| Abstract: |
Artificial intelligence (AI) remains an emerging technology in clinical hematology, with few implementations in broad clinical practice. There has been a rapid increase in attempts to design risk calculators to predict clinical outcomes and treatment responses, all with the goal of further personalizing therapy. However, these attempts have been constrained by the limited availability of high-quality patient data and both the technological challenge of building clinically interpretable models and the practical challenge of rigorously validating them in real-world settings. Here, a number of noteworthy or promising demonstrations of AI in clinical hematology have been described, and some of the opportunities and pitfalls that AI is currently facing in the hematology field have also been addressed. As the number of published AI models in the field increases exponentially, we anticipate the greatest utility in those models with intermediate-term outcomes such as toxicity and early treatment failure, since they might offer clear opportunities for clinical intervention. © 2024 Elsevier Inc. All rights reserved. |
