Deep-learning-based biomarker of spinal cartilage endplate health using ultra-short echo time magnetic resonance imaging

Abstract

Background T2* relaxation times in the spinal cartilage endplate (CEP) measured using ultra-short echo time magnetic resonance imaging (UTE MRI) reflect aspects of biochemical composition that influence the CEP’s permeability to nutrients. Deficits in CEP composition measured using T2* biomarkers from UTE MRI are associated with more severe intervertebral disc degeneration in patients with chronic low back pain (cLBP). The goal of this study was to develop an objective, accurate, and efficient deep-learning-based method for calculating biomarkers of CEP health using UTE images. Methods Multi-echo UTE MRI of the lumbar spine was acquired from a prospectively enrolled cross-sectional and consecutive cohort of 83 subjects spanning a wide range of ages and cLBP-related conditions. CEPs from the L4-S1 levels were manually segmented on 6,972 UTE images and used to train neural networks utilizing the u-net architecture. CEP segmentations and mean CEP T2* values derived from manually- and model-generated segmentations were compared using Dice scores, sensitivity, specificity, Bland-Altman, and receiver-operator characteristic (ROC) analysis. Signal-to-noise (SNR) and contrast-to-noise (CNR) ratios were calculated and related to model performance. Results Compared with manual CEP segmentations, model-generated segmentations achieved sensitives of 0.80–0.91, specificities of 0.99, Dice scores of 0.77–0.85, area under the receiver-operating characteristic curve values of 0.99, and precision-recall (PR) AUC values of 0.56–0.77, depending on spinal level and sagittal image position. Mean CEP T2* values and principal CEP angles derived from the model-predicted segmentations had low bias in an unseen test dataset (T2* bias =0.33±2.37 ms, angle bias =0.36±2.65°). To simulate a hypothetical clinical scenario, the predicted segmentations were used to stratify CEPs into high, medium, and low T2* groups. Group predictions had diagnostic sensitivities of 0.77–0.86 and specificities of 0.86–0.95. Model performance was positively associated with image SNR and CNR. Conclusions The trained deep learning models enable accurate, automated CEP segmentations and T2* biomarker computations that are statistically similar to those from manual segmentations. These models address limitations with inefficiency and subjectivity associated with manual methods. Such techniques could be used to elucidate the role of CEP composition in disc degeneration etiology and guide emerging therapies for cLBP.

High contrast cartilaginous endplate imaging in spine using three dimensional dual-inversion recovery prepared ultrashort echo time (3D DIR-UTE) sequence

“…The proposed 3D DIR-UTE sequence provides better CEP contrast in comparison to previously reported dual-echo subtraction UTE and IR-FS-UTE studies, especially with regard to the contrast between the CEP and BMF [ 19 , 32 , 33 ]. To the best of our knowledge, this is the first in vivo CEP imaging study that uses the 3D DIR-UTE Cones technique; other studies have been focused on the T 2 * quantification of CEP using multi-echo UTE sequences [ 14 , 34 , 35 ]. Our study has demonstrated that this technique has great potential for morphological evaluation of the CEP and IVD health in vivo.…”

Thresholding approaches for estimating paraspinal muscle fat infiltration using T1‐ and T2‐weighted MRI: Comparative analysis using water–fat MRI

“…All images were acquired in the axial plane. All sequences were prospectively applied to the same region of the lumbar spine (L1–S1) and were acquired contemporaneously, enabling accurate anatomic co‐location across sequences 10 …”

Cartilaginous endplates: A comprehensive review on a neglected structure in intervertebral disc research

“…General Tissue origin Mesoderm (sclerotome) 26 Mesoderm 24,25 Notochord 26 Mesoderm (sclerotome) 26 Vascularity Vascular at birth, avascular in adulthood 26,29,30 Avascular 13,37 regardless of the stage of development 39 Avascular 38 Vascular at outer one-third of the AF and avascular at inner AF 36 Nutrition source Diffusion from neighboring blood vessels 53 Diffusion from synovial fluid 12,13 and the subchondral bone 12 Anaerobic glycolysis, 51 diffusion from CEP 50,51,53 Diffusion from neighboring blood vessels and CEP 50 Imaging* Hypointense on T2-weighted MR indistinguishable from bony endplate. Specific sequences such as UTE assist in differentiating CEP from bony endplate on MR 114,116,222 Hypointense on T2-weighted MR and distinguishable from adjacent soft tissue and bone 113 Hyperintense on T2-weighted MR, clear delineation from CEP in healthy IVD and gradient in boundary between AF and NP 114 Hypointense on T2-weighted MR, indistinguishable from CEP, gradient in boundary between AF and NP in healthy IVD 114…”

“…Considering the limitations of T1‐ and T2‐weighted sequences and CT for analyzing CEP features, alternative quantitative MRI approaches have been implemented 115 . Ultrashort time to echo (UTE) MRI is the most widely reported as an effective means of visualizing the CEP and can show in vivo the delineation of the BEP and CEP 114,116 . Outside of the clinical setting, both CT and MRI are widely used in ex vivo animal or human studies.…”