The Use of AI in Flow Cytometry Research

Artificial intelligence is reaching into every branch of our existence. With AI, we can speed up our work, reduce human error, and save money on paid staff time. In flow cytometry research, AI has been trialled in classifying and diagnosing lymphomas from MFC and FCM data.

The research found that there was a good level of accuracy within the algorithm, however there were instances where the computer classified a number of lymphomas incorrectly.

The time taken was reduced so drastically however, that it was still quicker to run the algorithm and for a human to go through and check the results are correct than to not use it.

Over time the algorithm will be updated and we can expect it to become more accurate, and once accuracy is higher then we will see AI being used in the classification and diagnosis in the relevant illnesses.

AI in Flow Cytometry Research… Zhao et al.: FCM data + AI = somewhat accurate

Zhao et al. used AI within their flow analysis using FCM data.

They found that AI can help massively in getting the bulk of the work done – such as tagging cell populations and classifying B-cell lymphomas.

They ran the algorithm and found that whilst the bulk of classification results were right and it sped the process up as a whole, a human still had to go through and check results as the computer classed some lymphomas as normal and vice versa.

Source:

https://onlinelibrary.wiley.com/doi/10.1002/cyto.a.24191

AI in flow cytometry research

AI in Flow Cytometry Research… Hollein et al.: MFC data + AI = 74% accurate

Hollein et al. used computational algorithms with AWS to process and classify MFC data from samples that were suspected to contain B-cell lymphomas.

Their accuracy was 97% for 2 classes, and 74% for 9.

They suggested that some more work on the algorithm would improve these rates, but were happy overall with the experiment in that it proved that AI will work. The next step, in their words, is “to further improve the accuracy of classification a future algorithm will integrate distance metrics following SOM generation into the NN classifier.”

Source:https://ashpublications.org/blood/article/132/Supplement%201/2856/263692/An-Artificial-Intelligence-AI-Approach-for

Conclusion

Both studies found that the algorithm requires improvement, but show that AI has promise in classifying and diagnosing lymphomas in patients in the future.

high dimensional data analysis cytof analysis incyt247spectral flow cytometry