Correlation Analysis Between Magnetic Resonance Imaging-Based Anatomical Assessment and Behavioral Outcome in a Rat Contusion Model of Chronic Thoracic Spinal Cord Injury

Abstract

Although plenty of evidences from preclinical studies have led to potential treatments for patients with spinal cord injury (SCI), the failure to translate promising preclinical findings into clinical advances has long puzzled researchers. Thus, a more reliable combination of anatomical assessment and behavioral testing is urgently needed to improve the translational worth of preclinical studies. To address this issue, the present study was designed to relate magnetic resonance imaging (MRI)-based anatomical assessment to behavioral outcome in a rat contusion model. Rats underwent contusion with three different heights to simulate various severities of SCI, and their locomotive functions were evaluated by the grid-walking test, Louisville swim scale (LSS), especially catwalk gait analysis system and basic testing, and Basso, Beattie, Bresnahan (BBB) score. The results showed that the lesion area (LA) is a better indicator for damage assessment compared with other parameters in sagittal T2-weighted MRI (T2WI). Although two samples are marked as outliers by the box plot analysis, LA correlated closely with all of the behavioral testing without ceiling effect and floor effect. Moreover, with a moderate severity of SCI in a contusion height of 25 mm, the smaller the LA of the spinal cord measured on sagittal T2WI the better the functional performance, the smaller the cavity region and glial scar, the more spared the myelin, the higher the volatility, and the thicker the bladder wall. We found that LA significantly related with behavior outcomes, which indicated that LA could be a proxy of damage assessment. The combination of sagittal T2WI and four types of behavioral testing can be used as a reliable scheme to evaluate the prognosis for preclinical studies of SCI.

Keywords: MRI; behavioral testing; correlation analysis; rat thoracic spine; sagittal T2WI; spinal cord injury.

FIGURE 1

Schematic diagram of the experimental design.

FIGURE 2

MRI acquisition and analysis of lesion region. (A) Sagittal T2-weighted images of rats in three groups. (B) Schematic drawing of the parameters, which include lesion area (LA), lesion length (LL), lesion width (LW), and signal intensity (SI). LA was calculated as hyperintense region/overall region of T9–T11 vertebral canal × 100%. (C–F) Comparison among three groups on LA, LL, LW, and SI, respectively. n = 5 per group (*p < 0.05; ^**p < 0.01; ^***p < 0.001; ^#p < 0.0001).

FIGURE 3

Experience-dependent methods of behavioral testing and Spearman’s rank correlation coefficient test. Various levels of spontaneous recovery, comparison among three groups on (A) BBB score and (B) grid-walking test, were evaluated by experience-dependent methods in rats that underwent contusion of three different severities. (C,D) Photographs of swimming test and comparison among three groups on LSS scores. n = 5 per group (*p < 0.05; ^**p < 0.01; ^***p < 0.001; ^#p < 0.0001; ns represents no statistical significance). Spearman’s rank correlation coefficient test. (E–G) Correlation analysis between lesion area and experience-dependent methods of behavioral testing including outliers. (H–J) Correlation analysis between lesion area and experience-dependent methods of behavioral testing after removing outliers.

FIGURE 4

Catwalk test and Spearman’s rank correlation coefficient test. (A) Photographs of catwalk test in three groups. Various levels of spontaneous recovery, comparison among three groups on (B) swing time, (C) max contact area, (D) regularity index, and (E) print position, were evaluated by the catwalk test in rats that underwent contusion of three different severities. n = 5 per group (*p < 0.05; ^**p < 0.01; ^***p < 0.001; ^#p < 0.0001; ns represents no statistical significance). (F–I) Correlation analysis between lesion area and indicators of the catwalk including outliers. (H–J) Correlation analysis between lesion area and indicators of catwalk after removing outliers.

FIGURE 5

Spearman’s rank correlation coefficient test of the SCI-moderate group after seven additional rats were added. (A–G) Correlation analysis between lesion area and indicators of behavior (n = 12.)

FIGURE 6

Comparison between the SCI-mod-s group and SCI-mod-l group on (A) lesion area, (B) BBB score, (C) grid-walking test, (D) LSS scores, (E) swing time, (F) max contact area, (G) regularity index, (H) print position. n = 6 per group (*p < 0.05; ^**p < 0.01; ns represents no statistical significance).

FIGURE 7

Comparison between the SCI-mod-s group and SCI-mod-l group on (A,B) cavity region; (C) glia scar; (D,E) spared myelin; (F–H) electrophysiological detection; (I,J) thickness of bladder wall (*p < 0.05; ^**p < 0.01; ns represents no statistical significance).