Volume 3,Issue 7
基于 KAN 增强 YOLOv10的脊柱弯曲异常智能筛查系统研究
针对脊柱弯曲异常传统筛查方式主观性强、漏诊率高、智能化不足等问题，本文以 YOLOv10为基础框架，研究融合 KAN 网络的脊柱侧弯智能筛查协同设计方案。嵌入 KAN 模块可强化细微椎体非线性特征提取能力，一体化结构减少测量误差，轻量化改造实现 CPU 端落地运行。研究成果可为脊柱弯曲异常安全高效智能筛查提供技术参考。
[1]WangA,ChenH,LiuL,etal.YOLOv10:Real-TimeEnd-to-EndObjectDetection[J/OL].arXiv:2405.14458,2024.
[2]SapkotaR,KarkeeM.UltralyticsYOLOEvolution:AnOverviewofYOLO26,YOLO11,YOLOv8,andYOLOv5ObjectDetectorsforComputerVisionandPatternRecognition[EB/OL]. (2025-10-01).https://arxiv.org/pdf/2510.09653v2.
[3]RedmonJ,FarhadiA.YOLOv3:AnIncrementalImprovement[J/OL].arXiv:1804.02767,2018.
[4]BochkovskiyA,WangCY,LiaoHYM.YOLOv4:OptimalSpeedandAccuracyofObjectDetection[J/OL].arXiv:2004.10934,2020.
[5]LiuZ,WangY,VaidyaS,etal.KAN:Kolmogorov-ArnoldNetworks[J/OL].arXiv:2404.19756,2024.
[6]DrokinI.Kolmogorov-ArnoldConvolutions:DesignPrinciplesandEmpiricalStudies[EB/OL].(2024-07-02).https://arxiv.org/pdf/2407.01092.
[7]SomvanshiS,JavedSA,IslamMMandothers.ASurveyonKolmogorov-ArnoldNetwork[EB/OL].(2024-11-08).https://arxiv.org/pdf/2411.06078.
[8]PANYL,CHENQR,CHENTTandothers.Evaluationofacomputer-aidedmethodformeasuringtheCobbangleonchestX-rays[J].EuropeanSpineJournal,2019,28:3035-3043.
[9]KONIECZNYMR,SENYURTH,KRAUSPER.Epidemiologyofadolescentidiopathicscoliosis[J].JournalofChildren'sOrthopaedics,2013,7(1):3-9.
[10]Bernstein P, MetzlerJ, Weinzierl M, etal. Radio graphic Scoliosis AngleEstimation:Spline-BasedMeasurementRevealsSuperiorReliabilityComparedtoTraditionalCobbMethod[J].European Spine Journal, 2020, 29(12):3033-3042.