Towards faster, better and cheaper diagnosis of blood diseases by machine learning

Together with clinicians from the lab of Prof. Arjan van de Loosdrecht (VUMC Amsterdam), researchers from the team of CRIG group leader Prof. Yvan Saeys have developed an automated machine learning pipeline for the diagnosis of ‘myelodysplastic syndromes’ (MDS), a group of malignant blood disorders.

Currently, an accurate diagnosis of MDS mostly requires the study of the bone marrow morphology and cytogenetics, often complemented with manually interpreted immunophenotyping by flow cytometry. This approach has several drawbacks, including the time investment and the lack of reproducibility and objectivity associated with a manual inspection of flow cytometry data.

As machine learning techniques can overcome many of these challenges, the researchers developed an automated machine learning pipeline for MDS diagnosis. As such, they were able to improve the diagnostic power by 10% compared to classical analysis, with a time investment of only 30 seconds, compared to about 1 hour for manual analysis.

This research brings novel developments in automated phenotyping using machine learning methods one step closer to the clinic. The methodology developed here could be easily applied to other immune-related diseases as well, with the potential for faster, better, and cheaper patient diagnosis.

Read the original article via this link.