actual data transfer and application are difficult to meet. (5) The integrated automatic quantification scheme for UBM silicone oil emulsification specificity is lacking, and the general image processing method does not form a closed loop flow from acquisition source quality consistency guarantee to multidimensional feature extraction and comprehensive quantification aiming at gravity dependency distribution, characteristic vertical artifacts, tiny hyperechoic spots and other symptoms. Therefore, there is a need for an evaluation method and system that can objectively, continuously, comparatively quantify the acoustic signs related to silicone oil emulsification, reflect the azimuth risk difference, and meet the data compliance requirements, while ensuring the consistency of the acquisition quality. Disclosure of Invention The invention aims to solve the technical problem of how to convert the emulsification-related acoustic sign from manual discrete scoring to repeatable continuous physical quantification results and enable the quantification results to reflect clinical risk differences of different anatomical orientations (especially lower orientations) on the premise of ensuring consistency of acquisition conditions and data desensitization compliance in UBM examination after silicone oil filling. The technical scheme adopted for solving the technical problems is as follows: a UBM image quantitative analysis method for silicone oil emulsion evaluation, comprising the steps of: S0, image acquisition standardized closed-loop control, namely communicating with Ultrasonic Biological Microscope (UBM) equipment, and acquiring acquisition parameters and/or image quality