基于特征分析与优化技术的狼疮肾炎辅助诊断  

Auxiliary Diagnosis of Lupus Nephritis Based on Feature Analysis and Optimization Technology

作　　者：赵琳 石艳平 潘章磊 薛健 张浩 郝凡昌 ZHAO Lin;SHI Yanping;PAN Zhanglei;XUE Jian;ZHANG Hao;HAO Fanchang(School of Computer Science and Technology,Shandong Jianzhu University,Ji'nan 250101,China;The Second Affiliated Hospital of Shandong First Medical University,Tai'an 271099,China;Zibo General Research Institute of Inspection and Measurement,Zibo 255020,China)

机构地区：[1]山东建筑大学计算机科学与技术学院,山东济南250101 [2]山东第一医科大学第二附属医院,山东泰安271099 [3]淄博市检验检测计量研究总院,山东淄博255020

出　　处：《软件导刊》2025年第2期114-120,共7页Software Guide

基　　金：山东省自然科学基金面上项目(ZR2022MF272);泰安市科技创新发展项目(政策引导类)(2022NS194);山东省重点研发计划资助项目(2019GGX101068)。

摘　　要：狼疮肾炎的早期诊断对于患者的治疗和预后至关重要。目前传统诊断方式依赖于医生的临床经验,为提高诊断效率与准确性,提出一种基于特征分析与优化技术的狼疮肾炎辅助诊断模型。为对狼疮肾炎患者病情的进展程度进行分级,医学专家依据临床指标将485名患者标注成轻度、中度、重度三大类。针对临床指标冗余问题,提出一种基于调整余弦相似度的特征选择方法FSMACS,以提升模型检测效率。针对标注后数据类别不平衡问题,提出一种基于个体贝叶斯不平衡影响指数的过采样方法IBOA来减少分类模型误差。实验结果表明,经过FSMACS和IBOA优化的模型在各种常规经典分类器上均表现良好,在使用Adaboost分类时的精确度、召回率、F1值和几何平均值分别达到89.6%、89.4%、93.4%、92.1%,为狼疮肾炎患者的辅助诊断提供了一种高效准确的方法。Early diagnosis of lupus nephritis is crucial for the treatment and prognosis of patients.At present,traditional diagnostic methods rely on the clinical experience of doctors.In order to improve diagnostic efficiency and accuracy,a lupus nephritis auxiliary diagnostic model based on feature analysis and optimization techniques is proposed.To grade the progression of lupus nephritis,medical experts classified 485 patients as mild,moderate or severe based on clinical indicators.A feature selection method FSMACS based on adjusting cosine similarity is proposed to address the issue of redundant clinical indicators and improve model detection efficiency.Aiming at the problem of imbalanced data categories after annotation,a oversampling method called IBOA based on individual Bayesian imbalance impact index is proposed to reduce classification model errors.The experimental results show that the model optimized by FSMACS and IBOA performs well on various conventional classical classifiers,with accuracy,recall,F1 score,and geometric mean of 89.6%,89.4%,93.4%,and 92.1%,respectively,when using Adaboost for classification.This provides an efficient and accurate method for the auxiliary diagnosis of lupus nephritis patients.

关 键 词：狼疮肾炎 辅助诊断 特征选择 类别不平衡 分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]