Amphora AI Pre

Version 1.0.4 (Dec 6,2024)

About

The Amphora-AI Pre (AAP) is a software to automatically predict a recurrence (rehospitalization/readmission in 7 days) in heart failure (HF) patients using AI techniques. Its purpose is to verify integration (input, output, computational functionality) into the cloud service Amphora Cloud in the AMPHORA project, led by the Medical Data Transfer-MDT, s.r.o. private company. The AMPHORA is partially supported by TACR project FW06010766.

Authors: F. Plesinger, Z. Koscova, E. Vargova (AIMT Scientific group at ISI of the CAS, Brno, Czech Republic)

Usage

AAP is a Python command-line application. It is called once per each file to process. Input and output is xml file. The bullet[1] below shows general usage, the bullet[2]shows usage with flag to overwrite existing results.

• 1. main Input.xml Output.xml
• 2. main Input.xml Output.xml --Overwrite

When the AAP is executed, the input file is parsed, preprocessed, and pushed into a machine-learning (AI) model for inference. Output probability as well as computation details is stored in the output file. Output probability is also transformed into color zones/levels, where each of them refer to a different rehospitalization prevalence zones in a test set (green - less or equal to 5%, orange - between 5 and 50%, red - higher or equal than 50%). Optionally, the AAP can output image with SVG showing model probability in context of the test-data. This optional SVG output is controlled by the element GenerateImages in the input XML file. If used, the AAP produces image like this:

Figure: Image output of the Amphora-AI Pre software. Bold vertical is the output probability; colored zones (green, orange, and red) refer to 7-days HF rehospitalization prevalence (≤5%, 5-50%, and >50%).