A Three-Dimensional Digital Model for Early Diagnosis of Hepatic Fibrosis Based on Magnetic Resonance Elastography

Summary

The objective of this study was to develop a novel three-dimensional digital model for the early diagnosis of hepatic fibrosis, which includes the stiffness of each voxel in the patient's liver and can thus, be used to calculate the distribution ratio of the patient's liver at different fibrosis stages.

Abstract

Hepatic fibrosis is an early stage of liver cirrhosis, and there are no better non-invasive and convenient methods for the detection and evaluation of the disease. Despite the good progress made with the liver stiffness map (LSM) based on magnetic resonance elastography (MRE), there are still some limitations that need to be overcome, including manual focus determination, manual selection of regions of interest (ROIs), and discontinuous LSM data without structural information, which makes it impossible to evaluate the liver as a whole. In this study, we propose a novel three-dimensional (3D) digital model for the early diagnosis of hepatic fibrosis based on MRE.

MRE is a non-invasive imaging technique that employs magnetic resonance imaging (MRI) to measure the liver stiffness at the scanning site through human-computer interaction. Studies have indicated a significant positive correlation between the LSM obtained through MRE and the degree of hepatic fibrosis. However, for clinical purposes, a comprehensive and precise quantification of the degree of hepatic fibrosis is necessary. To address this, the concept of Liver Stiffness Distribution (LSD) was proposed in this study, which refers to the 3D stiffness volume of each liver voxel obtained by the alignment of 3D liver tissue images and MRE indicators. This provides a more effective clinical tool for the diagnosis and treatment of hepatic fibrosis.

Introduction

Hepatic fibrosis refers to the formation of excessive scar tissue in the liver, usually as a result of liver damage or disease^1,2,3,4. It frequently arises as a consequence of chronic liver injury and is commonly associated with liver diseases, such as chronic viral hepatitis, non-alcoholic fatty liver disease, and alcoholic liver disease. If left untreated, hepatic fibrosis can progress to cirrhosis, a potentially life-threatening condition associated with significant morbidity and mortality.

Active research in this area aims to elucidate the cellular and molecular mechanisms underlying the pathogenesis of hepatic fibrosis, as well as to develop novel diagnostic and therapeutic strategies to improve patient outcomes. Another objective is the noninvasive detection of the hepatic fibrosis stage, which is a critical aspect that directly correlates with disease diagnosis, treatment selection, and prognosis evaluation. Despite the importance of accurate diagnosis and the monitoring of hepatic fibrosis, traditional diagnostic methods, such as liver biopsy, are invasive and associated with significant risks. In contrast, magnetic resonance elastography^5,6 (MRE) is a promising non-invasive imaging technique that has demonstrated potential in the diagnosis and monitoring of hepatic fibrosis by quantifying liver stiffness.

In recent years, there has been significant research focused on evaluating the accuracy and reliability of MRE in the diagnosis of hepatic fibrosis, as well as its potential advantages over traditional diagnostic methods. The liver stiffness metric of MRE has been granted approval by the United States Food and Drug Administration (FDA) for clinical diagnosis, and extensive comparative analysis with pathological results has been conducted in clinical practice. The results have shown that the stiffness maps generated by MRE exhibit a strong positive correlation with various stages of liver fibrosis^{7,8,9,10,11,12}. Yet so far, the work of accurately evaluating and tracking the progression of liver fibrosis in patients through quantitative analysis of liver stiffness distribution (LSD) by matching liver structure images with MRE has not made much progress.

In this study, the medical imaging group analysis technique^13,14,15 is introduced to achieve accurate alignment of the liver structure images with the stiffness map generated by MRE in 3D space, enabling the calculation of liver stiffness values for each voxel of the entire liver. Based on the 3D-digital model of LSD, the exact distribution of patient-specific liver fibrosis staging can be calculated and evaluated. This lays a solid foundation for the precise quantitative diagnosis of early-stage liver fibrosis.

