Skip to content. | Skip to navigation

Python User Groups in Berlin

Personal tools

You are here: Home / Python Users Berlin / February 13, 2020

February 13, 2020

by Veit Schiele last modified Feb 18, 2020 02:03 PM

Prefree –  ML detection algorithm for medical conditions

Prefree, a project focused on building a ML detection algorithm for different medical conditions in pregnancy.
Prefree’s central aim is creating an application that offers a set of functions, on the one hand to physicians and pregnant women for recording their health data (vital parameters etc.) throughout their pregnancy.
This data will be analyzed with machine-learning methods in order to assess the pregnant woman’s individual risk profile and offer treatment recommendations to the physicians. This will reduce the number of misclassifications in the hospital wards and simultaneously offer a higher level of security to pregnant women at risk.
The central part consists of a Django-web-interface which implements the machine-learning algorithm on the server in order to classify the patients according to the results obtained from training. The algorithm itself is developed using the scipy-stack (pandas, sklearn, numpy) and consists of a Random-Forest-Classifier based on a sklearn-learn-interface-compatible machine-learning module named xgboost.
The implemented algorithm is a Random-Forest_Classifier which is suitable for this specific problem of classification based on medical data. The monitoring will consist of a simple application that transmits the user's data via defined endpoints of a REST-API to the server and presents the evaluation to the user.
For deployment, it is planned on simply launching the containerized application using docker and ultimately augmenting it using kubernetes.

Technology

Prefree implements a platform-solution for distributed information- and applicationsprocessing. The central part consists of a Django-web-interface which implements the machine-learning algorithm on the server in order to classify the patients according to the results obtained from training. The algorithm itself is developed using the scipy-stack (pandas, sklearn, numpy) and consists of a Random-Forest-Classifier based on a sklearn-learn-interface-compatible machine-learning module named xgboost. The implemented algorithm is a Random-Forest_Classifier which is suitable for this specific problem of classification based on medical data. The monitoring will consists of a simple application which transmits the users data via defined endpoints of a REST-API to the server and presents the evaluation to the user.
For deployment, we plan on simply launching the containerized application using docker and ultimately augmenting it using kubernetes.

Figure 1: Datastreams Prefree

Figure 1: Datastreams Prefree

Figure 2: Datastream within the application

Cooperation

The project is currently looking for research assistants who can support further development.

Contact Person

Prof. Dr. med. Stefan Verlohren
Phone: +49 30 450 564 294
Contact form