International testing and refinement of AI algorithms predicting acute leukemia subtypes from routine laboratory data
Abstract
Abstract Despite advances for patients with acute leukemia health disparities limit access to diagnosis and treatment. Artificial Intelligence (AI) approaches may address some disparities. We retrospectively assemble a diverse, international cohort of 6206 leukemia patients from 20 centers to test an AI tool designed to support leukemia diagnosis using standard laboratory results. Executing the pretrained algorithm results in varying accuracy metrics. With confidence cutoff predictions, 2000-fold bootstrapped area under the curve (AUROC) metrics are 0.94 for acute myeloid leukemia (AML), 0.98 for the promyelocytic subtype and 0.84 for acute lymphoblastic leukemia. However, this cutoff excludes 70.8–92.5% of patients from predictions. We improve accuracy and robustness, while maintaining generalizability via an ensemble of Isolation Forest and Local Outlier Factor increasing AUROC for AML from 0.72 to 0.84 (hold-out test set, patients below confidence threshold), while excluding only 12.1% of patients. Furthermore, we retrain the algorithm for pediatric patients.
Article Details
Authors (44)
Amin T. Turki
Yi Fan
Department of Radiation Oncology, University of Pennsylvania
Alberto Hernández-Sánchez
Wellington Silva
Shaun Fleming
Koray Yalcin
Catharina H.M.J. Van Elssen
Yazan Madanat
Magdalena Karasek
Mahmoud Aljurf
Matteo G. Della Porta
Alexandra Martinez-Roca
Luca Guarnera
Katarina Steffen
Evangelia Antoniou
Maria M. Rivas
Deepak K. Mishra
Ansgar T. Blum
Stephania Niry Manantsoa
Adeniyi Adiat
Amir Enshaei
Felicitas Thol
Maria Teresa Voso
Jia Chen
Tusneem Ahmed Elhassan
Anthony V. Moorman
María Belén Vidriales
Nina R. Neuendorff
Ahmet Koc
Pratyush Mishra
Dirk Strumberg
Roma S. Fourmanov
Lukas Heine
Jens Kleesiek
Daniel Munárriz
Gianluca Asti
Mridula Mokoonlall
Marisa Kometas
Eduardo Rego
Rabea Mecklenbrauck
Marta Sobas
Depei Wu
Felix Nensa
Merlin Engelke