Protocol

This study utilized 3D-digital LSD modeling to reconstruct the liver of a typical patient with clinically confirmed hepatic fibrosis. The patient was recruited from a well-known liver disease treatment institution, "You An Hospital" in Beijing, China, and underwent routine upper abdominal magnetic resonance imaging (MRI) and MRE imaging after providing consent. The patient was chosen as the case study for this research method due to the confirmation of hepatic fibrosis staging through pathological examination and the absence of obvious clinical symptoms, which emphasizes the applicability and clinical value of this research in diagnosing early-stage hepatic fibrosis patients. This paper also provides a quantitative comparison between the liver of this patient and a healthy liver. The software tools used in this study are listed in the Table of Materials.

1. Data collection and preparation

NOTE: The parameter difference is not sensitive to the research method.

1. MRI scanning strategies
   NOTE: This study utilized actual DICOM data obtained from clinical imaging using a magnetic resonance imaging (MRI) device manufactured by GE. The content of the data includes IDEAL (Iterative Decomposition of water and fat with Echo Asymmetry and Least-squares estimation) water-fat separation imaging and magnetic resonance elastography (MRE) imaging.
   1. Ensure that the IDEAL data have a horizontal resolution of 256 pixels by 256 pixels, with a pixel spacing of 1.5625 mm and a slice thickness of 10 mm.
      NOTE: The scanning strategy could be further optimized, but the methodology employed in this study is applicable to higher-precision medical imaging.
2. Rename the folder of every sequence.
   NOTE: As the DICOM data exported from the equipment does not explicitly provide sequence names, during the preprocessing stage, it is necessary to add explicit names for each sequence to facilitate subsequent analysis and processing.
   1. Copy all DICOM data to a customized working directory.
   2. Navigate to the directory containing the data in MATLAB's working directory.
   3. Execute the Description_Name function to add descriptive names to the folders for each sequence.
   4. See Figure 1 for a comparison before and after renaming. Add a Description Name to each image sequence folder to facilitate the identification of the necessary image sequences for various analytical purposes.
3. Quickly check images of IDEAL.
   1. Change the directory of different phases' folders, including the in-phase, out-phase, water, and fat phases, which were stored in separate folders for imaging using IDEAL.
   2. Execute the Slice_View function to view the impact sequences for each phase.
   3. See Figure 2 for an image of the interactive graphic user interface (GUI) for the MRI-IDEAL sequence. Use the scroll bar at the bottom of the GUI to quickly browse through the different sequences.
   4. Use the MRI-IDEAL out-phase sequence as the type of MRI sequence for providing clearer descriptions of liver tissue boundaries.
      ​NOTE: In the following operations, the focus will be on using IDEAL's out-phase sequence to delineate the 3D region of the liver.

2. Extract the 3D region of the liver

NOTE: The individual voxels in the 3D region of the liver serve as spatial carriers for LSD, with the stiffness value of each voxel being derived from MRE. Extracting the 3D region of the liver tissue is a necessary step before fusion. While deep learning can be used to accomplish this task more efficiently, it is not the focus of this study. Therefore, mature software tools (e.g., MIMICS) are still used here to extract the 3D region of the liver tissue.

1. To initiate the MIMICS software, select New Project and in the ensuing dialog box, navigate to the folder containing the IDEAL out-phase images. Proceed by clicking on NEXT | the Convert button, thereby gaining entry into the sequence-editing state.
2. To create an empty Mask, click on the New button in the MASK dialog box located on the right-hand side and select the maximum threshold.
3. To delimit the area of the liver in all horizontal views, utilize the Edit Masks tool located beneath the Segment label.
4. To generate the 3D spatial part of the liver, select the liver mask that has been delineated and click on the Calculate Part from Mask button. The extracted 3D region of the liver is shown in Figure 3.
5. Click on File | Export | select the Dicom command. In the popup dialog box, choose the liver mask, set the file path and files' names, and click the OK button to complete the export of the 3D region of the liver to the specified DICOM files.

3. The Liver stiffness map sequence

NOTE: The MRE stiffness range in patients with early fibrosis is typically below 8 kpa. To view this, the sequence image labeled 'SE27_ST8K_(Pa)' should be selected.

