Collaboration
We are open for possible research collaborations.
Contact UsDEMOInterested in a demonstration of any of our classifiers?
Contact usWe are developing a clinical decision support system (CDSS) that assists doctors in the diagnosis finding process and helps in the selection of the most relevant laboratory measurements to confirm or exclude a suspected diagnosis. The CDSS is based on machine learning models that were trained on 12 years worth of medical data.
Given age, gender, symptoms and blood laboratory results, the software predicts the most likely diagnoses for a patient. The process is adapted to clinical practice and works in two steps:
In the first step, a model (Model A in the illustration) takes 31 basic blood parameters (including relative and absolute values of some measurements) and symptoms to predict a diagnosis category and relevant measurements to disambiguate the diagnoses. This first-stage model currently reaches an average f1-score of 0.90 when considering 7 diagnosis categories.
The diagnostics within each diagnosis group is then handled by a separate specialized model (Model B) that given the initial information and further measurements can suggest specific diagnoses for a patient in the second step. Depending on the category the specialized models range from 0.71 to 0.94 in average f1-score.
We provide access to the models through a password-controlled API upon request.
Based on the initial input this model suggests the most likely disease category and additional laboratory measurements to further narrow the range of possible diagnoses.
Based on the initial input and the additional parameters this model suggests the most likely diagnoses for a patient.
©2025