Limits...
Prediction of Neurological Impairment in Cervical Spondylotic Myelopathy using a Combination of Diffusion MRI and Proton MR Spectroscopy.

Ellingson BM, Salamon N, Hardy AJ, Holly LT - PLoS ONE (2015)

Bottom Line: Parameters that did not add value to model performance were removed, then an optimized multiparametric model was established to predict mJOA.DTI fiber tract density, MD and FA at the site of compression, along with Cho/NAA at C2, were significantly correlated with mJOA score (R2 = 0.05939, P < 0.0001; R2 = 0.4739, P < 0.0001; R2 = 0.7034, P < 0.0001; R2 = 0.4649, P < 0.0001).A combination biomarker consisting of DTI fiber tract density, MD, and Cho/NAA showed the best prediction of mJOA (R2 = 0.8274, P<0.0001), with post-hoc tests suggesting fiber tract density, MD, and Cho/NAA were all significant contributors to predicting mJOA (P = 0.00053, P = 0.00085, and P = 0.0019, respectively).

View Article: PubMed Central - PubMed

Affiliation: Department of Radiological Sciences, David Geffen School of Medicine, University of California-Los Angeles, United States of America; Department of Biomedical Physics, David Geffen School of Medicine, University of California-Los Angeles, United States of America; Department of Bioengineering, Henri Samueli School of Engineering and Applied Sciences, University of California-Los Angeles, United States of America; Department of Psychiatry and Biobehavioral Sciences, David Geffen School of Medicine, University of California-Los Angeles, United States of America.

ABSTRACT

Purpose: In the present study we investigated a combination of diffusion tensor imaging (DTI) and magnetic resonance spectroscopic (MRS) biomarkers in order to predict neurological impairment in patients with cervical spondylosis.

Methods: Twenty-seven patients with cervical spondylosis were evaluated. DTI and single voxel MRS were performed in the cervical cord. N-acetylaspartate (NAA) and choline (Cho) metabolite concentration ratios with respect to creatine were quantified, as well as the ratio of choline to NAA. The modified mJOA scale was used as a measure of neurologic deficit. Linear regression was performed between DTI and MRS parameters and mJOA scores. Significant predictors from linear regression were used in a multiple linear regression model in order to improve prediction of mJOA. Parameters that did not add value to model performance were removed, then an optimized multiparametric model was established to predict mJOA.

Results: Significant correlations were observed between the Torg-Pavlov ratio and FA (R2 = 0.2021, P = 0.019); DTI fiber tract density and FA, MD, Cho/NAA (R2 = 0.3412, P = 0.0014; R2 = 0.2112, P = 0.016; and R2 = 0.2352, P = 0.010 respectively); along with FA and Cho/NAA (R2 = 0.1695, P = 0.033). DTI fiber tract density, MD and FA at the site of compression, along with Cho/NAA at C2, were significantly correlated with mJOA score (R2 = 0.05939, P < 0.0001; R2 = 0.4739, P < 0.0001; R2 = 0.7034, P < 0.0001; R2 = 0.4649, P < 0.0001). A combination biomarker consisting of DTI fiber tract density, MD, and Cho/NAA showed the best prediction of mJOA (R2 = 0.8274, P<0.0001), with post-hoc tests suggesting fiber tract density, MD, and Cho/NAA were all significant contributors to predicting mJOA (P = 0.00053, P = 0.00085, and P = 0.0019, respectively).

Conclusion: A linear combination of DTI and MRS measurements within the cervical spinal cord may be useful for accurately predicting neurological deficits in patients with cervical spondylosis. Additional studies may be necessary to validate these observations.

No MeSH data available.


Related in: MedlinePlus

Initial and Final Combination Diffusion and Spectroscopic Biomarker Results for Prediction of mJOA.A) Initial model consisting of four MR measurements found to be individually correlated with mJOA. Note that results suggested FA at the site of compression did not add significant predictive value to the model. B) Final model consisting of three MR measurements (excluding FA).
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4592013&req=5

pone.0139451.g003: Initial and Final Combination Diffusion and Spectroscopic Biomarker Results for Prediction of mJOA.A) Initial model consisting of four MR measurements found to be individually correlated with mJOA. Note that results suggested FA at the site of compression did not add significant predictive value to the model. B) Final model consisting of three MR measurements (excluding FA).