1. Change the directory to the folder of 'SE27_ST8K_(Pa)', which contains the liver stiffness map sequence.
2. To browse through each stiffness map, execute the MRE_show function in Matlab's workspace, with the function's argument being the filename located in the specified path.
3. The liver stiffness map shown in Figure 4 is an RGB true-color image, with a data structure of 512 pixels by 512 pixels by 3 matrix, where each pixel point has three values representing the three primary colors, RBG. Observe the color bar on the left that displays the corresponding stiffness values of different colored pixels. Calculate the exact stiffness of each pixel by using their respective correlations.
4. The supplementary information in Figure 4 includes data such as sequence description, scan position, time, patient information, and image parameters. Use these data, particularly the image parameters, to establish the spatial relationship between MRE and IDEAL sequences.

4. 3D-Volume of liver stiffness distribution

NOTE: Each voxel in the 3D liver stiffness volume represents the stiffness value of a corresponding voxel in the 3D liver region, which is derived from the stiffness value of each pixel in Figure 4. By aligning the 3D liver region in Figure 3 with the stiffness map in Figure 4, the stiffness value of each voxel can be extracted, resulting in the generation of the 3D liver stiffness volume.

1. Invoke the LSD_Slice function with the 3D liver region shown in Figure 3 and the Liver stiffness map in Figure 4 as input parameters to obtain the 3D-Volume of liver stiffness distribution, as shown in Figure 5.
2. View the stiffness map of each layer of the liver by dragging the scroll bar below the GUI shown in Figure 5.
   NOTE: However, unlike Figure 4, only liver tissue is accurately retained here.
3. Observe the icons in the upper right corner of the GUI (Figure 5) such as zooming in, zooming out, returning to the global view, and marking the coordinates of the pixel selected.
   NOTE: The default color bar is the colormap of "jet" which means that the corresponding values (Unit kpa) from blue to red are low to high.
4. Execute the LSD_Volume function with the same input as LSD_Slice to obtain the spatial distribution of the 3D liver LSD, as shown in Figure 6. View the 3D-volume of LSD from any perspective by holding down the left mouse button and dragging the screen (Figure 6).

5. LSD quantitative analysis

NOTE: An important quantitative analysis focus of this study is to provide the proportion of different stages of LSD voxels in the patient's liver. Figure 6 shows that the distribution of liver fibrosis in patients is uneven in different spatial locations. The reason why clinical symptoms are not yet obvious is mainly due to a considerable proportion of liver tissue being in a normal stage. Therefore, it is necessary to quantify precisely the difference between patients and healthy individuals. This is an important quantitative concept of this study.

1. Determine the numerical ranges of stiffness values for different stages of hepatic fibrosis, as shown in Figure 7.
2. Calculate the distribution of the patient's entire liver voxels in different fibrosis stages (Figure 8) by invoking the Hepatic_Fibrosis function with the input parameter of the 3D-volume of LSD shown in Figure 6.
3. Use the same steps to calculate and compare the results of a completely healthy liver with the typical liver fibrosis patient described above (Figure 9).

Results

By utilizing the information in the Description_Name field of DICOM files, the original MRI folder can be renamed to facilitate the rapid localization of the required imaging sequence during the analysis process in the imaging group. The MRI-IDEAL out-phase sequence is the type of MRI sequence used for providing clearer descriptions of liver tissue boundaries. This is because the MRI-IDEAL out-of-phase sequence can better differentiate the magnetization strength and angle of different tissues through spe...

Discussion

In clinical practice, it is challenging to accurately quantify and track the condition of early-stage hepatic fibrosis patients. The comparison shown in Figure 9 fully reflects the degree of hepatic fibrosis in the patient compared to a healthy liver; of course, this figure can also be a comparison between two different periods for the patient, used for evaluation of treatment efficacy. This precise quantification method is the core critical step of this study. Furthermore, the calculation m...

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
MATLAB	MathWorks	2022B	Computing and visualization
Mimics	Materialise	Mimics Research V20	Model format transformation
Tools for 3D_LSD	Intelligent Entropy	HepaticFibrosis V1.0	Beijing Intelligent Entropy Science & Technology Co Ltd. Modeling for CT/MRI fusion