Artificial intelligence-driven label-free detection of chronic myeloid leukemia cells using ghost cytometry

Area of Science:

• Hematology
• Biotechnology
• Artificial Intelligence

Background:

• Early diagnosis of chronic myeloid leukemia (CML) is crucial for deep molecular response but challenging due to subtle or absent symptoms and normal white blood cell counts.
• Current diagnostic methods may miss CML in its early stages, delaying critical treatment initiation.

Purpose of the Study:

• To explore the potential of artificial intelligence (AI)-based quantitative detection of CML cells using ghost cytometry (GC).
• To develop a method for identifying CML patients before peripheral blood count abnormalities appear.

Main Methods:

• Pre-trained AI models were developed using morphological data of peripheral blood leukocytes from newly diagnosed CML patients and healthy individuals.
• These AI models were applied to analyze peripheral blood samples from CML patients undergoing tyrosine kinase inhibitor treatment.

Main Results:

• The AI model demonstrated accurate detection of CML cells, even when present in small quantities after treatment initiation.
• A strong correlation was observed between AI-detected CML cell counts and BCR::ABL1 IS mRNA levels.

Conclusions:

• Multidimensional single-cell morphological data from GC, analyzed by machine learning, enables label-free, quantitative detection of CML cells.
• This approach may lead to a new screening test for early-stage CML detection, preceding traditional blood test anomalies.