439 Cytometry and machine learning based approaches for diagnosis of malignant cells in Sézary syndrome

In cutaneous T-cell lymphoma (CTCL), the lack of early diagnostic biomarkers is a challenge. This results in delayed diagnosis, and can seriously affect treatment prognosis. Therefore, defining accurate methods for identification malignant cells pivotal importance. We explore potential artificial intelligence (AI) to uncover tumor-defining Sézary syndrome (SS), leukemic type CTCL. established mass-imaging approach mass-cytometry method acquire large-scale single-cell data from peripheral blood, following datasets were analysed by trained AI model, called CellCNN. CellCNN supervised machine learning (ML) algorithm that trains convolutional neural network with single layer using labelled as inputs. approach, we enrolled 4 healthy individuals (HDs) 5 patients SS; study, included discovery cohort 60 (20 SS, 20 atopic dermatitis, HDs) validation 33 (11 each group). Algorithm performance was assessed specificity sensitivity. successfully developed first morphology based diagnosis SS samples. The delivered best separation Sezary (84.6% abnormality) specimens (13.9% compared other models. For more bench-to-bedside translation, applied techniques AI. Our accurately identify blood (sensitivity = 0.91 1.0) on pattern cell surface molecules. findings pave way an easy-to-implement sensitive facilitate tumors involvement.