Accelerated 3D qCEST of the Spine in a Porcine Model Using MR Multitasking at 3T.
Publication Year:
2025
PubMed ID:
40817426
Funding Grants:
Public Summary:
Researchers recently tested a new, much faster type of 3D MRI scan to see if it could accurately measure lower back pain in pigs with injured spinal discs. By comparing this new 30-minute whole-spine scan to an older, slower method, they found both produced very similar results, proving the new technique is highly reliable. The advanced scan works by detecting specific chemical changes inside the spinal discs, allowing researchers to easily tell the difference between healthy and injured tissue. Using these chemical markers, the team trained an artificial intelligence program that was able to predict the pigs' actual pain levels with 80% accuracy—performing significantly better than traditional MRI assessments. Ultimately, this study shows that this quick, advanced scanning method could one day be used to objectively measure exactly how much back pain a patient is experiencing based on the chemical health of their spine.
Scientific Abstract:
To assess lower back pain using quantitative chemical exchange saturation transfer (qCEST) imaging in a porcine model by comparing exchange rate maps obtained from multitasking qCEST with conventional qCEST. Use a permuted random forest (PRF) model trained on CEST-derived magnetization transfer ratio (MTR) and exchange rate (k(sw)) features to predict Glasgow pain scores. Six Yucatan minipigs were scanned at baseline and at four post-injury time points (weeks 4, 8, 12, and 16) following intervertebral disc injury. Conventional qCEST imaging was performed at four B1 powers using a two-dimensional reduced field of view turbo spin-echo (TSE) sequence, with a total acquisition time of 24 min per slice. Multitasking steady-state (SS) CEST imaging was performed with pulsed saturation to achieve a steady state, acquiring 32 slices at 59 offsets for 4 B1 powers in 36 min. Exchange rate maps were generated using omega plot analysis, and CEST images were analyzed using a multi-pool fitting model to produce MTR and k(sw) maps. Permuted random forest (PRF) model was trained on MTR and k(sw) values to predict pain scores. Modic changes were assessed using T2-weighted MR images. The Pearson correlation coefficient between exchange rate maps from multitasking qCEST and conventional qCEST was 0.82, demonstrating strong agreement. The 3D qCEST (SS-CEST) technique effectively differentiated between healthy and injured discs, with injured discs exhibiting significantly higher k(sw) values. Using MTR and k(sw), the PRF model achieved 80% accuracy in predicting pain scores disc-by-disc, outperforming the correlation with Modic changes (r = 0.45, p < 0.05); with a Cohen's Kappa of 0.4. 3D steady-state qCEST with whole-spine coverage can be done at 3T within 32 min using MR Multitasking (acceleration factor of 22), and qCEST-derived biomarkers (MTR and k(sw)) can predict pain scores with an accuracy of 80%.