Related Experiment Video
Updated: Feb 22, 2026
Database-guided Flow-cytometry for Evaluation of Bone Marrow Myeloid Cell Maturation
Published on: November 3, 2018
Artificial intelligence differentiates prefibrotic primary myelofibrosis with thrombocytosis from essential
Andrew Srisuwananukorn1, Giuseppe Gaetano Loscocco2,3, James M Dolezal4
1Division of Hematology, Department of Internal Medicine, The Ohio State University Comprehensive Cancer Center, Columbus, OH, USA. Andrew.srisuwananukorn@osumc.edu.
Abstract:
Prefibrotic primary myelofibrosis (prePMF) and essential thrombocythemia (ET) are distinct myeloproliferative neoplasms (MPNs) with overlapping clinical features, often leading to diagnostic uncertainty. We developed an artificial intelligence (AI) framework with human interpretability to distinguish prePMF from ET using digitized hematoxylin and eosin-stained bone marrow biopsy (BMB) slides. Trained on an initial cohort of MPN patients with thrombocytosis, the AI model achieved an AUROC of 0.89 and accuracy of 92.3%. To assess the image features guiding predictions, we generated synthetic images which potentially exaggerate disease-specific morphologies. In a blinded survey, hematopathologists reviewed both real and AI-generated images. While human experts frequently agreed with AI predictions on diagnosis with real images, diagnostic discordance reached up to 88% for AI-generated ET images despite being correctly predicted by AI. We further quantified marrow cellularity and adiposity in the real and generated images, which revealed a higher proportion of fat content in all ET images (42.0%) compared to prePMF (28.9%). These findings suggest that AI can utilize morphological cues distinct from current established diagnostic criteria, such as proportion of adiposity to distinguish types of MPNs. Thus, an AI-assisted diagnostic tool underscores the potential of AI to augment histopathologic evaluation and allow identification of more specific subpopulations of forms of MPNs.
Insights
Artificial intelligence can now differentiate prefibrotic primary myelofibrosis (prePMF) from essential thrombocythemia (ET) using bone marrow biopsy images. This AI tool aids in diagnosing these myeloproliferative neoplasms (MPNs) by identifying subtle morphological differences.
Area of Science:
- Hematopathology
- Artificial Intelligence in Medicine
- Myeloproliferative Neoplasms
Background:
- Prefibrotic primary myelofibrosis (prePMF) and essential thrombocythemia (ET) are distinct myeloproliferative neoplasms (MPNs) with overlapping clinical features, causing diagnostic challenges.
- Accurate differentiation of prePMF and ET is crucial for appropriate patient management and treatment strategies.
Purpose of the Study:
- To develop and validate an interpretable artificial intelligence (AI) framework for distinguishing prePMF from ET using digitized bone marrow biopsy (BMB) slides.
- To investigate the morphological features utilized by the AI model in differentiating these MPNs.
Main Methods:
- Development of an AI framework trained on digitized H&E-stained BMB slides from MPN patients with thrombocytosis.
- Validation of the AI model's performance using AUROC and accuracy metrics.
- Generation of synthetic images to visualize AI-driven morphological cues.
- Blinded survey of hematopathologists assessing AI predictions on real and synthetic images.
- Quantification of marrow cellularity and adiposity in both real and generated images.
Main Results:
- The AI model achieved an AUROC of 0.89 and 92.3% accuracy in distinguishing prePMF from ET.
- Hematopathologists showed high agreement with AI on real images but significant discordance (up to 88%) on AI-generated ET images.
- AI identified morphological cues, such as a higher proportion of marrow adiposity in ET (42.0%) compared to prePMF (28.9%), distinct from current diagnostic criteria.
Conclusions:
- AI can effectively distinguish prePMF from ET using BMB morphology, potentially identifying novel diagnostic features like adiposity proportion.
- AI-assisted diagnostic tools show promise in augmenting histopathologic evaluation for MPNs.
- This approach may lead to the identification of more specific subpopulations within MPNs, improving diagnostic precision.
