Proton MR CSF analysis and a new software as predictors for the differentiation of meningitis in children

• Published in: NMR in biomedicine Vol. 18; no. 4; pp. 213 - 225
• Main Authors: Subramanian, Arunachalam, Gupta, Abhishek, Saxena, Swapnil, Gupta, Ashish, Kumar, Raj, Nigam, Anjali, Kumar, Rashmi, Mandal, Sudhir K., Roy, Raja
• Format: Journal Article
• Language: English
• Published: Chichester, UK John Wiley & Sons, Ltd 01-06-2005
• Subjects: 1H NMR; Algorithms; Biomarkers - cerebrospinal fluid; cerebrospinal fluid; Child; Child, Preschool; Diagnosis, Computer-Assisted - methods; Diagnosis, Differential; differential diagnosis; discriminant function analysis; expert system; Female; Humans; Magnetic Resonance Spectroscopy - methods; Male; meningitis; Meningitis - cerebrospinal fluid; Meningitis - classification; Meningitis - diagnosis; metabolite quantification; Multivariate Analysis; Protons; Reproducibility of Results; Sensitivity and Specificity; Software
• ISSN: 0952-3480; 1099-1492
• DOI: 10.1002/nbm.944

This article describes proton MR spectroscopic analysis of cerebrospinal fluid of 167 children suffering from meningitis and 24 control cases. Quantification of 12 well‐separated and commonly observed cerebrospinal fluid metabolites viz., β‐hydroxybutyrate, lactate, alanine, acetate, acetone, acetoacetate, pyruvate, glutamine, citrate, creatine/creatinine, glucose (total) and urea was carried out using Bruker's NMRQUANT software with respect to a known concentration of sodium‐3‐(trimethylsilyl)‐2,2,3,3‐d4‐propionate (TSP), serving as an external reference. The assignment of urea in CSF is reported for the first time by NMR. The presence of cyclopropane, observed for the first time in tuberculous meningitis overall in 85.1% of cases, acts as a finger‐print marker for the differential diagnosis. Multivariate discriminant function analysis was carried out for the proton MR‐detected metabolite information and the clinical symptoms data of the meningitis and control cases to find the important descriptors for classification, followed by a re‐validation of the entire database. It was found that the control could be differentiated from the disease group with a success rate of 96.4%, followed by the differential diagnosis of tuberculous meningitis with a corresponding value of 77.2%. Excluding the presence of cyclopropane, bacterial meningitis could be classified 84.4% correct and viral meningitis with a rate of 83.3%. It is proposed that the NMR spectroscopic information, along with other routine clinical features, may serve as an additional diagnostic tool for the differential diagnosis of meningitis in children. Copyright © 2004 John Wiley & Sons, Ltd.