Reproducibility of the Recursive Segmentation Technique for Creating 3D Models of Thin Connective Tissue Structures from MRI Data: The Myosepta of Lampreys (Agnatha) as an Example

Abstract

The analysis, measuring, and recording of anatomical structures by means of dissection and other destructive techniques introduces inaccuracies due to the risk of damaging the original organization of a body's tissues. More so than other tissues, thin connective tissues are especially susceptible to breakage and distortions. MRI is the method of choice for visualizing and identifying connective tissue structures in 3D without the need for contrast staining, which dehydrates and shrinks soft tissues. The visualization of MRI data, however, comes with their own challenges because optimizing image resolution as well as contrast resolution in a particular data set generally results in non‐cubic voxels. Optimizing image resolution requires minimizing pixel size in the viewing plane, whereas optimizing contrast resolution among pixels in the viewing plane requires maximizing the number of hydrogen atoms and, hence, the voxel volume for a more accurate statistical summation of relaxing hydrogen atoms. The latter would create a non‐cubic voxel with a greater depth than the side length of the pixel in the viewing plane. Non‐cubic voxels, however, present a special challenge for building realistic 3D models based on MRI data, in particular for complexly configured thin connective tissue structures that are thinner than a voxel and, therefore, may not be discriminated from the surrounding tissue along their entire course. Recursive segmentation is a technique that allows the reconstruction of such structures in their entirety by using more than one MRI scan of the same specimen in the same position. While most segmentation can usually be done from 3D MRI scans, some portions of the structure may not have enough contrast resolution to be identified because of the almost cubic voxels. For such a portion, its segmentation from a 2D MRI scan with greater contrast resolution in the viewing plane, although lower image resolution in the planes perpendicular to the viewing plane, can serve as a scaffold for guiding the segmentation of the portion that is invisible in the 3D MRI scan. This technique allows the reconstruction of a realistic (“realitätsgetreues”) 3D model of the entire structure. The recursive technique requires skill and anatomical knowledge of the structures to be segmented by the operator and, therefore, needed to be tested as to its efficacy, reliability, and reproducibility even when used by different operators, some of whom may be less familiar with the anatomy of the structures and the segmentation software. By having a novice operator who is unfamiliar with the anatomy of lampreys execute the segmentation of myosepta with and without the help of the recursive technique, we were able to demonstrate that recursive segmentation using both 2D and 3D MRI data sets is essential to creating accurate and realistic 3D models of thin connective tissue structures. Realistic 3D models of myosepta of lampreys and other piscine vertebrates are necessary prerequisites for an elucidation of the biomechanics underlying undulatory swimming behavior.Support or Funding InformationLSU FoundationThis abstract is from the Experimental Biology 2019 Meeting. There is no full text article associated with this abstract published in The FASEB Journal.