Non-small cell lung cancer auxiliary decision-making method based on multi-modal causal representation

By extracting multimodal data features from non-small cell lung cancer patients through rotational position encoding, multi-level gating aggregation, and graph attention mechanisms, and combining random causal relationship networks and pessimistic estimation mechanisms, the semantic gap and causal interpretation problems of multimodal data are solved, enabling accurate decision support and safe treatment recommendations.

CN122158055APending Publication Date: 2026-06-05YANGTZE DELTA REGION INST (QUZHOU) UNIV OF ELECTRONIC SCI & TECH OF CHINA
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN Β· China
Patent Type
Applications(China)
Current Assignee / Owner
YANGTZE DELTA REGION INST (QUZHOU) UNIV OF ELECTRONIC SCI & TECH OF CHINA
Filing Date
2026-01-23
Publication Date
2026-06-05

AI Technical Summary

Technical Problem

Existing decision support technologies for non-small cell lung cancer face challenges such as the semantic gap in multimodal data, lack of causal reasoning ability, and lack of quantification mechanisms for unknown risks. These limitations prevent the construction of accurate holistic tumor profiles for patients and can easily lead to incorrect decisions.

Method used

We employ rotational position encoding and alternating attention mechanisms to extract textual features, multi-level gating aggregation to extract image features, and graph attention mechanisms to extract gene features. We also combine random causal relationship networks and pessimistic estimation mechanisms to generate auxiliary decisions, thereby achieving unified semantic alignment and causal interpretation of multimodal data.

Benefits of technology

It achieves efficient representation of multimodal data, can accurately assess the causal effects of treatment plans, reduce the risk of erroneous decisions, meet the requirements of evidence-based medicine, and reduce the number of model parameters and computational energy consumption.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122158055A_ABST
    Figure CN122158055A_ABST
Patent Text Reader

Abstract

The application discloses a non-small cell lung cancer auxiliary decision-making method based on multi-modal causal representation and belongs to the field of clinical auxiliary decision-making. The method comprises the following steps: S1, obtaining clinical texts of non-small cell lung cancer patients, inputting the texts into a text feature extraction network, and obtaining text features; S2, obtaining imaging examination data of the non-small cell lung cancer patients, inputting the data into an image feature extraction network, and obtaining image features; S3, obtaining genomic data of the non-small cell lung cancer patients, inputting the data into an omics feature extraction network, and obtaining gene features; S4, mapping and splicing the text features, the image features and the gene features to obtain mixed variable representation, inputting the mixed variable representation into a random causal relationship network, and combining a pessimistic estimation mechanism to generate an auxiliary decision. Through the introduction of long text analysis of rotary position coding, lesion image extraction of multi-order gating aggregation and gene pathway analysis of graph attention mechanism, efficient representation of multi-modal data is realized.
Need to check novelty before this filing date? Find Prior Art