Fully Automatic Method for Reliable Spinal Cord Compartment Segmentation in Multiple Sclerosis

Abstract

BACKGROUND AND PURPOSE: Fully automatic quantification methods of spinal cord compartments are needed to study pathologic changes of the spinal cord GM and WM in MS in vivo. We propose a novel method for automatic spinal cord compartment segmentation (SCORE) in patients with MS. MATERIALS AND METHODS:The cervical spinal cords of 24 patients with MS and 24 sex-and age-matched healthy controls were scanned on a 3T MR imaging system, including an averaged magnetization inversion recovery acquisition sequence. Three experienced raters manually segmented the spinal cord GM and WM, anterior and posterior horns, gray commissure, and MS lesions. Subsequently, manual segmentations were used to train neural segmentation networks of spinal cord compartments with multidimensional gated recurrent units in a 3-fold cross-validation fashion. Total intracranial volumes were quantified using FreeSurfer. RESULTS:The intra-and intersession reproducibility of SCORE was high in all spinal cord compartments (eg, mean relative SD of GM and WM: # 3.50% and #1.47%, respectively) and was better than manual segmentations (all P , .001). The accuracy of SCORE compared with manual segmentations was excellent, both in healthy controls and in patients with MS (Dice similarity coefficients of GM and WM: $ 0.84 and $0.92, respectively). Patients with MS had lower total WM areas (P , .05), and total anterior horn areas (P , .01 respectively), as measured with SCORE. CONCLUSIONS:We demonstrate a novel, reliable quantification method for spinal cord tissue segmentation in healthy controls and patients with MS and other neurologic disorders affecting the spinal cord. Patients with MS have reduced areas in specific spinal cord tissue compartments, which may be used as MS biomarkers.

Anterior horn atrophy in the cervical spinal cord: A new biomarker in progressive multiple sclerosis

“…We applied SCORE on all AMIRA TI images acquired at each slice (Figures 1 and 2). 15,16 SCORE is a fully automated software developed in-house, which provides delineations of the cSC/CSF, cSCGM/cSCWM, cSCAH/cSCPH/cSC gray commissure and normal-appearing/lesional cSC tissue borders. It has been developed using multidimensional gated recurrent unit neural networks trained on expert manual segmentations and has shown highly reliable results.…”

“…Moreover, an accurate fully automatic Spinal Cord cOmpaRtment sEgmentation (SCORE) method using neural segmentation networks for quantification of total and lesional cSCGM, cSCWM, as well as parcellations of the cSCGM such as the anterior and posterior horns (SCAH, SCPH) on AMIRA images has been developed. 15,16 In this work, we have exploited recent advances in hardware and software technologies to investigate in depth the characteristics of cSC pathology across the MS spectrum, as well as the clinical impact of pathological changes in cSC structures in both RRMS and PMS patients. In view of previous post-mortem and electrophysiologic studies pointing to a more pronounced neurodegenerative process primarily affecting motor neurons in MS, 17,18 we hypothesized that this effect would be most evident in the cSCAH at the level of cervical enlargement.…”

“…It has been developed using multidimensional gated recurrent unit neural networks trained on expert manual segmentations and has shown highly reliable results. 15,16 After automatic segmentation, a quality control was performed. Segmentation errors or segmentations in which the presence of lesions did not allow for confident cSCGM/cSCWM differentiation were excluded after visual inspection (Supplemental Appendix 1, Supplemental Figure 1).…”

“…Moreover, an accurate fully automatic S pinal C ord c O mpa R tment s E gmentation (SCORE) method using neural segmentation networks for quantification of total and lesional cSCGM, cSCWM, as well as parcellations of the cSCGM such as the anterior and posterior horns (SCAH, SCPH) on AMIRA images has been developed. 15,16…”