7.0 Tesla High-Field MRI for Quantitative Assessment of Posttraumatic Spinal Cord Injury Area Volume in Animal Models

is a debilitating CNS damage resulting in destruction of nervous fibers and development of neurological deficit. Regenerative medicine and especially cell therapy are commonly considered to be promising treatment options for SCI. Precise quantitative assessment of injury structure and volume is critical for evaluation of therapy efficiency. Current methods for calculation of the injured spinal cord parameters are based on manual delimitation of the interested area. However, this technique is often tedious and imprecise. New Method: Standard rat models of severe acute SCI were used. Spinal cord MRI was performed in 1 day after SCI and further for 4 times with 1-week interval. We designed software “Spinal cavity Searcher” based on the algorithm of MRI T2-WI scans analysis and implementing binarization and Freeman chain code. Results: ScS software semi-independently calculates the volume of SCI area and estimates the structure of the injured region. This simplifies the process of calculation and makes it more precise. We also compared results of implemented algorithm with manual calculation data and found no significant difference. Conclusion: Described method of quantitative calculation of rat spinal cord posttraumatic region structure and volume simplifies the assessment procedure due to automatization of region of interest (ROI) separation comparing to manual calculation technique. The level of preciseness is comparable in both methods. Clinical Relevance: Designed algorithm promotes the process of noninvasive control for therapy efficiency using MRI data.


Introduction
Spinal cord injury is a complex dynamic process occurring after mechanical damage of the spinal cord and followed by a series of events leading to secondary central nervous system (CNS) injury and subsequent death of spinal cord nervous cells and fibers. This is followed by manifestation of neurological deficit. Secondary injury of the spinal cord occurs due to ischemia, inflammation, oxidative stress and subsequent tissue necrosis and apotosis followed by fibrous scar [1,2]. Finally, the certain level of cystic-glial-fibrous transformation of the injured spinal cord is observed. Spinal cord injury is one of the major medical and social problems of modern medicine. No effective pathogenetic treatment alternatives are available today. All routinely used treatment techniques are either palliative or have significantly low efficiency. This demonstrates that new treatment alternatives are highly demanded [3]. One of the most promising options for spinal cord injury (SCI) treatment is regenerative medicine. In preclinical and/or clinical studies, while assessing the efficiency of neuroprotective and/or neuroregenerative therapy, it is important to estimate the internal structure and volume of posttraumatic spinal cord region in the most precise manner [4].
Today we consider MRI scanning to be the gold standard for non-invasive diagnostics of spinal cord injury [5]. The problem of quantitative calculation of posttraumatic spinal cord region parameters (including area and volume) using MRI is currently unsolved. Each contusion area having highly heterogeneous and complex structure is consuming for any visualization techniques due to inexact borders and completely nonlinear dynamics of structure changes in time. In most cases quantitative calculations of posttraumatic spinal cord parameters are performed manually, often using routine software such as ImageJ or VG Studio Max [4,6,7] or other scan viewers. All algorithms of manual calculation include the stage of manual delimitation of the interested region (ROI). This process is consuming and long-lasting, especially in large studies including multiple animals. Thus, the automatization of spinal cord contusion region volume calculation in animal models appears to be extremely important for further therapy optimization and is interesting both for preclinical investigators and for clinical practitioners.

Selection and Examination of Animal Models
Total number of 10 animals were included into the study:

Image Analysis Techniques
Automatic delimitation of the interested region (ROI) in processed images was implemented using binarization algorithm with lower limitation. Such conversion allows decreasing significantly the amount of scan information avoiding the loss of detailing. The obtained binary images explicitly determine the boundaries of the object. Contour analysis was performed using the contour encoding technique -Freeman Chain Code [9]. Chain codes allow representing the objects borders (contours) as the sequence of strait line segments with certain length and direction. This representation is based on 8-linked (octagonal) grid ( Figure 1). The length of each segment is determined by grid resolution while directions are set by the chosen code (e.g. all directions of 8-linked chain code require 3 bytes). We used Otsu method (Nobuyuki Otsu) [10] to select the optimal value of binarization lower limits. This algorithm separates two pixel classes ("valuable" and "background" pixels) using exact level of limitation within minimal intra-class dispersion. Manual calculation of isolated ROI area was realized using Gauss formula (calculation of polygonal shape area which vertexes are determined on a plane by Cartesian coordinates).   We estimated the dynamics of contusion area volume changes using developed software. We also compared the obtained results with manual calculation technique to estimate the preciseness of the developed algorithm (Figure 4). Mann-Whitney criteria evaluation demonstrated no significant differences of contusion area volume in both calculation methods (e.g. computed vs manual) ( Table 1).

Figure 4: Dynamics of contusion area volume changes.
Blue line -Result of programed calculation using ScS implemented algorithm.
Red line -Result of manual calculation within the ScS program.

Conclusion
Developed computer algorithm for spinal cord contusion volume calculation provides the automatic process of objective data collection concerning structure, size and volume of spinal cord injured region using MRI imaging. This algorithm is available for future implementation in analytical preclinical and clinical studies of various regenerative technologies related to MRI imaging.