A pancreatic cystic lesion data processing method and system

By using a risk progression matrix and hierarchical risk weights to encode multi-category variables in pancreatic cystic lesion data processing, and combining dynamic reference intervals and Gaussian kernel focusing to encode continuous numerical variables, the problem that traditional coding methods cannot effectively capture risk relationships is solved, thus improving the prediction accuracy and reliability of the model.

CN122177478BActive Publication Date: 2026-07-07TIANJIN TUMOR HOSPITAL
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
TIANJIN TUMOR HOSPITAL
Filing Date
2026-05-13
Publication Date
2026-07-07

AI Technical Summary

Technical Problem

When processing pancreatic cystic lesion data, existing technologies, such as traditional one-hot coding and quantile mapping coding, cannot effectively capture the risk relationships between multi-category variables and continuous numerical variables. This results in high model learning costs, a high risk of overfitting, and an inability to accurately assess risk jumps, thus affecting prediction accuracy.

Method used

Multi-category variables are encoded using a risk progression matrix and hierarchical risk weights, while continuous numerical variables are encoded using a dynamic reference interval and Gaussian kernel focusing method. The most appropriate encoding method is selected for different types of variables, and clinical knowledge is introduced to guide model learning.

Benefits of technology

It improves the model's predictive accuracy and sensitivity in risk assessment, especially when dealing with highly imbalanced or risk-differentiated categories, significantly enhancing the ability to identify high-risk categories and increasing the reliability of the encoding in real-world scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122177478B_ABST
    Figure CN122177478B_ABST
Patent Text Reader

Abstract

The present application relates to the technical field of medical data processing, and particularly relates to a pancreatic cystic lesion data processing method and system.In the present application, when encoding the multi-classification variable, the clinical knowledge is directly injected into the data features through the quantifiable risk progression matrix based on external evidence and the hierarchical risk weight, which is equivalent to providing a learning guide for the artificial intelligence model, so that the model does not need to start from scratch, thereby accelerating the convergence and improving the prediction accuracy; meanwhile, the encoding method provided by the present application has significantly different numerical characteristics, and such differences can be more effectively captured by the model, especially when dealing with highly unbalanced or highly different risk categories, the sensitivity of the model to high-risk categories can be significantly improved.
Need to check novelty before this filing date? Find Prior Art