Artificial intelligence (AI) has been used to identify blast cells in acute myeloid leukaemia (AML) blood smears, with results are at least as good as those by a trained cytologist, German researchers have reported.
In a pilot study, a team in Munich trained a deep neuronal network, using almost 20,000 single white blood cell images, to classify each cell by leukocyte type and whether it had any blast-like characteristics. The overall aim of their work is to improve the efficiency of the evaluation of blood smears for the diagnosis of blood cancers.
Drs Carsten Marr and Christian Matek from the Institute of Computational Biology at Helmholtz Zentrum München, and Professor Karsten Spiekermann and Simone Schwarz from the Department of Medicine III, University Hospital, LMU Munich, used images extracted from blood smears from 100 AML patients and 100 controls.
They found the performance of the AI to be at least as good as that of trained human cytologists, according to research published in Nature Machine Intelligence.
Dr Marr said: “To bring our approach to clinics, digitisation of patients’ blood samples has to become routine. Algorithms have to be trained with samples from different sources to cope with the inherent heterogeneity in sample preparation and staining.
“Together with our partners we could prove that deep learning algorithms show a similar performance as human cytologists. In a next step, we will evaluate how well other disease characteristics, such as genetic mutations or translocations, can be predicted with this new AI-driven method.”
As part of the study, the team produced the very first large dataset of images which can be used to test their deep neuronal network and others like it. Now Dr Marr and his team are working with the Department of Medicine III at the University Hospital of LMU Munich and the Munich Leukaemia Laboratory to digitalise hundreds more patient blood smears.
Matek C, Schwarz S, Spiekermann K, Marr C (2019) “Human-level recognition of blast cells in acute myeloid leukaemia with convolutional neural networks”, Nature Machine Intelligence, doi: 10.1038/s42256-019-0101-9