A system and device for constructing a prediction index of neoadjuvant efficacy of malignant tumors

By using an improved deep convolutional neural network and multi-region analysis, a predictive index system for neoadjuvant efficacy assessment of malignant tumors was constructed. This system enables accurate identification and combined feature prediction of tumor, necrosis, and lymphatic infiltration areas, solving the problems of subjectivity and insufficient quantitative accuracy in existing assessment methods and providing accurate efficacy prediction support.

CN122436244APending Publication Date: 2026-07-21NANCHANG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
NANCHANG UNIV
Filing Date
2026-06-24
Publication Date
2026-07-21

AI Technical Summary

Technical Problem

Existing methods for assessing neoadjuvant efficacy rely on manual interpretation, which suffers from significant observer variability, insufficient quantitative precision, and neglect of the prognostic value of multi-regional combinations, making it difficult to meet the clinical need for accurate prediction of efficacy in malignant tumors.

Method used

An improved deep convolutional neural network is used for feature extraction. Combined with multi-region classification and binarization modules, an 11-dimensional structured index system is constructed. Through parallel prediction of multiple models, accurate identification and combined feature prediction of tumor, necrosis and lymphatic infiltration areas are achieved.

Benefits of technology

It provides an objective, accurate, and interpretable assessment tool that can systematically analyze the spatial combination patterns of tumor, necrosis, and lymphatic infiltration areas, supporting the efficacy assessment after neoadjuvant therapy and overcoming the subjectivity and limitations of traditional assessment methods.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122436244A_ABST
    Figure CN122436244A_ABST
Patent Text Reader

Abstract

The application discloses a system and device for constructing a prediction index of neoadjuvant efficacy of malignant tumors. The method comprises: acquiring a pathological image, synchronously identifying tumor regions, necrotic regions and lymphatic infiltration regions therein, and binarizing the three regions into three-dimensional logical variables based on an area proportion threshold; constructing an 11-item structured index system composed of 8 combined features and 3 independent region features according to the three-dimensional logical variables, wherein the 8 combined features are generated by full permutation logic and are used to represent the existence mode of the tumor, necrotic and lymphatic infiltration regions, and the 3 independent region features are used to reflect the global existence of a single pathological element; screening the 11 features and combining a prediction module to evaluate the residual tumor load grade. The application integrates multi-region information, solves the problem that the existing single index is difficult to quantize the complex interaction of the tumor microenvironment, and improves the prediction performance and model interpretability.
Need to check novelty before this filing date? Find Prior Art