The survival of acute myeloid leukemia (AML) patients is poor. In chronic myeloid leukemia (CML), it is unclear which patients could safely discontinue therapy. Previous personalized medicine initiatives have not been able to improve treatment responses likely due to unreliable statistical models. By combining large amounts of image and clinical data, we hypothesize that we can build algorithms predicting the disease course of AML and CML patients. We aim to (1) scan the world's largest collection of routine MGG-stained samples into images from 13 clinical centers in Finland, Sweden, Norway, Japan, and Australia; (2) train algorithms to detect blood cells, their abnormal subtypes, and surrounding patterns (”cytomorphological fingerprint”) in all hematological diseases and healthy subjects; (3) develop algorithms observing both clinical and image data to predict survival, treatment response, and complications; (4) and finally evaluate these algorithms in a clinical setting.

2023

2027