Computationally derived cytological image markers for predicting risk of relapse in acute myeloid leukemia patients following bone marrow transplantation

摘要

Allogenic hematopoietic stem cell transplant (HCT) is a curative therapy for acute myeloid leukemia (AML).Relapse after HCT is the most common cause of treatment failure and is associated with poor prognosis. Earlyidentification of which patients are at elevated risk of relapse may justify use of aggressive post-HCT treatmentoptions, potentially preventing relapse and treatment failure. In this study, our goal was to predict relapseafter HCT in AML patients using quantitative features extracted from digitized Wright-Giemsa stained post-transplant aspirate smears. We collected 39 aspirate specimens from a cohort of 39 AML patients after HCT,of which 25 experienced relapse, while 14 did not. Our approach comprised the following main steps. First, adeep learning model was developed to segment myeloblasts, a cell type in bone marrow that accumulates andcharacterizes AML. A total of 161 texture and shape descriptors were then extracted from these segmentedmyeloblasts. The top eight predictive features were identi ed using a Wilcoxon rank sum test over 100 iterationsof 3-fold cross validation. A model was subsequently built employing these features and yielded an average areaunder the receiver operating characteristic curve of 0.80 0:05 in cross validation. The top eight features includefour Haralick texture features and four fractal dimension features. The texture features appear to characterizechromatin patterns in myeloblasts while the fractal features quantify morphological irregularity and complexityof myeloblasts, in alignment with ndings previously reported for AML patients post-treatment.