Predicting Genetic Mutations from Single-Cell Bone Marrow Images in Acute Myeloid Leukemia Using Noise-Robust Deep Learning Models

2025-06-15

AI Summary

This study addresses the dual challenges of inaccurate clinical labels and image noise in single-cell bone marrow image-based mutation prediction for acute myeloid leukemia (AML). We propose a two-stage noise-robust learning framework: (1) a deep learning binary classifier achieves 90% accuracy in identifying myeloid blasts; (2) a weakly supervised four-class mutation classifier—trained on coarse, patient-level mutation labels—demonstrates robustness to 20% label noise, attaining 85% mutation prediction accuracy. The method integrates pathologist-in-the-loop validation, noise-robust optimization, and interpretability design. To our knowledge, it is the first approach to directly infer AML driver mutations from blast cell images without per-cell molecular annotations, thereby balancing diagnostic reliability and clinical interpretability. This work establishes a novel paradigm for imaging genomics.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Medical and Biological Imaging
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

In this study, we propose a robust methodology for identification of myeloid blasts followed by prediction of genetic mutation in single-cell images of blasts, tackling challenges associated with label accuracy and data noise. We trained an initial binary classifier to distinguish between leukemic (blasts) and non-leukemic cells images, achieving 90 percent accuracy. To evaluate the models generalization, we applied this model to a separate large unlabeled dataset and validated the predictions with two haemato-pathologists, finding an approximate error rate of 20 percent in the leukemic and non-leukemic labels. Assuming this level of label noise, we further trained a four-class model on images predicted as blasts to classify specific mutations. The mutation labels were known for only a bag of cell images extracted from a single slide. Despite the tumor label noise, our mutation classification model achieved 85 percent accuracy across four mutation classes, demonstrating resilience to label inconsistencies. This study highlights the capability of machine learning models to work with noisy labels effectively while providing accurate, clinically relevant mutation predictions, which is promising for diagnostic applications in areas such as haemato-pathology.

Problem

• Predict genetic mutations from single-cell bone marrow images
• Handle label inaccuracies and data noise robustly
• Classify leukemic vs non-leukemic cells accurately

Innovation

• Noise-robust deep learning for mutation prediction
• Binary classifier distinguishes leukemic cells accurately
• Four-class model classifies mutations despite label noise