Label-free, rapid and highly accurate identification and categorization of leukemia cells via Raman spectroscopy

Leukemia, a highly malignant form of cancer, relies on early diagnosis for effective treatment and patient survival. Currently, leukemia is diagnosed by manually identifying blood cells through recognizing morphological features using staining techniques. In this work, we achieve rapid, accurate, label-free identification and categorization of four important leukemia cell lines to a subclass level using Raman spectroscopy and an identification model. The flat gold film-covered glass substrates ensure cell intact, good morphology and high-quality Raman spectra. Although the shapes of unstained erythrocytes, K562, U937 and Raji B cells are similar, their spectra are different. The sensitivity, the specificity and the accuracy of the unsupervised identification model (PCA-KMCA) of these four cell lines were 93.75%, 96.67%, and 97.50%, respectively. The accuracy of this model is higher than that of the supervised identification model (LDA-KNN) with an accuracy of 90.62%. Further, the unsupervised identification (PCA-KMCA) of Raji B and Jurkat cells could be greatly achieved, with sensitivity of 96.25%, specificity of 97.50% and accuracy of 96.88%. This method has great potential in the early diagnosis of leukemia, which can help doctors determine treatment plans faster and improve the prognosis of patients.