Detection of melanoma lymph node metastases by proton magnetic resonance spectroscopy of fine-needle aspirates

Abstract

Introduction Non-surgical sentinel node (SN) evaluation offers advantages over SN removal. Proton Magnetic Resonance Spectroscopy (MRS) of lymph node needle aspirates accurately identifies the presence of metabolites associated with melanoma micrometastases. Validation of this technique would reduce the morbidity and costs of SN evaluation. Methods Fine needle aspiration biopsies (FNAB) from 70 malignant and 42 benign nodes from patients undergoing node resection for metastatic melanoma were obtained. Proton MRS (8.5 T) was carried out using standard protocols (4). Four 5~tm sections from each node block (3mm thick) were stained with H&E (2 sections) and for S 100 protein and HMB45. MR spectra and histopathology were correlated using a statistical classification strategy (SCS) (3). Results Proton MRS of FNAB from benign and metastatic nodes are shown in Figure 1. Resonances include those from lipid (Lip), amino acids, lactate, creatine (Cre), phosphocreatine, choline (Chol) metabolites and inositol. An SCS-based classifier was generated for benign and metastatic nodes. In four random training sets, spectra from 47 metastatic and 28 benign nodes were subjected to SCS. Using five optimally discriminatory spectral regions, metastases was predicted with a sensitivity of 97.3%, a specificity of 90.2% and an accuracy of 94.7%. In independent validation sets (samples not in training sets), including 23 metastatic and 14 benign nodes, the presence of metastases was predicted with a sensitivity of 93.5%, a specificity of 87.5% and an accuracy of 91.2%. The crispness of the data (% samples with a class probability >75%) was ~88% for training and validation sets. These data indicate that SN staging of melanoma may be achieved without surgical biopsy and histopathology. Conclusions Pro- ton MRS of FNAB of lymph nodes provides robust, accurate diagnosis of metastatic disease in nodes from melanoma patients.