Machine Learning Models May Improve CLL Diagnosis and Management

A review of machine learning (ML) models developed to support the management of chronic lymphocytic leukemia (CLL) have demonstrated positive outcomes, including accurate diagnosis and improved work flow, according to a report published in the Health Informatics Journal.

“Given the challenges in CLL management, such as inefficiencies in flow cytometry gating, inconsistent genomic profiling, and difficulties in tailoring therapy for heterogeneous patient populations, ML models offer significant potential to address these gaps and enhance patient care across the complex CLL spectrum,” the researchers wrote in their report.

However, the researchers noted that continued work is needed to broaden datasets and leverage unstructured clinical data. Current models are specialized and not able to provide support across the continuum of CLL management.

The researchers reviewed 20 studies of supervised ML models designed to predict outcomes or guide treatment decisions for patients with CLL that were published between 2014 and 2023. Each study included a minimum of 100 real patients or samples.

There were 12 studies published between 2022 and 2023, 5 were published between 2020 and 2021, 2 were published during 2019, and 1 was published during 2014. No publications meeting the criteria were published between 2015 and 2018.

Most studies included 100 to 999 patients or samples, whereas 4 included 1000 to 9999 and 2 included 10,000 or more. Of the 20 studies, 14 developed ML for the diagnosis of CLL, 5 created a prediction model type, and 1 developed a model to aid treatment decisions. The most common data source for the models were flow cytometry (30%), morphology (20%), genomics (20%), mixed (15%), lab tests (10%), and free text (5%).

The researchers noted that “there has been a clear trend towards increasing complexity and diversity in the application of ML techniques in CLL. More recent studies have explored the integration of multiple data types.”

The researchers also discussed a novel study that developed a model to predict first-line treatment of CLL by using unsupervised clustering was used with demographics and laboratory test results. However, they noted that a single-center data set was used to develop the model, “which raises concerns about the generalizability of their findings.”

“The role of ML models is likely to change from isolated tools and become part of a more comprehensive, ML-assisted CLL management system,” the researchers concluded. “Such a system would not only improve diagnostics, but also enhance treatment planning, predict outcomes, and ultimately improve patient care across the entire CLL journey.”