Training Device and Attribute Data Generation Device Using AI for Health Analytics

ChangeBridge: Patent Apps - Health Informatics (G16H)

Published April 2nd, 2026

Detected April 2nd, 2026

Summary

USPTO published patent application US20260093784A1 by NEC Corporation disclosing an AI-based training device and attribute data generation system for health analytics. The system uses machine learning encoders and decoders with latent variable clustering to generate health attribute data for disease risk assessment and activity decision support. No regulatory obligations or compliance requirements are created.

View original document View source feed page

What changed

USPTO published patent application US20260093784A1 (Application No. 19332472, filed September 18, 2025) disclosing technology developed by NEC Corporation inventors Chenhui Huang, Kensuke Wagata, and Fumiyuki Nihey. The invention comprises a training device using AI/machine learning models to train an attribute data generation system. The system includes acquisition means for first and second attribute data, a first encoder converting data to stochastic latent variables, a second encoder projecting to latent space with clustering and centroid output, and a decoder for reconstruction. Optimization means train the encoders and decoder based on cluster relationships.

This is a published patent application with no regulatory or compliance implications. No action is required by regulated entities. The document represents prior art disclosure for the healthcare AI/ML technology sector, specifically classified under CPC G06F 18/2415 and G16H 50/30 (health informatics). Companies developing similar health data generation or disease risk assessment systems should consider this publication for freedom-to-operate analysis.

Source document (simplified)

← USPTO Patent Applications

TRAINING DEVICE, TRAINING METHOD, ATTRIBUTE DATA GENERATION DEVICE, ATTRIBUTE DATA GENERATION METHOD, AND PROGRAM

Application US20260093784A1 Kind: A1 Apr 02, 2026

Assignee

NEC Corporation

Inventors

Chenhui HUANG, Kensuke WAGATA, Fumiyuki NIHEY

Abstract

In order to generate insufficient attribute data by existing attribute data, a training device uses AI or machine learning models to train an attribute data generation device. An acquisition means acquires first and second attribute data other than the first attribute data. A first encoder converts the second attribute data into a stochastic latent variable. A second encoder projects the stochastic latent variable to a latent space, clusters projection points into clusters, and outputs centroids indicating centers of gravity of the clusters. A decoder reconstructs the second attribute data based on the projection points. An optimization means optimizes the first and second encoders, and the decoder based on relationships between the projection points and the centers of gravity and a relationship between the clusters. An analysis result of a health condition and a disease risk using the attribute data is used for supporting a decision making regarding a subject's activity.