Researchers Review AI Potential for Blood Disease
New study evaluates the effectiveness of machine learning models in diagnosing and predicting outcomes for myeloproliferative neoplasms.
Updated on Oct. 31, 2026 in Artificial Intelligence
Live Poll
Should medical AI tools require clinical validation before being used to inform patient care?
A recently published review explores how artificial intelligence models are currently being applied to classify blood disease subtypes, grade bone marrow fibrosis, and estimate survival. While these tools show promise in research cohorts, experts note that they have yet to reach the stage of interventional testing.
Why it matters
The study aims to reconcile rapid technical advancements in AI with the practical clinical requirements for managing myeloproliferative neoplasms. This gap must be bridged before these models can reliably inform patient care or standard medical practice.
The FIBOM-AI model achieves 85.2% accuracy in prospective practice, while the AIPSS-MF prognostic model reports a 0.814 concordance index. Additionally, various digital pathology models demonstrate high performance with AUC scores ranging between 0.89 and 0.94.
The players
Blood Reviews
This is a peer-reviewed academic journal that publishes comprehensive reports on the latest developments in hematology and blood science.
The details
Investigators from Italy, the United Kingdom, and Spain analyzed numerous AI applications, including those using complete blood count variables to detect reticulin fibrosis. They concluded that while deep learning systems excel at classifying disease subtypes, future progress relies on prospective multicenter validation and head-to-head comparisons against standard prognostic scores.
Timeline
The AIPSS-MF prognostic model study was published in 2024.
Reviewers completed their search of databases and literature in August 2026.
The final narrative review was published in Blood Reviews on October 23, 2026.
The Tech Race
This research positions machine learning against the clinical validation standards for hematologic prognostic scores to determine if AI can replace manual diagnostics. It marks a shift from experimental modeling toward the rigorous requirements needed to integrate deep learning into routine pathology.
While these AI tools are not yet available for direct patient diagnosis, their development could eventually lead to faster and more precise bone marrow reporting. Patients may ultimately benefit from improved prognostic accuracy, potentially leading to more personalized treatment plans for blood disorders.
The takeaway
Artificial intelligence shows significant potential to enhance the precision of blood disease diagnostics and survival estimation. However, widespread clinical adoption requires ongoing head-to-head validation against existing medical standards to ensure patient safety and efficacy.
Further reading
Learn more about the latest innovations in Artificial Intelligence impacting global health research.
Source note: This article includes information reported by Ajmc.
Live Poll
Should medical AI tools require clinical validation before being used to inform patient care?
