Search

CN-122029566-A - Method for quantifying irregularities of a surface of an anatomical region of a patient

CN122029566ACN 122029566 ACN122029566 ACN 122029566ACN-122029566-A

Abstract

The invention relates to a computer-implemented method for quantifying irregularities of a surface of an anatomical region of a patient in at least one medical image, comprising the steps of (a) obtaining at least one medical image representing the anatomical region, (b) segmenting the anatomical region or at least a part of the anatomical region, (c) detecting an actual contour of the anatomical region from the segmented anatomical region, (d) calculating a smoothed contour of the anatomical region based on the actual contour, and (e) calculating a score based on a distance between the actual contour and the smoothed contour.

Inventors

  • Xixi Yang
  • Alexander. Bonnet
  • Joan Alexis Graunes

Assignees

  • 加栢公司
  • 巴黎西岱大学
  • 国家科学研究中心

Dates

Publication Date
20260512
Application Date
20241009
Priority Date
20231011

Claims (12)

  1. 1. A method implemented by computer means for quantifying irregularities of a surface of an anatomical region of a patient in at least one medical image, the method comprising the steps of: (a) Obtaining at least one medical image representing the anatomical region, and Automatic execution: (b) Segmenting the anatomical region or at least a portion of the anatomical region, (C) An actual contour of the anatomical region is detected from the segmented anatomical region, (D) A smoothed contour of the anatomical region is calculated based on the actual contour, (E) A score is calculated based on the distance between the actual contour and the smoothed contour.
  2. 2. The method of claim 1, wherein the image is a 3D image consisting of a stack of slices, each slice forming a 2D image.
  3. 3. The method of claim 2, wherein step (b) is performed on a 3D image and steps (c), (D), and (e) are performed on each 2D slice of the plurality of slices of the 3D image.
  4. 4. Method according to any of the preceding claims, characterized in that a smooth contour is obtained by applying a Savitzky-Golay filter.
  5. 5. The method of any preceding claim, wherein during step (b) segmentation is performed using a Unet model.
  6. 6. The method according to any of the preceding claims, wherein during step (c) detection of the actual contour of the anatomical region is performed using a marching square algorithm.
  7. 7. A method according to any one of the preceding claims, wherein during step (c) unsuitable parts of the contour are excluded from the detected actual contour to obtain a remaining actual contour.
  8. 8. The method according to claim 7, wherein the unsuitable portion comprises a contoured portion surrounding certain portions of the anatomical region, a contoured portion having a curvature above a defined threshold, and/or a contoured portion having a low contrast between the inner anatomical region portion and the outer anatomical region portion.
  9. 9. Method according to any of claims 7 or 8, characterized in that after removing an unsuitable portion of the contour, a further segmentation of the anatomical region is performed on the region of interest surrounding the remaining actual contour, the actual contour being updated based on detecting the actual contour from the further segmented anatomical region.
  10. 10. Computer software comprising instructions which when executed by a processor implement at least part of the method according to any of the preceding claims.
  11. 11. A computer apparatus, comprising: An input interface for receiving a medical image, Memory for storing at least computer program instructions according to claim 10, A processor for accessing a memory for reading the above-mentioned instructions and for performing the method according to any one of claims 1 to 9, -An output interface for providing an indication based on the score.
  12. 12. A computer readable non-transitory recording medium having computer software registered thereon for implementing the method according to any of claims 1 to 9 when the computer software is executed by a processor.

Description

Method for quantifying irregularities of a surface of an anatomical region of a patient Technical Field The invention relates to a computer-implemented method for automatically and objectively quantifying irregularities of a surface of an anatomical region of a patient in at least one medical image. Background The present invention improves a method of assisting an expert user (i.e., radiologist) in calculating a liver surface nodular score (LSN score) from a medical image (e.g., CT scan). The article "nodular quantification of liver surface in conventional CT images" by Smith AD et al served as a biomarker for liver cirrhosis detection and assessment. Radiology （Liver Surface Nodularity Quantification from Routine CT Images as a Biomarker for Detection and Evaluation of Cirrhosis. Radiology）. 2016 Sep;280（3）:771-81" discloses a method of calculating a Liver Surface Nodular (LSN) score that is the average of the distances between the liver contours and a smooth polynomial curve that mimics the expected normal liver surface. The method requires the radiologist to manually delineate a region of interest (ROI) on the liver surface. Fig. 1 shows a flow chart 10 describing this prior art method. In step 11, an abdominal CT scan is performed and a 3D image is generated from the scan. In the step (12) of the process, the radiologist reviews the images generated from the CT scan and manually marks (e.g., circling, picture frame.) a region inside the liver to indicate a "region of interest" (ROI). Next, in step 13, the radiologist manually draws an area around the liver surface (or edge). Liver contour detection is performed in step 14. The prior art method uses an automated process to detect liver contours, generating an image depicting the liver edges with lines. Fig. 2 shows an example of this generated image 21, wherein a wire 22 is manually inserted by a radiologist on the edge of the liver 23. The prior art method allows detecting only a part (i.e. a local part) of the external liver contour, since specific areas where automatic detection cannot work properly have to be avoided (e.g. the method may create artefacts). These areas include areas of contact with the abdominal wall, areas with natural sharp turns or fissures, image artifacts, or liver domes. Next, in step 15, a "smoothed" liver contour is automatically generated using the image generated in step 14, which contour is shown by line 33 in fig. 3. Fig. 3 also includes a computer-generated line 32 outlining the liver. In step 16, the radiologist manually examines the region of interest (ROI) by scrolling the slices of the 3D image layer by layer to ensure that the automated process has correctly detected the liver contours within the ROI and mapped the correct distance to calculate. If a contour detection error or a rendering distance error occurs on a slice, the radiologist will need to cancel the measurement on that particular slice. After the radiologist manually identifies the ROI on the liver surface, the liver edges are automatically detected on the selected cross-section and adjacent consecutive cross-sections by propagating the delineated region of interest. The number of adjacent slices is defined by the user. Typically, the radiologist needs to calculate distances over multiple slices to obtain reliable LSN scores. In step 17, the average distance between the liver contour (line 32 in fig. 3) and the smoothed liver contour (line 33 in fig. 3) is calculated. Next, in step 18, an LSN score is generated based on the calculated average distance. In some cases, the contrast between the liver (e.g., liver) surface and its environment in the image obtained by CT scanning is not ideal, making placement of the ROI (region of interest) difficult or limiting the size of the ROI. Therefore, deriving LSN scores from medical images is both time consuming and highly specialized. Furthermore, the computed LSN scores are subjective, with different users or the same user getting different results at different times. This document proposes a method implemented by computer means for automatically and objectively quantifying irregularities of a surface of an anatomical region of a patient (such as but not limited to an organ) in at least one medical image, without interaction or supervision by an expert user (i.e. radiologist), which can solve the above-mentioned problems. Disclosure of Invention To this end, this document proposes a method implemented by computer means for quantifying irregularities of a surface (e.g. an inner surface and/or an outer surface) of an anatomical region (or a part of an anatomical region) of a patient in at least one medical image, the method comprising the steps of: (a) Obtaining at least one medical image representative of the anatomical region and automatically performing: (b) Segmenting the anatomical region or at least a portion of the anatomical region, (C) An actual contour of the anatomical region is detected from the segment