Acute leukemia diagnosis through AI-enhanced ATR-FTIR spectroscopy of peripheral blood smear

Abstract

Abstract Background Acute leukemia is a highly perilous cancer, currently diagnosed using invasive procedures like bone marrow aspirate and biopsy (BMA/BMB). There is the pressing need for non-invasive, reagent-free diagnostic approaches with exceptional sensitivity and specificity. Hence, this study explored the potential of combining attenuated total reflectance-Fourier transform infrared (ATR-FTIR) spectroscopy and artificial intelligence (AI) for diagnosing acute lymphoblastic leukemia (ALL) and acute myelogenous leukemia (AML). Methods The study analyzed the spectral data from 50 healthy individuals and 50 BMA/BMB-confirmed acute leukemia patients. Six trained models were used to assess the diagnostic performance, focusing on accuracy, positive predictive value, negative predictive value, F1-score, and area under the ROC curve (AUC). Spectral peak patterns were examined in the 1800 𝑐𝑚−1 to 850 𝑐𝑚−1 range. Results Of the six (6) trained models, the SVM model showed remarkable diagnostic performance, including accuracy, positive predictive value, negative predictive value, F1-score and AUC of 83%, 80%, 86%, 82.47% and 90.76%, respectively. Leukemia and healthy blood samples exhibited distinguishable spectral peak patterns in the amides I and II, glycogen, and phosphorylated protein regions. Conclusion This study underscores the potential of AI-enhanced FTIR spectroscopy as a valuable adjunct diagnostic tool for acute leukemia. By providing a less invasive and faster alternative to BMA/BMB, this approach offers the possibility of enhancing leukemia diagnosis and ultimately improving patient outcomes through efficient and minimally intrusive diagnostic practices, especially in pediatric and geriatric cases.