Machine Learning Diagnostic for Automated Identification and Classification of Bone Marrow Cells

Application

Assist in the diagnosis of hematologic disorders of bone marrow.

Key Benefits

  • Automated machine learning approach for performing bone marrow differential counts that accounts for all viable cells on the smear.
  • Preliminary data generated by a prototype shows the system is fast and demonstrates precision compared to manual counting.
  • Potential to be a rapid, automated approach for the diagnosis of many hematologic diseases, facilitating earlier and more efficacious treatment.

Market Summary

Bone marrow aspiration biopsy is performed on over half a million patients annually in the US and is a critical aid in the diagnosis of a myriad of hematologic diseases, including anemias, leukemia, lymphoma, and multiple myeloma. The biopsy is performed by inserting a needle into the bone to withdraw a liquid bone marrow sample. The gold-standard approach for analysis is manual microscopy in which a subset of marrow cells smeared on a glass slide is assessed. However, this is a labor-intensive process that is prone to inter-observer variability and potentially inaccurate diagnosis. While widely available automated methods exist for similar microscopic studies of blood cells, clinically useful automated tools for bone marrow assessment are not readily available. In part, this is due to the complexity of bone marrow cell types and their crowding on pathology slides. Hence, there is an unmet need for more rapid and accurate automated methods for diagnostic assessment of bone marrow cells.

Technical Summary

Emory researchers have developed a novel bone marrow analysis tool with improved precision and efficiency compared to current manual microscopy methods. This tool was built using artificial intelligence (AI)/ machine learning with whole-slide digital pathology images of bone marrow aspirate smears to automate the identification and accurate classification of 16 important bone marrow cell types. The tool employs three new neural network models which together create a powerful pipeline to facilitate an essential step in bone marrow evaluation for hematologic disease. A highly promising prototype has been developed.

Developmental Stage

Pre-clinical stage of development.

Publication

  • Lewis JE, Shebelut CW, Drumheller BR, Zhang X, Shanmugam N, Attieh M, Horwath MC, Khanna A, Smith GH, Gutman DA, Aljudi A, Cooper LAD, Jaye DL. An Automated Pipeline for Differential Cell Counts on Whole-Slide Bone Marrow Aspirate Smears. Modern Pathology. 2023; 36(2):100003.
  • Chandradevan R, Abdulrahman AA, Drumheller BR, Kunananthaseelan N, Amgad M, Gutman DA, Cooper LAD, Jaye DL. Machine-Based Detection and Classification for Bone Marrow Aspirate Differential Counts: Initial Development Focusing on Non-Neoplastic Cells. Laboratory Investigation 2020, 100:98-109

Patent Information

Tech ID: 22132
Published: 6/5/2023