Worldwide, there has been an increase in the number of cases of non-Hodgkin lymphoma (NHL). Burkitt lymphoma comprises of 30-40% of pediatric NHL cases and is a rapidly growing tumor. Access to efficient diagnostic paradigms are therefore crucial for quick therapeutic intervention. Currently, the identification of Burkitt lymphoma and other NHL involves histologic and genetic testing which can be costly and slow. Also, the process of fixing tissue and staining biopsy samples can lead to inconsistent results. Recently, Raman spectroscopy has exposed potential biomarkers in B-cells that could be indicative of cancer. However, slow acquisition speed limits the viability of adapting Raman spectroscopy in a clinical setting. Here we demonstrate a high-speed method to visualize Burkitt lymphoma cells and non-malignant B-cells which does not involve chemical alteration or destruction of cells. Preliminary results indicate higher collection of lipid droplets in malignant B-cells compared to normal B-cells. Using a support-vector machine learning algorithm, we were able to exploit these chemical differences and classify malignant cells from non-malignant cells with a sensitivity of 80% and specificity of 81.2%. Further work into refining this process can lead towards faster identification of cells and could potentially provide deeper insights into the chemical processes that occur within malignant blood cells.
|