Search

CN-122024989-A - Medical image quality control report generation method

CN122024989ACN 122024989 ACN122024989 ACN 122024989ACN-122024989-A

Abstract

The invention relates to the technical field of medical image detection and discloses a medical image quality control report generation method which comprises the steps of firstly carrying out soft tissue background field correction on a proximal femur scanning image to remove low-frequency interference, determining the stress conduction direction of a main pressure-bearing bone trabecula based on an anatomical geometric position, then converting the image into a spatial frequency domain, respectively extracting a structural frequency spectrum and a full-angle average structural frequency spectrum along the stress conduction direction, calculating effective texture cut-off frequencies of the structural frequency spectrum and the full-angle average structural frequency spectrum by using a breakpoint regression model, obtaining a quantization index reflecting the structural anisotropy retention degree by calculating the logarithmic ratio of the two cut-off frequencies, and comparing the quantization index with an adaptive threshold constructed based on the number of frequency points to generate a quality control report indicating whether the image quality is qualified or not.

Inventors

  • TANG CHUNQIN

Assignees

  • 南京中医药大学第二附属医院(江苏省第二中医院)

Dates

Publication Date
20260512
Application Date
20260128

Claims (9)

  1. 1. A method for generating a medical image quality control report, comprising: Acquiring a dual-energy X-ray proximal femur scanning image to be analyzed and a corresponding proximal femur anatomy segmentation mask, and performing soft tissue background field correction processing on image data in the proximal femur anatomy segmentation mask to obtain a background bone-removed microstructure distribution map; determining the stress conduction direction of the trabecula of the main bearing bone based on the geometrical relative position of the femoral head anatomical center and the femoral distance bearing region in the background bone microstructure distribution map; Performing spatial frequency domain transformation on the background-removed bone microstructure distribution map to obtain two-dimensional bone texture frequency domain feature distribution, and respectively extracting a structural spectrum and an all-angle average structural spectrum along the stress transmission direction of the main bearing bone trabecula based on the two-dimensional bone texture frequency domain feature distribution; Respectively calculating effective texture cut-off frequency of the stress direction corresponding to the structure spectrum and average effective texture cut-off frequency corresponding to the full-angle average structure spectrum by using a piecewise linear breakpoint fitting model; Calculating the logarithmic ratio of the effective grain cut-off frequency of the stress direction to the average effective grain cut-off frequency to obtain the anisotropic retention index of the main bearing bone trabecular structure; And comparing the anisotropic retention index of the main bearing bone trabecular structure with a preset minimum identifiable structure threshold value, and generating a quality control report indicating whether the image quality is qualified or unqualified according to a comparison result.
  2. 2. The method for generating a medical image quality control report according to claim 1, wherein obtaining a dual-energy X-ray proximal femur scan image to be analyzed and a corresponding proximal femur anatomy segmentation mask comprises: The method comprises the steps of obtaining gray matrix data and pixel interval attributes of a dual-energy X-ray near-end femur scanning image to be analyzed, obtaining a near-end femur anatomy segmentation mask which is aligned with the space of the dual-energy X-ray near-end femur scanning image, carrying out pixel-level multiplication operation on the dual-energy X-ray near-end femur scanning image and the near-end femur anatomy segmentation mask, retaining gray information in a mask area, and constructing to-be-processed image data only comprising the near-end femur area.
  3. 3. The method for generating a medical image quality control report according to claim 1, wherein performing soft tissue background field correction processing on the image data in the proximal femur anatomy segmentation mask to obtain a background bone-removed microstructure distribution map comprises: The method comprises the steps of establishing a three-dimensional polynomial curved surface model based on pixel space coordinates and used for fitting a low-frequency background field caused by soft tissue thickness change, solving coefficients of the three-dimensional polynomial curved surface model based on gray values of all effective pixel points in the near-end femur anatomy segmentation mask by using a least square method to obtain an estimated background field, subtracting the estimated background field from image data in the near-end femur anatomy segmentation mask, and obtaining a background bone microstructure distribution map after removing low-frequency offset.
  4. 4. The method of claim 1, wherein determining the main bearing bone trabecular stress conduction direction based on the geometric relative position of the femoral head anatomical center and the femoral distance bearing region in the background bone removal microstructure distribution map comprises: the method comprises the steps of calculating the gray level gravity center of a background bone microstructure distribution diagram by using a gray level weighted centroid algorithm, taking the gray level gravity center as a femoral head anatomical center, searching a pixel point with a maximum gray level value in a sector area with a preset fixed angle relative to the femoral head anatomical center, taking the pixel point as a maximum gray level response point of a femoral distance bearing area, calculating a connecting line vector pointing to the maximum gray level response point of the femoral distance bearing area from the femoral head anatomical center, and taking the angle of the connecting line vector under a polar coordinate system as the stress conduction direction of a main bearing bone trabecula.
