Neural Network Training Corpus for Radiation Treatment Platform Control

ChangeBridge: Patent Apps - Health Informatics (G16H)

Published October 2nd, 2024

Detected April 2nd, 2026

Summary

The USPTO published patent application US20260094688A1 on April 2, 2026, covering neural network training methodology using physical settings data from a specific radiation treatment platform. Inventors Shahab Basiri and Esa Kuusela describe a training corpus that excludes data from other platforms, with retraining triggered by time, usage count, or maintenance events. The trained neural network maps optimized treatment plan control points to corresponding machine control points. No compliance deadlines or penalties apply.

View original document View source feed page

What changed

The USPTO published patent application US20260094688A1 for a neural network training system using physical settings data from a specific radiation treatment platform as a training corpus. The corpus excludes information from other radiation treatment platforms. The training process can be repeated based on passage of time, number of therapeutic uses, or completion of maintenance activities. The trained neural network maps optimized radiation treatment plan control points to corresponding machine control points for platform control.

Healthcare providers and medical device manufacturers involved in radiation therapy systems should review this patent for potential licensing opportunities or competitive assessment. Patent applications do not create immediate compliance obligations but may influence future product development strategies. No deadlines or penalties are associated with this publication.

Source document (simplified)

← USPTO Patent Applications

NEURAL NETWORK TRAINING CORPUS DEVELOPMENT AND USE, AND USE OF A TRAINED NEURAL NETWORK RE RADIATION TREATMENT PLATFORM MACHINE CONTROL POINTS

Application US20260094688A1 Kind: A1 Apr 02, 2026

Inventors

Shahab Basiri, Esa Kuusela

Abstract

Information comprising physical settings for a particular radiation treatment platform (such as machine control points) form a training corpus. A neural network is trained using that training corpus. By one approach, the training corpus does not include any information that pertains to any radiation treatment platform other than the particular radiation treatment platform. The training may be repeated as a function of at least one of a passage of time, a particular number of therapeutic uses of the particular radiation treatment platform, and/or completion of at least one maintenance activity. A radiation treatment plan can be optimized to provide optimized plan control points. These optimized plan control points can be mapped to corresponding machine control points as a function, at least in part, of the aforementioned trained neural network.