Utilization of chemical shift MRI in the diagnosis of disorders affecting pediatric bone marrow.

OBJECTIVE: MRI signal intensity of pediatric bone marrow can be difficult to interpret using conventional methods. Chemical shift imaging (CSI), which can quantitatively assess relative fat content, may improve the ability to accurately diagnose bone marrow abnormalities in children. METHODS: Consecutive pelvis and extremity MRI at a children's hospital over three months were retrospectively reviewed for inclusion of CSI. Medical records were reviewed for final pathological and/or clinical diagnosis. Cases were classified as normal or abnormal, and if abnormal, subclassified as marrow-replacing or non-marrow-replacing entities. Regions of interest (ROI) were then drawn on corresponding in and out-of-phase sequences over the marrow abnormality or over a metaphysis and epiphysis in normal studies. Relative signal intensity ratio for each case was then calculated to determine the degree of fat content in the ROI. RESULTS: In all, 241 MRI were reviewed and 105 met inclusion criteria. Of these, 61 had normal marrow, 37 had non-marrow-replacing entities (osteomyelitis without abscess n = 17, trauma n = 9, bone infarction n = 8, inflammatory arthropathy n = 3), and 7 had marrow-replacing entities (malignant neoplasm n = 4, bone cyst n = 1, fibrous dysplasia n = 1, and Langerhans cell histiocytosis n = 1). RSIR averages were: normal metaphyseal marrow 0.442 (0.352-0.533), normal epiphyseal marrow 0.632 (0.566-698), non-marrow-replacing diagnoses 0.715 (0.630-0.799), and marrow-replacing diagnoses 1.06 (0.867-1.26). RSIR for marrow-replacing entities proved significantly different from all other groups (p < 0.05). ROC analysis demonstrated an AUC of 0.89 for RSIR in distinguishing marrow-replacing entities. CONCLUSION: CSI techniques can help to differentiate pathologic processes that replace marrow in children from those that do not.