Zhou Xinsen, Chen Yi, Heidari Ali Asghar, Chen Huiling, Chen Xiaowei
Department of Computer Science and Artificial Intelligence, Wenzhou University, Wenzhou 325035, China.
School of Surveying and Geospatial Engineering, College of Engineering, University of Tehran, Tehran, Iran.
Artif Intell Med. 2025 Feb;160:103042. doi: 10.1016/j.artmed.2024.103042. Epub 2024 Nov 27.
Systemic lupus erythematosus (SLE) is an autoimmune inflammatory disease. Lupus nephritis (LN) is a major risk factor for morbidity and mortality in SLE. Proliferative and pure membranous LN have different prognoses and may require different treatments. This study proposes a binary rough hypervolume-driven spherical evolution algorithm with groupwise intelligent sampling (bRGSE). The efficient dimensionality reduction capability of the bRGSE is verified across twelve datasets. These datasets are from the public datasets, with feature dimensions ranging from seven hundred to fifty thousand. The experimental results indicate that bRGSE performs better than seven high-performing alternatives. Then, the bRGSE was combined with adaptive boosting (AdaBoost) to form a new model (bRGSE_AdaBoost), which analyzed clinical records collected from 110 patients with LN. Experimental results show that the proposed bRGSE_AdaBoost can identify the most critical indicators, including urine latent blood, white blood cells, endogenous creatinine clearing rate, and age. These indicators may help differentiate between proliferative LN and membranous LN. The proposed bRGSE algorithm is an efficient dimensionality reduction method. The developed bRGSE_AdaBoost model, a computer-aided model, achieved an accuracy of 96.687 % and is expected to provide early warning for the treatment and diagnosis of LN.
系统性红斑狼疮（SLE）是一种自身免疫性炎症性疾病。狼疮性肾炎（LN）是SLE发病和死亡的主要危险因素。增殖性和单纯膜性LN有不同的预后，可能需要不同的治疗方法。本研究提出了一种具有分组智能采样的二元粗糙超体积驱动球形进化算法（bRGSE）。bRGSE的有效降维能力在12个数据集中得到验证。这些数据集来自公共数据集，特征维度从700到50000不等。实验结果表明，bRGSE的性能优于7种高性能替代算法。然后，将bRGSE与自适应增强（AdaBoost）相结合，形成一个新模型（bRGSE_AdaBoost），该模型分析了从110例LN患者收集的临床记录。实验结果表明，所提出的bRGSE_AdaBoost可以识别出最关键的指标，包括尿潜血、白细胞、内生肌酐清除率和年龄。这些指标可能有助于区分增殖性LN和膜性LN。所提出的bRGSE算法是一种有效的降维方法。所开发的bRGSE_AdaBoost模型是一种计算机辅助模型，准确率达到96.687%，有望为LN的治疗和诊断提供早期预警。