  5. 5. The method of claim 1, wherein performing a spatial-frequency-domain transformation on the background-removed bone microstructure distribution map to obtain a two-dimensional bone texture frequency-domain feature distribution comprises: The method comprises the steps of obtaining a background bone microstructure distribution map, applying a two-dimensional smooth window function to the background bone microstructure distribution map to inhibit spectrum leakage effect caused by image boundary truncation to obtain a windowed bone microstructure distribution map, performing two-dimensional discrete Fourier transform on the windowed bone microstructure distribution map to obtain a complex frequency spectrum containing amplitude and phase, calculating the square of a modulus of the complex frequency spectrum to obtain two-dimensional bone texture frequency domain feature distribution representing texture energy distribution in a spatial frequency domain, and converting frequency coordinates into frequency units.
  6. 6. The method for generating a medical image quality control report according to claim 1, wherein the extracting of the structural spectrum and the full-angle average structural spectrum along the main bearing bone trabecular stress conduction direction based on the two-dimensional bone texture frequency domain feature distribution respectively comprises: Setting a wedge-shaped statistical window with a preset fixed angle width by taking the stress conduction direction of the main bearing bone trabecula as a central axis, calculating radial power average values of all angles of the two-dimensional bone texture frequency domain characteristic distribution in the wedge-shaped statistical window to obtain a structural spectrum along the stress conduction direction of the main bearing bone trabecula, and calculating radial power average values of the two-dimensional bone texture frequency domain characteristic distribution in all angle ranges to obtain a full-angle average structural spectrum.
  7. 7. The method for generating a medical image quality control report according to claim 1, wherein calculating the effective texture cut-off frequency of the stress direction corresponding to the structural spectrum and the average effective texture cut-off frequency corresponding to the full-angle average structural spectrum by using a piecewise linear breakpoint fitting model respectively comprises: The method comprises the steps of determining a frequency interval for fitting, converting a frequency value and a power value of the structural spectrum or the full-angle average structural spectrum into natural logarithmic coordinates, constructing a continuous two-segment linear model comprising a low-frequency power law attenuation segment and a high-frequency attenuation segment, wherein the continuous two-segment linear model keeps numerical continuity at a break point of a logarithmic coordinate system, setting a plurality of candidate break points in the frequency interval, solving model parameters by using a linear least square method for each candidate break point, calculating a fitting residual square sum, selecting a candidate break point corresponding to the minimum fitting residual square sum, and converting the candidate break point from the logarithmic coordinates back to the linear frequency coordinates to respectively obtain the effective texture cut-off frequency in the stress direction and the average effective texture cut-off frequency.
  8. 8. The method for generating a medical image quality control report according to claim 1, wherein calculating a logarithmic ratio of the stress direction effective texture cut-off frequency to the average effective texture cut-off frequency to obtain a main bearing bone trabecular structure anisotropy retention index comprises: Dividing the effective grain cut-off frequency of the stress direction by the average effective grain cut-off frequency to obtain a frequency ratio, and calculating the natural logarithm of the frequency ratio to obtain the anisotropic retention index of the main bearing bone trabecular structure.
  9. 9. The method for generating a quality control report of medical image according to claim 1, wherein comparing the anisotropic retention index of the trabecular structure of the main bearing bone with a preset minimum recognizable structure threshold, and generating a quality control report indicating whether the image quality is acceptable or unacceptable according to the comparison result, comprises: The method comprises the steps of calculating the total number of discrete frequency points used for calculating the effective texture cut-off frequency of the stress direction, constructing the minimum identifiable structure threshold based on the total number of the discrete frequency points, enabling the minimum identifiable structure threshold to be reduced along with the increase of the total number of the discrete frequency points, judging whether the anisotropic retention index of the main bearing bone trabecular structure is larger than or equal to the minimum identifiable structure threshold or not, if yes, outputting a qualified conclusion in a quality control report, and if not, outputting a disqualified conclusion in the quality control report.

