# International testing and refinement of AI algorithms predicting acute leukemia subtypes from routine laboratory data

## TL;DR

This study tests an AI tool for diagnosing acute leukemia using routine lab data from a diverse international patient group to reduce health disparities.

## Contribution

The study introduces an ensemble AI model that improves diagnostic accuracy while maintaining generalizability across diverse populations.

## Key findings

- The AI algorithm achieved high AUROC metrics for specific leukemia subtypes, such as 0.98 for promyelocytic leukemia.
- An ensemble of Isolation Forest and Local Outlier Factor improved AML AUROC from 0.72 to 0.84 in the hold-out test set.
- The model was retrained to better suit pediatric patients, addressing differences in leukemia diagnosis between adults and children.

## 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.

Artificial intelligence (AI) could help improve risk predictions in acute leukaemia and reduce systemic health disparities in the diagnostic process. Here, the authors assemble a diverse, international cohort of 6,206 patients with acute leukaemias and deploy an AI tool to support diagnosis based on standard laboratory results, with refinements for both adult and peadiatric leukaemias.