Portfolio

Our portfolio encompasses different projects like pattern recognition in billing systems, recommender systems in medicine, as well as R&D in AI explainability and novel AI methods.

Pattern Recognition in Billing Systems

Pattern recognition in billing systems and automatic discovery of deviations in accounting, helping to control how the billing is performed as well as to make projections of future spending.

Recommender Systems

We help physicians to learn from other colleagues which are the best suited medical procedures for the patients.

  • Reduce the workload and get a better overview among an ever complex amount of alternative procedures
  • Improve the quality of the patient’s therapy
  • Improve billing process
  • Reduce costs by discovering alternative and effective procedures

Apps Based on Descentralized Architectures

Applications that store the information directly in the preferred customer’s data facility, without depending on central platforms and with high data protection standards.

  • No data fragmentation, as the dara resides with the customers and not on the platform
  • High flexibility in the app selection and design for a personalized user experience

In our R&D program we are discovering and developing novel methods in artificial intelligence, as well as on mathematical modelling and systems biology / systems medicine.

From Personalization to Patient Centered Systems Toxicology and Pharmacology

The current paradigm of precision and personalization will evolve to a novel cyclic paradigm, where patient therapies are adjusted in short time cycles thanks to the availability of ubiquitous data. Naturally, this concept aims to improve the quality of the patient’s therapies. We describe here how it could work.

Simulation of Human Cognitive Processes for AI Explainability

How to emulate rational process for decision making in medicine? We present in this work a method to simulate logical processes integrated in deep learning methods, also considering uncertainty in the data and the model output, for safe and explainable recommender systems in medicine.

Help to Reduce the Amount of Data Required for Training

The use of large amounts of data is necessary for good deep learning models. However, large amounts of data require a lot of resources (energy and storage size) and could contain persistent distortions that affect the quality of the models. In this work, we explain how to identify patterns using mathematical methods (persistent topology) to identify persistent defects in time series and evaluate which data can be used for model training. 

Graph Neural Network Modelling for Procedures based on Patient Diagnoses

In this work define a patient Graph structure based on basic patient’s information like age and gender as well as the diagnoses and trained GNNs models to identify the corresponding patient’s procedures. This method is not only more accurate than other conventional baseline models but is also helpful to identify patient clusters and make a detailed analysis of the diagnose hierarchy leading to each procedure.