Method for generating a trained machine learning algorithm

US12165314B2 · kind B2 · utility

Assignee

• Siemens Healthineers AG · DE

Inventors

• Daniel Lerch · Ebermannstadt, DE
• Robert Tamas Derzsi · Oberfellendorf, DE

Key dates

• Filing date: Sep 28, 2021
• Grant date: Dec 10, 2024
• Expiry date: Nov 9, 2042

Classification

• Technology area (CPC G): Physics
• CPC primary: G06V2201/03
• WIPO field: Computer technology
• WIPO sector: Electrical engineering

Abstract

At least one embodiment relates to a method for training a machine learning algorithm. In an embodiment, the method includes provisioning a training apparatus with a machine learning algorithm to be trained, having a secure data connection to a medical-engineering imaging system; creating a plurality of training datasets based on image recordings, each training dataset of the plurality of training datasets being furnished with a ground truth or being linked by data link; sending training datasets, of the plurality of training datasets, to the training apparatus via the secure data connection; training the machine learning algorithm via the training apparatus, with the training datasets to create a trained machine learning algorithm; and sending a parameter dataset of the trained machine learning algorithm to the central computing apparatus via the data interface. A corresponding machine learning algorithm, a control facility, an imaging system and a multi-device network are also disclosed.