Fourier-transform infrared spectroscopy of buffy coat distinguishes acute myeloid and lymphoblastic leukemia

Abstract Acute myeloid leukemia (AML) and acute lymphoblastic leukemia (ALL) are distinct hematologic malignancies requiring accurate differentiation. This proof-of-concept study investigated Fourier-transform infrared (FTIR) spectral differences in peripheral blood buffy-coat specimens from 30 patients with AML, 24 with ALL, and 10 healthy controls. Spectra were preprocessed using extended multiplicative signal correction and Savitzky–Golay second derivatives and analyzed using peak-area analysis, hierarchical cluster analysis (HCA), principal component analysis (PCA), and partial least-squares discriminant analysis (PLS-DA). Group-associated differences were observed in protein-, ester-carbonyl-, phosphate-, and carbohydrate-associated spectral regions. PCA and HCA showed exploratory clustering, with clearer separation between healthy-control and leukemia observations than between AML and ALL. Using a spectral-observation-level 67:33 model-development/held-out prediction split, PLS-DA achieved accuracies of 98.92–100.00%, sensitivities of 97.79–100.00%, and specificities of 98.02–100.00%. These findings demonstrate group-associated spectral variation and high internal spectral-level classification performance but represent analytical discrimination rather than patient-level diagnostic accuracy. Participant-wise model development and independent external validation in larger, demographically balanced cohorts are required before clinical utility can be established.