Mentions: Multiple linear regression was then performed using the individual imaging measurements determined to be significantly correlated with mJOA, namely the ratio of maximum tract density at the site of compression to tract density at C2, FA at the site of compression, MD at the site of compression, and Cho/NAA at C2. Results suggested a linear combination of these parameters was able to predict mJOA with high accuracy (Fig 3A; P = 1.19x10-8) using the model mJOA = -3.0494•[TDI] + 7.5408 •[FA]– 4.2580•[MD]– 0.9335•[Cho/NAA]+21.7699, where TDI is the ratio of maximum fiber tract density at the site of compression to fiber tract density at C2, FA is the fractional anisotropy measured at the site of compression, MD is the mean diffusivity at the site of compression, and Cho/NAA is the measured Cho/NAA at C2. A high concordance was observed between the model predicted mJOA and actual mJOA (R2= 0.8460, P<0.0001). Individual t-tests on the regression coefficients showed that the ratio of maximum fiber tract density at the site of compression to fiber tract density at C2 (P = 0.0307), MD at the site of compression (P = 0.0047), and Cho/NAA at C2 (P = 0.0299) all added predictive value, whereas FA at the site of compression did not significantly contribute to added model performance (P = 0.1173). Since FA at the site of compression did not contribute significantly to the performance of the model, presumably due to the high correlation between FA and other parameters, this parameter was removed and the model was rerun. Results of this optimized model showed a higher accuracy in predicting mJOA (Fig 3B; P = 6.11x10-9) using the model mJOA = -4.3623•[TDI]– 5.0188•[MD]– 1.2613•[Cho/NAA]+28.5691. A similarly high concordance was observed between the model predicted mJOA and actual mJOA (R2= 0.8274, P<0.0001) with this optimized model. Individual t-tests performed on the regression coefficients in the optimized model showed that the ratio of maximum fiber tract density (P = 0.00053), MD at the site of compression (P = 0.00085), and Cho/NAA (P = 0.0019) all provided added value for predicting mJOA score.


Prediction of Neurological Impairment in Cervical Spondylotic Myelopathy using a Combination of Diffusion MRI and Proton MR Spectroscopy.

Ellingson BM, Salamon N, Hardy AJ, Holly LT - PLoS ONE (2015)

Initial and Final Combination Diffusion and Spectroscopic Biomarker Results for Prediction of mJOA.A) Initial model consisting of four MR measurements found to be individually correlated with mJOA. Note that results suggested FA at the site of compression did not add significant predictive value to the model. B) Final model consisting of three MR measurements (excluding FA).
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4592013&req=5

pone.0139451.g003: Initial and Final Combination Diffusion and Spectroscopic Biomarker Results for Prediction of mJOA.A) Initial model consisting of four MR measurements found to be individually correlated with mJOA. Note that results suggested FA at the site of compression did not add significant predictive value to the model. B) Final model consisting of three MR measurements (excluding FA).
Mentions: Multiple linear regression was then performed using the individual imaging measurements determined to be significantly correlated with mJOA, namely the ratio of maximum tract density at the site of compression to tract density at C2, FA at the site of compression, MD at the site of compression, and Cho/NAA at C2. Results suggested a linear combination of these parameters was able to predict mJOA with high accuracy (Fig 3A; P = 1.19x10-8) using the model mJOA = -3.0494•[TDI] + 7.5408 •[FA]– 4.2580•[MD]– 0.9335•[Cho/NAA]+21.7699, where TDI is the ratio of maximum fiber tract density at the site of compression to fiber tract density at C2, FA is the fractional anisotropy measured at the site of compression, MD is the mean diffusivity at the site of compression, and Cho/NAA is the measured Cho/NAA at C2. A high concordance was observed between the model predicted mJOA and actual mJOA (R2= 0.8460, P<0.0001). Individual t-tests on the regression coefficients showed that the ratio of maximum fiber tract density at the site of compression to fiber tract density at C2 (P = 0.0307), MD at the site of compression (P = 0.0047), and Cho/NAA at C2 (P = 0.0299) all added predictive value, whereas FA at the site of compression did not significantly contribute to added model performance (P = 0.1173). Since FA at the site of compression did not contribute significantly to the performance of the model, presumably due to the high correlation between FA and other parameters, this parameter was removed and the model was rerun. Results of this optimized model showed a higher accuracy in predicting mJOA (Fig 3B; P = 6.11x10-9) using the model mJOA = -4.3623•[TDI]– 5.0188•[MD]– 1.2613•[Cho/NAA]+28.5691. A similarly high concordance was observed between the model predicted mJOA and actual mJOA (R2= 0.8274, P<0.0001) with this optimized model. Individual t-tests performed on the regression coefficients in the optimized model showed that the ratio of maximum fiber tract density (P = 0.00053), MD at the site of compression (P = 0.00085), and Cho/NAA (P = 0.0019) all provided added value for predicting mJOA score.

