ASH 2024 | Smart Cytoflow: a machine learning decision support system for myeloma diagnosis and monitoring

Smart Cytoflow is a virtual assistant for flow cytometry analysis in multiple myeloma. What Smart Cytoflow does is to automatically analyze MRD data from Euroflow panels for multiple myeloma, and automatically clusterizes cells, identifies plasma cells, and detects MRD in multiple myeloma. So we have seen very high accuracy for both diagnosis and MRD detection. Probably diagnosis is better because MRD is a little bit more tricky when you have very few cells, but still in our lab it’s working quite well. So aside from doing the algorithm, which is a combined clusterization and supervised machine learning system for atypical plasma cell identification, we have also generated an automatic software. So this is totally automatized in our lab. So this software is constantly supervising folders where we load the flow cytometry data for MRD detection in multiple myeloma. And when it identifies new data, it just anonymizes it, encrypts the data, sends that to a cluster computing service, runs the algorithm, and provides you with the results in an interactive platform, which explains whether the prediction is pathogenic or not, and whether the uncertainty about that prediction is high or not. Uncertainty is very important because when machine learning predicts something, this prediction can be very accurate or we might have a lot of uncertainty about it. And if the prediction is not very certain, if the algorithm is not very certain about the result, you should be very careful with the interpretation, with the manual interpretation of the results. So it’s just a virtual assistant for humans, particularly experts in flow cytometry in decision making for multiple myeloma diagnosis and MRD detection.

This transcript is AI-generated. While we strive for accuracy, please verify this copy with the video.