Advancements in Machine Learning (ML): Transforming the Future of Blood Cancer Detection and Outcome Prediction

Abstract

The diagnosis and treatment of hematologic malignancies are becoming more and more complex. Growing knowledge of pathophysiology, diagnostic methods and, last but not least, treatment options offer many opportunities for patients, but integrating the growing amount of knowledge into daily practice can be challenging. Artificial intelligence (AI) technologies, specifically machine learning (ML) and deep learning (DL), have the potential to revolutionize the field of hematology-oncology by providing innovative solutions for diagnosis prediction, risk assessment, screening, patient prognosis estimation and treatment selection. Recent studies demonstrate that ML algorithms can rapidly predict hematologic malignancies and patient outcomes, matching or exceeding the accuracy of experienced hematologists. Moreover, ML-based prediction approaches have the potential to identify redundant tests, thus reducing the number of laboratory tests required for diagnosis. DL is a subset of machine learning that uses artificial neural networks to mimic the learning process of the human brain. DL has shown the potential to outperform ML methods in predicting patient survival rates and discovering new prognostic and diagnostic biomarkers. As research progresses, ML and DL technologies hold immense promise to significantly enhance our ability to predict the diagnosis of hematologic malignancies, leading to earlier detection and more personalized and effective treatment strategies. This review focuses on recent advancements in ML and DL for the prediction and early detection of hematologic malignancies and discusses relevant challenges regarding the generalizability of clinical prediction models. PEER REVIEWED ARTICLE **Peer reviewers:** Prof. Stefan Balabanov, University Hospital Zurich, Zurich, Switzerland Dr Martin Schneider: University Hospital Carl Gustav Carus Dresden, Dresden, Germany Received on March 04, 2024; accepted after peer review on May 13, 2024; published online on June 12, 2024.