Description

Medical image quality control report generation method Technical Field The invention relates to the technical field of medical image detection, in particular to a medical image quality control report generation method. Background Dual-energy X-ray absorption (DXA) has been widely used for bone density measurement and fracture risk assessment as a clinical gold standard for osteoporosis diagnosis. With the advancement of medical image analysis technology, analysis of proximal femur has not been limited to simple bone mineral content determination, but advanced applications including evaluation of bone trabecular microstructure, finite element mechanical analysis, and three-dimensional reconstruction have been developed. These advanced analysis means place very high demands on the sharpness, contrast and structural integrity of the input image, as the original quality of the image data determines the accuracy and reliability of the subsequent mechanical inferences and pathological diagnostics. However, uneven distribution of soft tissue thickness in the proximal femoral region results in significant differences in X-ray penetration attenuation, often resulting in low frequency brightness bias and dark image interference in the image, and DXA imaging is often accompanied by higher quantum noise to reduce radiation dose. The existing image quality evaluation method mostly depends on general signal-to-noise ratio calculation or edge sharpness detection, and isotropic degradation caused by blurring of an imaging system and bone fracture Liang Xishu caused by osteoporosis are difficult to distinguish effectively. The quality degradation and pathological change are similar in isomorphism, so that the traditional full-image statistical mode is extremely easy to generate misjudgment, and a low-quality image which does not really meet the structural analysis requirement cannot be identified. In addition, existing texture analysis techniques often consider skeletal microstructure as a randomly distributed signal, ignoring the inherent mechanical conduction paths inside the proximal femur that follow Wolff's law distribution, particularly the primary bearing bone trabeculae. During bone loss, the directionality of the primary bearing trabeculae is usually relatively preserved, and structural information of the directionality is synchronously erased during image blurring or excessive smoothing. Disclosure of Invention The invention provides a medical image quality control report generation method, which solves the technical problems in the background technology. The invention provides a medical image quality control report generation method, which comprises the following steps: Acquiring a dual-energy X-ray proximal femur scanning image to be analyzed and a corresponding proximal femur anatomy segmentation mask, and performing soft tissue background field correction processing on image data in the proximal femur anatomy segmentation mask to obtain a background bone-removed microstructure distribution map; determining the stress conduction direction of the trabecula of the main bearing bone based on the geometrical relative position of the femoral head anatomical center and the femoral distance bearing region in the background bone microstructure distribution map; Performing spatial frequency domain transformation on the background-removed bone microstructure distribution map to obtain two-dimensional bone texture frequency domain feature distribution, and respectively extracting a structural spectrum and an all-angle average structural spectrum along the stress transmission direction of the main bearing bone trabecula based on the two-dimensional bone texture frequency domain feature distribution; Respectively calculating effective texture cut-off frequency of the stress direction corresponding to the structure spectrum and average effective texture cut-off frequency corresponding to the full-angle average structure spectrum by using a piecewise linear breakpoint fitting model; Calculating the logarithmic ratio of the effective grain cut-off frequency of the stress direction to the average effective grain cut-off frequency to obtain the anisotropic retention index of the main bearing bone trabecular structure; And comparing the anisotropic retention index of the main bearing bone trabecular structure with a preset minimum identifiable structure threshold value, and generating a quality control report indicating whether the image quality is qualified or unqualified according to a comparison result. The invention has the beneficial effects that the technical problem that isotropic degradation caused by blurring of an imaging system and bone small Liang Xishu caused by osteoporosis are difficult to distinguish by the traditional full-view statistics means are effectively solved by introducing an anatomic mechanical anchoring mechanism, and the discernability of an image structure is quantified by utilizing the fact that mai