Leukaemia Stem Cells and Their Normal Stem Cell Counterparts Are Morphologically Distinguishable by Artificial...

Area of Science:

• Hematology
• Artificial Intelligence
• Cell Biology

Background:

• Leukaemia stem cells (LSCs) drive disease initiation and treatment resistance in hematologic malignancies.
• Current methods lack the ability to precisely identify and quantify LSCs due to indistinct morphological features.
• Accurate LSC identification is crucial for guiding treatment and assessing prognosis.

Purpose of the Study:

• To develop an AI-driven method for distinguishing LSCs from normal stem cells.
• To demonstrate that LSCs possess unique, AI-detectable morphological features.
• To establish a foundation for AI-based morphology in identifying primitive leukaemia cells.

Main Methods:

• Utilized artificial intelligence (AI) deep learning, specifically a combination of 19 convolutional neural networks.
• Developed AI models to analyze single-cell images from a polycythemia vera (PV) mouse model.
• Compared AI analysis with traditional microscopic visualization by pathologists.

Main Results:

• AI models accurately distinguished LSCs from normal stem cells with >99% accuracy.
• AI identified unique morphological features of LSCs imperceptible to human pathologists.
• AI could also differentiate LSCs from other cell lineages in PV mice.

Conclusions:

• AI deep learning can identify unique morphological signatures of LSCs.
• AI-based morphology offers a novel approach for identifying primitive leukaemia cells.
• This technology holds potential for assessing therapy response and disease prognosis in leukaemia.