Correlation of MRI and Histological Examination of Physeal Bars in a Rabbit Model

Introduction. Partial or complete physeal arrest with formation of a bony bridge, or bar, between two cartilaginous segments is frequently observed in pediatric disorders. Bone bar formation may lead to angular deformity and/or limb-length discrepancy [1]. Plain radiographs are used to evaluate bar formation, but image interpretation is difficult due to the irregular three-dimensional (3D) shape of the physis [2]. Computed tomography (CT) and magnetic resonance imaging (MRI) are also used to image the physis. MRI does not use ionizing radiation for multi-plane image acquisition; however, the shape of the 3D physis must still be interpreted from individual sagittal, coronal and transverse images. Previous investigators have used 3D MR image reconstruction and analysis to estimate physeal bar areas in humans [2, 3]. Animal models have also been used to correlate the MR appearance of the physis and bar with histological morphology [4,5], but no study to date has correlated bar areas measured from both MRI and histology. Therefore, the purpose of this study was to create a physeal bar, using a radiofrequency (RF) ablation technique, and validate the indirect measurement of bar area from MR images, using a prototype reconstruction algorithm, with direct histological measurements in a rabbit model. Methods. All methods were approved by the local IACUC. To date, 10 of 18 10-week-old male New Zealand white rabbits have been evaluated. Surgical Method: The rabbits were anesthetized, and a 0.045" K-wire was used to create a track from the lateral bony cortex of the tibiae to approximately 1⁄2 to 1/3 of the way from the lateral edge of the physis, under fluoroscopic guidance. The K-wire was removed and a RF needle inserted into the same position. RF ablation was performed at 90C for 4 minutes (2 consecutive 2-minute sessions). The rabbits were allowed free cage movement post-op and were sacrificed 6 weeks thereafter. MR imaging of the excised tibiae was performed on the same day. Image Acquisition: Imaging was performed using a 3T clinical MRI system and a prototype 2cm x 7cm birdcage coil. A 3D T1-weighted fatsuppressed spoiled gradient-recalled echo (SPGR) sequence was used to generate a volumetric dataset for physeal segmentation and area quantification. Imaging parameters were: TR 21.1ms, TE 4.6ms, FOV 6cm, flip angle 10, slice thickness 0.2mm, matrix 512x512, receiver bandwidth 31.25kHz. These parameters displayed the physeal growth plate as hyper-intense voxels. Approximately 92 images were acquired for each tibia. Image Analysis: Custom written software (Applied Science Lab, GE Healthcare, Waukesha, WI) was used for semiautomated segmentation of the normal physes. A 3D representation of the segmented image data set was constructed for each knee, the physeal bar was then defined and its area quantified. The bar area was also expressed as a percentage of a flattened projection of the whole physis [6]. Histology: Following MR imaging, the tibiae were fixed, decalcified, cut in half sagittally and embedded in paraffin blocks. Sagittal histological slices (stained with hematoxylin-eosin) were prepared from each block, with two consecutive 5μm slices taken at 220μm intervals to cover the entire area of the bar. The linear length of physeal disruption on each slice was measured with reference to a known calibrated 1mm standard. The lengths of the physeal disruption at each 5μm interval location were averaged, and the total physeal bar area calculated as the summation of a series of trapezoids thus formed (Fig. 2). This method of calculating the histological bar area is independent of the side-to-side alignment of the slices, as the size of the individual trapezoids is dependent only on the length of the sides and its height. Statistics: A paired t-test was performed to detect differences of MRI and histological bar areas. The Pearson correlation coefficient (r) was calculated between the physeal bar areas derived from MRI and histology. Results. From the 10 rabbits evaluated, the mean bar area from MRI measurements was 40.3±26.3mm (mean±st.dev.), and the mean bar area from histology measurements was 29.7±20.0mm. One tibia was too fragmented for histological measurement and was excluded from statistical analysis. Representative MR and histological images are shown in Figure 1. The bar comprised 22.2±13.8 % of the total physis in the MR analysis. The bar area from histological measurements was 26.3±23.9 % smaller than the corresponding MR measurements; however, this difference was not significant (p=0.18). A strong correlation, r=0.82 (p=0.006), was found between the area measurements (Fig. 3). Discussion. This study used MR images to quantify ablated areas of physis, and correlated the indirect MR measurements with direct histological measurements. A good correlation was found between the two measurement methods and the small differences encountered were likely due to the effects of histological sample preparation. Segmentation of the MR images enabled a 3D representation of the physis and bar to be created. The reconstructions (Fig. 2) provided a quantitative and objective method to evaluate the local bar and overall physis area. The irregular and undulating pattern of the physis was clearly visible in these reconstructions. Viewing the physis in 3D may avoid errors in interpretation or localization of physis and bar, as compared to traditional single plane imaging methods. Furthermore, the 3D MRI dataset was reformatted to increase the accuracy of defining the lateral borders of the ablated/damaged area, a known difficulty in physeal imaging [2]. The results of this study provide the foundation for quantitative evaluation of in-vivo human physes in future studies and development of predictive models for limb length discrepancy.