Bottom Line: Parameters that did not add value to model performance were removed, then an optimized multiparametric model was established to predict mJOA.DTI fiber tract density, MD and FA at the site of compression, along with Cho/NAA at C2, were significantly correlated with mJOA score (R2 = 0.05939, P < 0.0001; R2 = 0.4739, P < 0.0001; R2 = 0.7034, P < 0.0001; R2 = 0.4649, P < 0.0001).A combination biomarker consisting of DTI fiber tract density, MD, and Cho/NAA showed the best prediction of mJOA (R2 = 0.8274, P<0.0001), with post-hoc tests suggesting fiber tract density, MD, and Cho/NAA were all significant contributors to predicting mJOA (P = 0.00053, P = 0.00085, and P = 0.0019, respectively).

View Article: PubMed Central - PubMed

Affiliation: Department of Radiological Sciences, David Geffen School of Medicine, University of California-Los Angeles, United States of America; Department of Biomedical Physics, David Geffen School of Medicine, University of California-Los Angeles, United States of America; Department of Bioengineering, Henri Samueli School of Engineering and Applied Sciences, University of California-Los Angeles, United States of America; Department of Psychiatry and Biobehavioral Sciences, David Geffen School of Medicine, University of California-Los Angeles, United States of America.

ABSTRACT

Purpose: In the present study we investigated a combination of diffusion tensor imaging (DTI) and magnetic resonance spectroscopic (MRS) biomarkers in order to predict neurological impairment in patients with cervical spondylosis.

Methods: Twenty-seven patients with cervical spondylosis were evaluated. DTI and single voxel MRS were performed in the cervical cord. N-acetylaspartate (NAA) and choline (Cho) metabolite concentration ratios with respect to creatine were quantified, as well as the ratio of choline to NAA. The modified mJOA scale was used as a measure of neurologic deficit. Linear regression was performed between DTI and MRS parameters and mJOA scores. Significant predictors from linear regression were used in a multiple linear regression model in order to improve prediction of mJOA. Parameters that did not add value to model performance were removed, then an optimized multiparametric model was established to predict mJOA.

Results: Significant correlations were observed between the Torg-Pavlov ratio and FA (R2 = 0.2021, P = 0.019); DTI fiber tract density and FA, MD, Cho/NAA (R2 = 0.3412, P = 0.0014; R2 = 0.2112, P = 0.016; and R2 = 0.2352, P = 0.010 respectively); along with FA and Cho/NAA (R2 = 0.1695, P = 0.033). DTI fiber tract density, MD and FA at the site of compression, along with Cho/NAA at C2, were significantly correlated with mJOA score (R2 = 0.05939, P < 0.0001; R2 = 0.4739, P < 0.0001; R2 = 0.7034, P < 0.0001; R2 = 0.4649, P < 0.0001). A combination biomarker consisting of DTI fiber tract density, MD, and Cho/NAA showed the best prediction of mJOA (R2 = 0.8274, P<0.0001), with post-hoc tests suggesting fiber tract density, MD, and Cho/NAA were all significant contributors to predicting mJOA (P = 0.00053, P = 0.00085, and P = 0.0019, respectively).

Conclusion: A linear combination of DTI and MRS measurements within the cervical spinal cord may be useful for accurately predicting neurological deficits in patients with cervical spondylosis. Additional studies may be necessary to validate these observations.

No MeSH data available.


Related in: MedlinePlus